CN109214420A - The high texture image classification method and system of view-based access control model conspicuousness detection - Google Patents

The high texture image classification method and system of view-based access control model conspicuousness detection Download PDF

Info

Publication number
CN109214420A
CN109214420A CN201810845341.7A CN201810845341A CN109214420A CN 109214420 A CN109214420 A CN 109214420A CN 201810845341 A CN201810845341 A CN 201810845341A CN 109214420 A CN109214420 A CN 109214420A
Authority
CN
China
Prior art keywords
color
image
texture
gradient
saliency
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN201810845341.7A
Other languages
Chinese (zh)
Inventor
王瑜
陈肖蒙
李长胜
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Technology and Business University
Original Assignee
Beijing Technology and Business University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing Technology and Business University filed Critical Beijing Technology and Business University
Priority to CN201810845341.7A priority Critical patent/CN109214420A/en
Publication of CN109214420A publication Critical patent/CN109214420A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/56—Extraction of image or video features relating to colour

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Multimedia (AREA)
  • Image Analysis (AREA)

Abstract

本发明公开了一种基于视觉显著性检测的高纹理图像分类方法及系统，其中，方法包括：通过颜色对比度的图像像素显著性值检测方法分割出显著性区域；通过完备局部二值模式算子和方向梯度直方图算法提取显著性区域的纹理特征和梯度特征，并通过有效串联融合策略共同表示图像细节信息；以及通过最近邻分类器对提取到的融合向量进行分类，以获得识别率。该方法利用拟合人类视觉机制的视觉注意模型来定位图像中的显著目标区域，以摒除背景冗余信息，对高纹理图像进行主体分割，再对该显著区域进行特征提取并完成图像分类，从而实现了自动将图像(视野)中各种不同的纹理区分开，使得随后的场景描述或目标识别成为可能。

The invention discloses a high-texture image classification method and system based on visual saliency detection, wherein the method comprises: segmenting saliency regions through a color contrast image pixel saliency value detection method; using a complete local binary pattern operator And the directional gradient histogram algorithm extracts the texture features and gradient features of the salient regions, and jointly represents the image details through an effective concatenated fusion strategy; and the extracted fusion vectors are classified by the nearest neighbor classifier to obtain the recognition rate. This method uses the visual attention model that fits the human visual mechanism to locate the salient target area in the image, so as to eliminate the background redundant information, perform subject segmentation on the high-texture image, and then perform feature extraction on the salient area and complete the image classification. Automatically distinguish various textures in the image (field of view), making subsequent scene description or object recognition possible.

Description

High-texture image classification method and system based on visual saliency detection
Technical Field
The invention relates to the technical field of image processing and computer vision, in particular to a high-texture image classification method and system based on visual saliency detection.
Background
The essence of digital image processing is image texture analysis, and in the visual perception of the human visual system, the characteristics of texture mainly include roughness, directionality, contrast and regularity. Because natural textures are numerous and diverse in type, form and structure, and on the other hand, the mechanism of perceiving the textures by a human visual system is not well known, so for digital image processing, how to extract numerical features capable of effectively describing the textures is always the key point and the difficulty of texture research. Texture features can be broadly divided into: the four categories of the statistical-based feature, the template convolution-based feature, the frequency-domain-based feature and the model-based feature are summarized as the following ten texture features: uniformity, density, thickness, roughness, regularity, linearity, directionality, frequency, and phase of the texture.
Because the textures existing in the image are distributed compactly and have various types, the whole image is analyzed and identified directly through texture feature description, and the complexity is higher. If only the target area can be processed, the analysis complexity can be reduced and the identification accuracy can be improved. The core of the texture image segmentation research, which is to distinguish the texture from other textures, is to make the calculation like a human being, and automatically distinguish different textures in an image (view field), so as to enable the subsequent scene description or target identification. Research into this subject has been conducted for over thirty years, with hundreds of various texture descriptions and segmentation methods proposed in the meantime. Unfortunately, the above-mentioned goals have so far not been achieved, and texture image segmentation remains a problem that is not well solved. The main difficulties faced by the process of texture image segmentation by partitioning texture features can be summarized as: efficiency and effectiveness. From the aspect of efficiency, most texture image segmentation algorithms have higher time complexity, which is because on one hand, extraction of texture features takes time, and on the other hand, the higher feature dimension causes the calculation amount of the feature division process to be generally larger; in effect, texture is a regional characteristic, and visual experience shows that the texture in an image must be reflected in a certain size region. Psychological studies have also demonstrated that human perception of texture is performed simultaneously over a relatively large area. Therefore, the texture features must also be extracted within a certain area. When a region contains a plurality of textures, the obtained feature is a 'four-way difference', so that the division error of different textures in the same image by extracting texture features is large.
Research finds that when a complex scene is observed by a human complex system, attention can be rapidly focused on certain salient targets and the salient targets are preferentially processed, namely a visual attention mechanism, and the determination of the salient targets by human eyes is usually from color differences between a target area and a background, so that the idea that a main body segmentation is carried out on a high-texture image by using a visual saliency detection method based on color contrast, and then the extraction of texture features of the target area is carried out for analysis is feasible.
Disclosure of Invention
The present invention is directed to solving, at least to some extent, one of the technical problems in the related art.
To this end, it is an object of the present invention to propose a high-texture image classification method based on visual saliency detection, which can automatically distinguish various textures in an image (field of view) so that subsequent scene description or object recognition is possible.
Another object of the present invention is to provide a high-texture image classification apparatus based on visual saliency detection.
In order to achieve the above object, an embodiment of an aspect of the present invention provides a high-texture image classification method based on visual saliency detection, including the following steps: segmenting a saliency region by an image pixel saliency value detection method of color contrast; extracting texture features and gradient features of the salient region through a complete local binary pattern operator and a direction gradient histogram algorithm, and jointly representing image detail information through an effective series fusion strategy; and classifying the extracted fusion vector through a nearest neighbor classifier to obtain the recognition rate.
According to the high-texture image classification method based on visual saliency detection, a saliency region is segmented by using an image pixel saliency value detection method based on color contrast, then the texture and gradient features of the saliency region are extracted by using a complete local binary pattern algorithm and a direction gradient histogram algorithm, an effective series fusion strategy is carried out, image detail information is jointly represented, and finally a nearest neighbor classifier is used for classifying the extracted texture feature vectors to obtain an identification rate, so that the purpose that a computer automatically classifies the extracted textures is achieved.
In addition, the high-texture image classification method based on visual saliency detection according to the above embodiment of the present invention may also have the following additional technical features:
further, in an embodiment of the present invention, the method for detecting a saliency value of an image pixel through color contrast segments a saliency region, further comprising: quantizing the color space to obtain a set of representative colors; acquiring the occurrence frequency of colors corresponding to the representative colors in an input image, and forming a histogram; calculating a significance value of the representative color according to the difference between each representative color and other representative colors; and assigning the significance value of each representative color to the corresponding pixel.
Further, in an embodiment of the present invention, the color space quantization method is octree color quantization, wherein the octree color quantization is divided into establishing a color octree, generating a color palette, generating a quantization file, sequentially reading in pixel colors, establishing a color octree with a leaf node smaller than the number of quantized colors, and traversing the color octree, if any color in an image does not exist in the color octree, newly inserting a leaf node to represent the any color, and if the number of leaf nodes of the color octree exceeds the number of quantized colors after inserting a pixel color, performing a merging operation of the leaf nodes according to a merging strategy, so that only colors not exceeding the number of quantized colors are saved as the color palette after all pixels are inserted, and scanning the file again to map each color to the color palette, a new image after quantization is generated.
Further, in an embodiment of the present invention, the complete local Binary pattern CLBP (complete local Binary pattern) algorithm is used to extract texture features of the salient region, the complete Binary pattern describes texture features of pixel points from a gray-value magnitude relation feature CLBP _ S (CLBP-Sign), a gray-value difference magnitude feature CLBP _ M (CLBP-Magnitudes), and a pixel gray-value and global average gray-value magnitude relation feature CLBP _ C (CLBP-Center), so as to maximally extract image gray texture information of a single pixel point, and mathematical description of the complete local Binary pattern features is as follows:
wherein, gi(i-1, 2, …, N) denotes the number gcThe gray value of the neighborhood pixel point as the center, R is the neighborhood radius, mNRepresenting the difference value between the central pixel point and the neighborhood pixel point, and c represents m in the local imageNAverage value of clRepresenting a global gray mean.
Further, in an embodiment of the present invention, the gradient feature of the saliency region is extracted through the directional gradient histogram algorithm hog (histogram of ordered gradient), and a formula for calculating the gradient of the pixel point by the directional gradient histogram algorithm is as follows:
Gx(x,y)＝H(x+1,y)-H(x-1,y)，
Gy(x,y)＝H(x,y+1)-H(x,y-1)，
wherein, G in the formulax(x, y), Gy (x, y), H (x, y) respectively represent the horizontal direction gradient, the vertical direction gradient and the pixel value at the pixel point (x, y) in the input image;
the gradient amplitude and the gradient direction at the pixel point are respectively as follows:
further, in one embodiment of the present invention, the texture feature vector and the gradient feature vector are fused in series to form the fused feature vector, and the formula is as follows:
H＝[Hclbp,Hhog]，
wherein, Hclbp，HhogRespectively representing a fusion feature vector, a CLBP feature vector and a HOG feature vector.
Further, in an embodiment of the present invention, the classifying the extracted fusion vector by the nearest neighbor classifier further includes: calculating, by the nearest neighbor classifier, a similarity and a dissimilarity between the two histograms.
In order to achieve the above object, another embodiment of the present invention provides a high-texture image classification system based on visual saliency detection, including: the detection module is used for segmenting a saliency region according to an image pixel saliency value detection method of the color contrast; the extraction module is used for extracting the texture features and the gradient features of the salient region through a complete local binary pattern operator and a direction gradient histogram algorithm, and jointly representing image detail information through an effective series fusion strategy; and the classification module is used for classifying the extracted fusion vector through a nearest neighbor classifier so as to obtain the recognition rate.
According to the high-texture image classification system based on visual saliency detection, a saliency region is segmented by using an image pixel saliency value detection method based on color contrast, then the texture and gradient features of the saliency region are extracted by using a complete local binary pattern algorithm and a direction gradient histogram algorithm, an effective series fusion strategy is carried out, image detail information is jointly represented, and finally a nearest neighbor classifier is used for classifying the extracted texture feature vectors to obtain an identification rate, so that the purpose that a computer automatically classifies the extracted textures is achieved.
In addition, the high-texture image classification system based on visual saliency detection according to the above embodiment of the present invention may also have the following additional technical features:
further, in an embodiment of the present invention, the complete local binary pattern algorithm is used to extract the texture features of the significant region, the complete binary pattern describes the texture features of the pixel points from the gray value magnitude relation feature CLBP _ S, the gray value difference amplitude feature CLBP _ M, and the magnitude relation feature CLBP _ C between the pixel gray value and the global average gray value, the image gray texture information of a single pixel point is maximally extracted, and the mathematical description of the complete local binary pattern features is as follows:
wherein, gi(i-1, 2, …, N) denotes the number gcThe gray value of the neighborhood pixel point as the center, R is the neighborhood radius, mNRepresenting the difference value between the central pixel point and the neighborhood pixel point, and c represents m in the local imageNAverage value of clRepresenting a global gray mean.
Further, in an embodiment of the present invention, the gradient feature of the significant region is extracted through the direction gradient histogram algorithm HOG, and a formula for calculating a gradient of a pixel point by the direction gradient histogram algorithm is as follows:
Gx(x,y)＝H(x+1,y)-H(x-1,y)，
Gy(x,y)＝H(x,y+1)-H(x,y-1)，
wherein, G in the formulax(x, y), Gy (x, y), H (x, y) respectively represent the horizontal direction gradient, the vertical direction gradient and the pixel value at the pixel point (x, y) in the input image;
the gradient amplitude and the gradient direction at the pixel point are respectively as follows:
additional aspects and advantages of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
The foregoing and/or additional aspects and advantages of the present invention will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:
FIG. 1 is a flow diagram of a method for high-texture image classification based on visual saliency detection according to one embodiment of the present invention;
FIG. 2 is a flow diagram of visual saliency detection according to one embodiment of the present invention;
FIG. 3 is a flow diagram of octree color quantization according to one embodiment of the invention;
fig. 4 is a schematic diagram of a significance detection result of a plant picture according to an embodiment of the present invention, wherein (a) is an original picture and (b) is a picture after significance detection;
FIG. 5 is a flow diagram of high-texture image classification according to one embodiment of the invention;
fig. 6 is a schematic structural diagram of a high-texture image classification system based on visual saliency detection according to an embodiment of the present invention.
Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the drawings are illustrative and intended to be illustrative of the invention and are not to be construed as limiting the invention.
The high-texture image classification method and system based on visual saliency detection proposed according to an embodiment of the present invention will be described below with reference to the accompanying drawings, and first, the high-texture image classification method based on visual saliency detection proposed according to an embodiment of the present invention will be described with reference to the accompanying drawings.
Fig. 1 is a flowchart of a high-texture image classification method based on visual saliency detection according to an embodiment of the present invention.
As shown in fig. 1, the high-texture image classification method based on visual saliency detection according to the embodiment of the present invention includes the following steps:
in step S101, a saliency region is segmented by an image pixel saliency value detection method of color contrast.
Further, in an embodiment of the present invention, segmenting out the saliency region by the image pixel saliency value detection method of the color contrast, as shown in fig. 2, may further include: quantizing the color space to obtain a set of representative colors; acquiring the occurrence frequency of colors corresponding to the representative colors in the input image, and forming a histogram; calculating a significance value of the representative color according to the difference between each representative color and other representative colors; the saliency value of each representative color is assigned to the corresponding pixel.
Further, in one embodiment of the present invention, the color space quantization method is octree color quantization, wherein, the octree color quantization comprises establishing color octree, generating palette and generating quantization file, reading pixel color in sequence, establishing a color octree with leaf node smaller than quantized color number, traversing the color octree, if any color in the image does not exist in the color octree, a leaf node is newly inserted to represent any color, and if the number of leaf nodes of the color octree exceeds the number of quantized colors after the pixel color is inserted, merging the leaf nodes according to a merging strategy so that only a plurality of colors which do not exceed the quantized colors are saved as a color palette after all the pixels are inserted, and the file is scanned again to map each color to the palette to produce a new quantized image.
Specifically, as shown in fig. 3, quantization is performed on the color space using an octree structure to obtain a set of representative colors, pixel colors are sequentially read in (R, G, B), a color octree with leaf nodes smaller than 256 (the number of colors after quantization) is created, then the color octree is traversed, and if a certain color in the image does not exist in the octree, a leaf node is newly inserted to represent the color, so as to remove the influence of the repeated colors. If the number of leaf nodes of the octree exceeds 256 after the pixel colors are inserted, merging operation of the leaf nodes is carried out according to a certain merging strategy, therefore, after all the pixels are inserted, only less than 256 colors are saved as a color palette, finally, the file is scanned again, each color is mapped to the color palette, and a quantized new image is generated.
And calculating the appearance frequency of the color corresponding to the representative color in the input image to form a histogram, wherein the proportion of the pixel corresponding to a certain representative color in all the pixels in the input image is the appearance frequency of the representative color. Each representative color has a frequency. The frequency of occurrence of this set of representative colors is called a histogram. In order to save computing resources, a representative color with a high frequency of appearance is usually reserved, and the frequencies of appearance of the remaining representative colors are added to the frequencies of appearance of the reserved representative colors with the closest color. When selecting the representative color with a high frequency of appearance, the frequency of appearance of the representative color is sorted from large to small. A representative color is then selected from front to back that is sufficient to cover a certain proportion of the image pixels. This proportion is usually chosen to be 95% in the experiments.
Calculating the significance value S of the representative color according to the difference between each representative color and other representative colors, wherein the specific calculation formula is as follows:
wherein, cjTo remove clOther representative colors than fjIs cjFrequency of occurrence of D (c)l,cj) Is cl、cjEuclidean distance in color space.
For each representative color, its saliency value is assigned to the corresponding pixel.
For example, as shown in fig. 4, it is a schematic diagram of the results of the experiment performed on the plant picture by using the visual saliency detection method.
In step S102, texture features and gradient features of the salient region are extracted through a complete local binary pattern operator and a directional gradient histogram algorithm, and image detail information is jointly represented through an effective series fusion strategy.
Further, in an embodiment of the present invention, a complete local binary pattern algorithm is used to extract texture features of a significant region, where the complete binary pattern describes texture features of pixel points from a gray value magnitude relation feature CLBP _ S, a gray value difference amplitude feature CLBP _ M, and a magnitude relation feature CLBP _ C between a pixel gray value and a global average gray value, so as to maximally extract image gray texture information of a single pixel point, and mathematical description of the complete local binary pattern features is as follows:
wherein, gi(i-1, 2, …, N) denotes the number gcThe gray value of the neighborhood pixel point as the center, R is the neighborhood radius, mNRepresenting the difference value between the central pixel point and the neighborhood pixel point, and c represents m in the local imageNAverage value of clRepresenting a global gray mean.
Further, in an embodiment of the present invention, the gradient feature of the significant region is extracted by a direction gradient histogram algorithm, and a formula for calculating the gradient of the pixel point by the direction gradient histogram algorithm is as follows:
Gx(x,y)＝H(x+1,y)-H(x-1,y)，
Gy(x,y)＝H(x,y+1)-H(x,y-1)，
wherein, G in the formulax(x, y), Gy (x, y), H (x, y) respectively represent the horizontal direction gradient, the vertical direction gradient and the pixel value at the pixel point (x, y) in the input image;
the gradient magnitude and gradient direction at the pixel point are respectively:
further, in one embodiment of the present invention, the texture feature vector and the gradient feature vector are fused in series to form a fused feature vector, and the formula is as follows:
H＝[Hclbp,Hhog]，
wherein, Hclbp，HhogRespectively representing a fusion feature vector, a CLBP feature vector and a HOG feature vector.
Specifically, as shown in fig. 5, a complete binary pattern algorithm is used to extract the salient region texture features. The complete binary pattern describes texture features of pixel points from the gray value magnitude relation feature CLBP _ S, the gray value difference amplitude feature CLBP _ M and the pixel gray value and global average gray value magnitude relation feature CLBP _ C, and image gray texture information of a single pixel point is extracted to the maximum extent. The mathematical description of the CLBP features is as follows:
wherein gi (i ═ 1,2, …, N) denotes gcThe gray value of the neighborhood pixel point which is the center; r is the neighborhood radius; m isNIndicating a large difference between the central pixel and the neighborhood pixelSmall; c represents m in the partial imageNThe mean value of (a); c. ClRepresenting a global gray mean.
And (5) extracting the gradient characteristics of the salient region by using an HOG algorithm. The formula for calculating the gradient of the pixel point (x, y) by the HOG algorithm is as follows:
Gx(x,y)＝H(x+1,y)-H(x-1,y)
Gy(x,y)＝H(x,y+1)-H(x,y-1)
in the formula Gx(x, y), Gy (x, y), H (x, y) respectively represent the horizontal direction gradient, the vertical direction gradient, and the pixel value at the pixel point (x, y) in the input image. The gradient amplitude and gradient direction at pixel point (x, y) are respectively:
then extracting a CLBP characteristic histogram HclbpAnd HOG feature histogram HhogAnd performing serial connection to form a fusion feature vector H, wherein the formula is as follows:
H＝[Hclbp,Hhog]
in step S103, the extracted fusion vector is classified by the nearest neighbor classifier to obtain a recognition rate.
Further, in an embodiment of the present invention, classifying the extracted fusion vector by a nearest neighbor classifier may further include: the similarity and difference between the two histograms is calculated by the nearest neighbor classifier.
That is, nearest neighbor is a simple and effective classification criterion, and the similarity and difference between two histograms can be quickly calculated by a nearest neighbor classifier, for example, the most common euclidean distance, mahalanobis distance. The mahalanobis distance was used as a measure in the experiments described herein and is shown below:
wherein,is a feature vector, H ', of the kth class of images in the training sample set'testD is the Mahalanobis distance between two characteristic vectors for the characteristic vectors of the image to be identified in the test sample set.
According to the high-texture image classification method based on visual saliency detection provided by the embodiment of the invention, a saliency region is segmented by using an image pixel saliency value detection method based on color contrast, then the texture and gradient features of the saliency region are extracted by using a complete local binary pattern algorithm and a direction gradient histogram algorithm, an effective series fusion strategy is carried out, image detail information is jointly represented, and finally the extracted texture feature vectors are classified by using a nearest neighbor classifier to obtain an identification rate, so that the purpose of automatically classifying the extracted texture by using a computer is realized.
Next, a high-texture image classification system based on visual saliency detection proposed according to an embodiment of the present invention is described with reference to the drawings.
Fig. 6 is a schematic structural diagram of a high-texture image classification system based on visual saliency detection according to an embodiment of the present invention.
As shown in fig. 6, the high-texture image classification system 10 based on visual saliency detection includes: a detection module 100, an extraction module 200 and a classification module 300.
The detection module 100 is configured to segment a saliency region according to an image pixel saliency value detection method of color contrast. The extraction module 200 is configured to extract texture features and gradient features of the salient region through a complete local binary pattern operator and a directional gradient histogram algorithm, and collectively represent image detail information through an effective series fusion strategy. The classification module 300 is configured to classify the extracted fusion vector by a nearest neighbor classifier to obtain a recognition rate. The system 10 of embodiments of the present invention enables automatic differentiation of various textures in an image (field of view) so that subsequent scene description or object recognition is possible.
Further, in an embodiment of the present invention, a complete local binary pattern algorithm is used to extract texture features of a significant region, where the complete binary pattern describes texture features of pixel points from a gray value magnitude relation feature CLBP _ S, a gray value difference amplitude feature CLBP _ M, and a magnitude relation feature CLBP _ C between a pixel gray value and a global average gray value, so as to maximally extract image gray texture information of a single pixel point, and mathematical description of the complete local binary pattern features is as follows:
wherein, gi(i-1, 2, …, N) denotes the number gcThe gray value of the neighborhood pixel point as the center, R is the neighborhood radius, mNRepresenting the difference value between the central pixel point and the neighborhood pixel point, and c represents m in the local imageNAverage value of clRepresenting a global gray mean.
Further, in an embodiment of the present invention, the gradient feature of the significant region is extracted by a direction gradient histogram algorithm, and a formula for calculating the gradient of the pixel point by the direction gradient histogram algorithm is as follows:
Gx(x,y)＝H(x+1,y)-H(x-1,y)，
Gy(x,y)＝H(x,y+1)-H(x,y-1)，
wherein, G in the formulax(x, y), Gy (x, y), H (x, y) respectively represent the horizontal direction gradient, the vertical direction gradient and the pixel value at the pixel point (x, y) in the input image;
the gradient magnitude and gradient direction at the pixel point are respectively:
it should be noted that the foregoing explanation of the embodiment of the high-texture image classification method based on visual saliency detection is also applicable to the apparatus of this embodiment, and is not repeated here.
According to the high-texture image classification system based on visual saliency detection provided by the embodiment of the invention, a saliency region is segmented by using an image pixel saliency value detection method based on color contrast, then the texture and gradient features of the saliency region are extracted by using a complete local binary pattern algorithm and a direction gradient histogram algorithm, an effective series fusion strategy is carried out, image detail information is jointly represented, and finally the extracted texture feature vectors are classified by using a nearest neighbor classifier to obtain an identification rate, so that the purpose of automatically classifying the extracted texture by using a computer is realized.
Furthermore, the terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present invention, "a plurality" means at least two, e.g., two, three, etc., unless specifically limited otherwise.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.
Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention, and that variations, modifications, substitutions and alterations can be made to the above embodiments by those of ordinary skill in the art within the scope of the present invention.

Claims (10)

1.一种基于视觉显著性检测的高纹理图像分类方法，其特征在于，包括以下步骤：1. a high texture image classification method based on visual saliency detection, is characterized in that, comprises the following steps: 通过颜色对比度的图像像素显著性值检测方法分割出显著性区域；The saliency region is segmented by the detection method of image pixel saliency value of color contrast; 通过完备局部二值模式算子和方向梯度直方图算法提取所述显著性区域的纹理特征和梯度特征，并通过有效串联融合策略共同表示图像细节信息；以及The texture features and gradient features of the saliency region are extracted by a complete local binary pattern operator and a directional gradient histogram algorithm, and image details are jointly represented by an effective series fusion strategy; and 通过最近邻分类器对提取到的融合向量进行分类，以获得识别率。The extracted fusion vectors are classified by the nearest neighbor classifier to obtain the recognition rate. 2.根据权利要求1所述的基于视觉显著性检测的高纹理图像分类方法，其特征在于，所述通过颜色对比度的图像像素显著性值检测方法分割出显著性区域，进一步包括：2. The high-texture image classification method based on visual saliency detection according to claim 1, wherein the saliency region is segmented by the image pixel saliency value detection method of color contrast, further comprising: 对颜色空间进行量化，以得到一组代表性色彩；Quantize the color space to obtain a representative set of colors; 获取所述代表性色彩对应的颜色在输入图像中的出现频率，并组成一个直方图；Obtain the frequency of occurrence of the color corresponding to the representative color in the input image, and form a histogram; 根据每个代表性色彩与其它代表性色彩的差异计算代表性色彩的显著性值；Calculate the significance value of the representative color according to the difference between each representative color and other representative colors; 将所述每个代表性色彩的显著性值赋予对应的像素。The saliency value of each representative color is assigned to the corresponding pixel. 3.根据权利要求1所述的基于视觉显著性检测的高纹理图像分类方法，其特征在于，所述颜色空间量化方法为八叉树色彩量化，其中，所述八叉树颜色量化分为建立色彩八叉树、产生调色板、生成量化文件，顺序读入像素颜色，建立一棵叶结点小于量化后颜色数的色彩八叉树，且对所述色彩八叉树进行遍历，如果图像中任一种颜色在所述色彩八叉树中不存在，则新插入一个叶子结点来表示所述任一种颜色，及若在插入像素颜色后，所述色彩八叉树的叶子节点数超过所述量化后颜色数，则根据归并策略做叶子结点的归并操作，以在所有像素插入后，只有不超过所述量化后颜色数种颜色被保存为调色板，并再次扫描文件，将每一颜色映射到所述调色板上，产生量化后的新图像。3. The high-texture image classification method based on visual saliency detection according to claim 1, wherein the color space quantization method is octree color quantization, wherein, the octree color quantization is divided into established Color octree, generate palette, generate quantization file, read pixel colors sequentially, build a color octree whose leaf node is less than the number of colors after quantization, and traverse the color octree, if the image Any one of the colors does not exist in the color octree, then a new leaf node is inserted to represent the any color, and if the pixel color is inserted, the number of leaf nodes of the color octree If the number of colors after quantization exceeds the number of colors after quantization, the merge operation of leaf nodes is performed according to the merge strategy, so that after all pixels are inserted, only the number of colors not exceeding the number of colors after quantization is saved as a palette, and the file is scanned again, Each color is mapped onto the palette, resulting in a new quantized image. 4.根据权利要求1所述的基于视觉显著性检测的高纹理图像分类方法，其特征在于，使用所述完备局部二值模式CLBP算法提取所述显著性区域的纹理特征，完备二值模式从灰度值大小关系特征CLBP_S、灰度值差值幅值特征CLBP_M和像素点灰度值与全局平均灰度值的大小关系特征描述像素点的纹理特征CLBP_C，最大化提取单个像素点的图像灰度纹理信息，完备局部二值模式特征的数学描述如下：4. The high-texture image classification method based on visual saliency detection according to claim 1, wherein the complete local binary pattern CLBP algorithm is used to extract the texture features of the saliency region, and the complete binary pattern is from The gray value size relationship feature CLBP_S, the gray value difference amplitude feature CLBP_M, and the size relationship feature between the gray value of the pixel point and the global average gray value describe the texture feature CLBP_C of the pixel point, and maximize the extraction of the image gray of a single pixel point. The mathematical description of the complete local binary pattern feature is as follows: CLBP_CN,R＝t(gc,cl)； CLBP_CN ,R =t(g c ,cl ); 其中，gi(i＝1,2,…,N)表示以gc为中心的邻域像素点的灰度值，R为邻域半径，mN表示中心像素点和邻域像素点差值的大小，c代表局部图像中mN的均值，cl表示全局灰度均值。Among them, g i (i=1,2,...,N) represents the gray value of the neighborhood pixel point with g c as the center, R is the neighborhood radius, m N represents the difference value between the center pixel point and the neighborhood pixel point The size of , c represents the mean of m N in the local image, and c l represents the global grayscale mean. 5.根据权利要求1所述的基于视觉显著性检测的高纹理图像分类方法，其特征在于，通过所述方向梯度直方图HOG算法提取所述显著性区域的梯度特征，所述方向梯度直方图算法计算像素点的梯度的公式如下：5. The high-texture image classification method based on visual saliency detection according to claim 1, wherein the gradient feature of the saliency region is extracted by the directional gradient histogram HOG algorithm, and the directional gradient histogram The formula for the algorithm to calculate the gradient of a pixel is as follows: Gx(x,y)＝H(x+1,y)-H(x-1,y)，G x (x,y)=H(x+1,y)-H(x-1,y), Gy(x,y)＝H(x,y+1)-H(x,y-1)，Gy(x,y)=H(x,y+1)-H(x,y-1), 其中，式中Gx(x,y)，Gy(x,y)，H(x,y)分别表示输入图像中像素点(x,y)处的水平方向梯度、垂直方向梯度和像素值；Among them, G x (x, y), Gy (x, y), H (x, y) respectively represent the horizontal gradient, vertical gradient and pixel value at the pixel point (x, y) in the input image; 所述像素点处的梯度幅值和梯度方向分别为：The gradient magnitude and gradient direction at the pixel points are: 6.根据权利要求1所述的基于视觉显著性检测的高纹理图像分类方法，其特征在于，将纹理特征特征向量和梯度特征向量串联融合，以形成所述融合特征向量，公式如下：6. the high-texture image classification method based on visual salience detection according to claim 1, is characterized in that, the texture feature feature vector and gradient feature vector are fused in series to form the fusion feature vector, and the formula is as follows: H＝[Hclbp,Hhog]，H=[H clbp , H hog ], 其中，式中H，Hclbp，Hhog分别表示融合特征向量，CLBP特征向量和HOG特征向量。where H, H clbp , and H hog represent the fusion feature vector, the CLBP feature vector and the HOG feature vector, respectively. 7.根据权利要求1所述的基于视觉显著性检测的高纹理图像分类方法，其特征在于，所述通过最近邻分类器对提取到的融合向量进行分类，进一步包括：7. The high-texture image classification method based on visual saliency detection according to claim 1, wherein the method for classifying the extracted fusion vector by the nearest neighbor classifier further comprises: 通过所述最近邻分类器计算两个直方图之间的相似性和差异性。The similarity and dissimilarity between the two histograms are calculated by the nearest neighbor classifier. 8.一种基于视觉显著性检测的高纹理图像分类系统，其特征在于，包括：8. A high-texture image classification system based on visual saliency detection, characterized in that, comprising: 检测模块，用于根据颜色对比度的图像像素显著性值检测方法分割出显著性区域；The detection module is used to segment the saliency area according to the detection method of the saliency value of the image pixels of the color contrast; 提取模块，用于通过完备局部二值模式算子和方向梯度直方图算法提取所述显著性区域的纹理特征和梯度特征，并通过有效串联融合策略共同表示图像细节信息；以及an extraction module for extracting texture features and gradient features of the saliency region through a complete local binary pattern operator and a directional gradient histogram algorithm, and jointly representing image detail information through an effective series fusion strategy; and 分类模块，用于通过最近邻分类器对提取到的融合向量进行分类，以获得识别率。The classification module is used to classify the extracted fusion vector by the nearest neighbor classifier to obtain the recognition rate. 9.根据权利要求8所述的基于视觉显著性检测的高纹理图像分类系统，其特征在于，使用所述完备局部二值模式算法提取所述显著性区域的纹理特征，完备二值模式从灰度值大小关系特征CLBP_S、灰度值差值幅值特征CLBP_M和像素点灰度值与全局平均灰度值的大小关系特征CLBP_C描述像素点的纹理特征，最大化提取单个像素点的图像灰度纹理信息，完备局部二值模式特征的数学描述如下：9 . The high-texture image classification system based on visual saliency detection according to claim 8 , wherein the complete local binary pattern algorithm is used to extract the texture features of the saliency region, and the complete binary pattern changes from gray to gray. 10 . Degree value size relationship feature CLBP_S, gray value difference amplitude feature CLBP_M, and the size relationship feature between pixel gray value and global average gray value CLBP_C Describe the texture feature of the pixel point, maximize the extraction of the image gray level of a single pixel point Texture information, the mathematical description of the complete local binary pattern feature is as follows: CLBP_CN,R＝t(gc,cl)； CLBP_CN ,R =t(g c ,cl ); 其中，gi(i＝1,2,…,N)表示以gc为中心的邻域像素点的灰度值，R为邻域半径，mN表示中心像素点和邻域像素点差值的大小，c代表局部图像中mN的均值，cl表示全局灰度均值。Among them, g i (i=1,2,...,N) represents the gray value of the neighborhood pixel point with g c as the center, R is the neighborhood radius, m N represents the difference value between the center pixel point and the neighborhood pixel point The size of , c represents the mean of m N in the local image, and c l represents the global grayscale mean. 10.根据权利要求8所述的基于视觉显著性检测的高纹理图像分类系统，其特征在于，通过所述方向梯度直方图HOG算法提取所述显著性区域的梯度特征，所述方向梯度直方图算法计算像素点的梯度的公式如下：10. The high-texture image classification system based on visual saliency detection according to claim 8, wherein the gradient feature of the saliency region is extracted by the directional gradient histogram HOG algorithm, and the directional gradient histogram The formula for the algorithm to calculate the gradient of a pixel is as follows: Gx(x,y)＝H(x+1,y)-H(x-1,y)，G x (x,y)=H(x+1,y)-H(x-1,y), Gy(x,y)＝H(x,y+1)-H(x,y-1)，Gy(x,y)=H(x,y+1)-H(x,y-1), 其中，式中Gx(x,y)，Gy(x,y)，H(x,y)分别表示输入图像中像素点(x,y)处的水平方向梯度、垂直方向梯度和像素值；Among them, G x (x, y), Gy (x, y), H (x, y) respectively represent the horizontal gradient, vertical gradient and pixel value at the pixel point (x, y) in the input image; 所述像素点处的梯度幅值和梯度方向分别为：The gradient magnitude and gradient direction at the pixel points are:
CN201810845341.7A 2018-07-27 2018-07-27 The high texture image classification method and system of view-based access control model conspicuousness detection Pending CN109214420A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810845341.7A CN109214420A (en) 2018-07-27 2018-07-27 The high texture image classification method and system of view-based access control model conspicuousness detection

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810845341.7A CN109214420A (en) 2018-07-27 2018-07-27 The high texture image classification method and system of view-based access control model conspicuousness detection

Publications (1)

Publication Number Publication Date
CN109214420A true CN109214420A (en) 2019-01-15

Family

ID=64990302

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810845341.7A Pending CN109214420A (en) 2018-07-27 2018-07-27 The high texture image classification method and system of view-based access control model conspicuousness detection

Country Status (1)

Country Link
CN (1) CN109214420A (en)

Cited By (27)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109993128A (en) * 2019-04-04 2019-07-09 北京航空航天大学 A Method for Quantifying Attribute Features of Pedestrian Image
CN110084160A (en) * 2019-04-16 2019-08-02 东南大学 A kind of video forest rocket detection method based on movement and brightness significant characteristics
CN110766009A (en) * 2019-10-31 2020-02-07 深圳市捷顺科技实业股份有限公司 Tail plate identification method and device and computer readable storage medium
CN110852994A (en) * 2019-10-15 2020-02-28 江门市得实计算机外部设备有限公司 Visual detection method and system for jet printing card
CN111080722A (en) * 2019-12-11 2020-04-28 中山大学 A color migration method and system based on saliency detection
CN111428778A (en) * 2020-03-20 2020-07-17 华侨大学 Automatic classification method for standard face section of fetus face
CN111783791A (en) * 2020-01-22 2020-10-16 北京沃东天骏信息技术有限公司 Image classification method, apparatus, and computer-readable storage medium
CN112164068A (en) * 2020-07-17 2021-01-01 中国人民解放军63892部队 Adaptive extraction method for target/interference region of alert radar PPI image
CN112183231A (en) * 2020-09-09 2021-01-05 广州杰赛科技股份有限公司 A method and device for vehicle identification based on high-definition video
CN112288040A (en) * 2020-01-10 2021-01-29 牧今科技 Method and system for performing image classification for object recognition
CN112288038A (en) * 2020-01-10 2021-01-29 牧今科技 Object recognition or object registration method based on image classification and computing system
CN112417931A (en) * 2019-08-23 2021-02-26 河海大学常州校区 A method for detection and classification of water surface objects based on visual saliency
CN112734722A (en) * 2021-01-08 2021-04-30 昆明理工大学 Flame endpoint carbon content prediction method based on improved complete local binary pattern
CN112884002A (en) * 2021-01-18 2021-06-01 四川大学 Intelligent electric meter version classification method
CN113095160A (en) * 2021-03-23 2021-07-09 中国大唐集团科学技术研究院有限公司华东电力试验研究院 Power system personnel safety behavior identification method and system based on artificial intelligence and 5G
CN113362354A (en) * 2021-05-07 2021-09-07 安徽国际商务职业学院 Method, system, terminal and storage medium for evaluating quality of tone mapping image
CN113689455A (en) * 2021-07-01 2021-11-23 上海交通大学 Thermal fluid image processing method, system, terminal and medium
CN114092442A (en) * 2021-11-22 2022-02-25 南京工程学院 Detection device and detection method for large-size leather surface defects
CN114240929A (en) * 2021-12-28 2022-03-25 季华实验室 Color difference detection method and device
US11538238B2 (en) 2020-01-10 2022-12-27 Mujin, Inc. Method and system for performing image classification for object recognition
CN116664960A (en) * 2023-07-28 2023-08-29 深圳弗瑞奇科技有限公司 A Cat Poop Recognition Method Based on Image Processing
CN117237265A (en) * 2022-06-08 2023-12-15 腾讯科技（深圳）有限公司 An image frozen frame detection method, device, equipment and storage medium
US12080006B2 (en) 2020-01-10 2024-09-03 Mujin, Inc. Method and system for performing image classification for object recognition
CN118839010A (en) * 2024-08-01 2024-10-25 北京远景视点科技有限公司 Automatic data cleaning system based on big data
CN119520799A (en) * 2023-08-18 2025-02-25 杭州阿里云飞天信息技术有限公司 Image recognition method, device, electronic device and storage medium
CN119851206A (en) * 2024-12-24 2025-04-18 北京鑫博腾飞科技有限公司 Intelligent detection method, system, device and product based on portrait monitoring
CN120088163A (en) * 2025-04-30 2025-06-03 江西华视光电有限公司 A method for quickly detecting out-of-control pixel points on a display screen

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102129693A (en) * 2011-03-15 2011-07-20 清华大学 Image vision significance calculation method based on color histogram and global contrast
EP3070643A1 (en) * 2015-03-20 2016-09-21 Thales Method and device for object recognition by analysis of digital image signals representative of a scene
CN107016409A (en) * 2017-03-20 2017-08-04 华中科技大学 A kind of image classification method and system based on salient region of image

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102129693A (en) * 2011-03-15 2011-07-20 清华大学 Image vision significance calculation method based on color histogram and global contrast
EP3070643A1 (en) * 2015-03-20 2016-09-21 Thales Method and device for object recognition by analysis of digital image signals representative of a scene
CN107016409A (en) * 2017-03-20 2017-08-04 华中科技大学 A kind of image classification method and system based on salient region of image

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
M.A.HASSAN, I.PARDIANSYAH, A.S.MALIK, I.FAYE AND W.RASHEE，: "Enhanced people counting system based head-shoulder detection in dense crowd scenario", 《INTERNATIONAL CONFERENCE ON INTELLIGENT AND ADVANCED SYSTEMS》 *
Z. GUO, L. ZHANG AND D. ZHANG，: "A Completed Modeling of Local Binary Pattern Operator for Texture Classification", 《IEEE TRANSACTIONS ON IMAGE PROCESSING》 *
ZHENGXIANG WANG, YIQUN HU, AND LIANG-TIEN CHIA: "Image-to-class distance metric learning for image classification", 《2010 EUROPEAN CONFERENCE ON COMPUTER VISION》 *
刘尚旺,李名,胡剑兰,崔艳萌: "基于视觉显著性检测的图像分类方法", 《计算机应用》 *
耿国华,周明全: "基于八叉树结构的色彩量化算法", 《小型微型计算机系统》 *

Cited By (37)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109993128A (en) * 2019-04-04 2019-07-09 北京航空航天大学 A Method for Quantifying Attribute Features of Pedestrian Image
CN110084160A (en) * 2019-04-16 2019-08-02 东南大学 A kind of video forest rocket detection method based on movement and brightness significant characteristics
CN112417931B (en) * 2019-08-23 2024-01-26 河海大学常州校区 Method for detecting and classifying water surface objects based on visual saliency
CN112417931A (en) * 2019-08-23 2021-02-26 河海大学常州校区 A method for detection and classification of water surface objects based on visual saliency
CN110852994A (en) * 2019-10-15 2020-02-28 江门市得实计算机外部设备有限公司 Visual detection method and system for jet printing card
CN110852994B (en) * 2019-10-15 2023-03-28 江门市得实计算机外部设备有限公司 Visual detection method and system for jet printing card
CN110766009A (en) * 2019-10-31 2020-02-07 深圳市捷顺科技实业股份有限公司 Tail plate identification method and device and computer readable storage medium
CN111080722A (en) * 2019-12-11 2020-04-28 中山大学 A color migration method and system based on saliency detection
CN111080722B (en) * 2019-12-11 2023-04-21 中山大学 Color migration method and system based on significance detection
US11772271B2 (en) 2020-01-10 2023-10-03 Mujin, Inc. Method and computing system for object recognition or object registration based on image classification
CN112288040A (en) * 2020-01-10 2021-01-29 牧今科技 Method and system for performing image classification for object recognition
CN112288038A (en) * 2020-01-10 2021-01-29 牧今科技 Object recognition or object registration method based on image classification and computing system
US11538238B2 (en) 2020-01-10 2022-12-27 Mujin, Inc. Method and system for performing image classification for object recognition
CN112288038B (en) * 2020-01-10 2021-05-18 牧今科技 Object recognition or object registration method based on image classification and computing system
US12080006B2 (en) 2020-01-10 2024-09-03 Mujin, Inc. Method and system for performing image classification for object recognition
CN111783791A (en) * 2020-01-22 2020-10-16 北京沃东天骏信息技术有限公司 Image classification method, apparatus, and computer-readable storage medium
CN111428778B (en) * 2020-03-20 2023-04-07 华侨大学 Automatic classification method for standard face section of fetus face
CN111428778A (en) * 2020-03-20 2020-07-17 华侨大学 Automatic classification method for standard face section of fetus face
CN112164068A (en) * 2020-07-17 2021-01-01 中国人民解放军63892部队 Adaptive extraction method for target/interference region of alert radar PPI image
CN112183231A (en) * 2020-09-09 2021-01-05 广州杰赛科技股份有限公司 A method and device for vehicle identification based on high-definition video
CN112734722A (en) * 2021-01-08 2021-04-30 昆明理工大学 Flame endpoint carbon content prediction method based on improved complete local binary pattern
CN112884002A (en) * 2021-01-18 2021-06-01 四川大学 Intelligent electric meter version classification method
CN113095160A (en) * 2021-03-23 2021-07-09 中国大唐集团科学技术研究院有限公司华东电力试验研究院 Power system personnel safety behavior identification method and system based on artificial intelligence and 5G
CN113362354A (en) * 2021-05-07 2021-09-07 安徽国际商务职业学院 Method, system, terminal and storage medium for evaluating quality of tone mapping image
CN113362354B (en) * 2021-05-07 2024-04-30 安徽国际商务职业学院 Quality evaluation method, system, terminal and storage medium for tone mapping image
CN113689455A (en) * 2021-07-01 2021-11-23 上海交通大学 Thermal fluid image processing method, system, terminal and medium
CN113689455B (en) * 2021-07-01 2023-10-20 上海交通大学 Thermal fluid image processing method, system, terminal and medium
CN114092442A (en) * 2021-11-22 2022-02-25 南京工程学院 Detection device and detection method for large-size leather surface defects
CN114240929B (en) * 2021-12-28 2024-07-19 季华实验室 A color difference detection method and device
CN114240929A (en) * 2021-12-28 2022-03-25 季华实验室 Color difference detection method and device
CN117237265A (en) * 2022-06-08 2023-12-15 腾讯科技（深圳）有限公司 An image frozen frame detection method, device, equipment and storage medium
CN116664960A (en) * 2023-07-28 2023-08-29 深圳弗瑞奇科技有限公司 A Cat Poop Recognition Method Based on Image Processing
CN119520799A (en) * 2023-08-18 2025-02-25 杭州阿里云飞天信息技术有限公司 Image recognition method, device, electronic device and storage medium
CN118839010A (en) * 2024-08-01 2024-10-25 北京远景视点科技有限公司 Automatic data cleaning system based on big data
CN119851206A (en) * 2024-12-24 2025-04-18 北京鑫博腾飞科技有限公司 Intelligent detection method, system, device and product based on portrait monitoring
CN120088163A (en) * 2025-04-30 2025-06-03 江西华视光电有限公司 A method for quickly detecting out-of-control pixel points on a display screen
CN120088163B (en) * 2025-04-30 2025-07-25 江西华视光电有限公司 Rapid detection method for out-of-control point of display screen pixel

Similar Documents

Publication Publication Date Title
CN109214420A (en) The high texture image classification method and system of view-based access control model conspicuousness detection
CN110543837B (en) Visible light airport airplane detection method based on potential target point
JP3740065B2 (en) Object extraction device and method based on region feature value matching of region-divided video
CN112686812A (en) Bank card inclination correction detection method and device, readable storage medium and terminal
CN112686248B (en) Certificate increase and decrease type detection method and device, readable storage medium and terminal
CN109711268B (en) Face image screening method and device
CN110717896A (en) Surface defect detection method of strip steel based on saliency label information propagation model
CN108629286B (en) Remote sensing airport target detection method based on subjective perception significance model
CN104166841A (en) Rapid detection identification method for specified pedestrian or vehicle in video monitoring network
CN109033972A (en) A kind of object detection method, device, equipment and storage medium
CN110400322A (en) Fruit Point Cloud Segmentation Method Based on Color and 3D Geometric Information
CN115937552B (en) An image matching method based on fusion of manual features and deep features
CN116503622A (en) Data Acquisition and Reading Method Based on Computer Vision Image
CN107578011A (en) Method and device for judging video key frames
CN114820987A (en) Three-dimensional reconstruction method and system based on multi-view image sequence
CN114511925B (en) A pig drinking behavior identification method based on the mouth area and multi-feature fusion
CN107622280A (en) Scene classification based image saliency detection method based on modular prescription
CN115170839B (en) Image data enhancement method and system
CN107977608B (en) Method for extracting road area of highway video image
Saputra et al. Integration GLCM and geometric feature extraction of region of interest for classifying tuna
CN115035390A (en) Aerial photography image detection method based on GAN and feature enhancement
Deng et al. Saliency Detection by Selective Strategy for Salient Object Segmentation.
CN116563591A (en) An Optical Smoke Detection Method Based on Feature Extraction in Sea and Sky Background
CN104778468B (en) Image processing device, image processing method, and monitoring device
CN114266703A (en) Image splicing method and system

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20190115