WO2020151529A1

WO2020151529A1 - Clothing image classification and image classification methods and apparatuses, and device

Info

Publication number: WO2020151529A1
Application number: PCT/CN2020/071925
Authority: WO
Inventors: 赵永飞; 龙一民; 袁炜; 徐博文; 吴剑; 胡露露; 张民英; 神克乐; 刘志敏; 陈新; 尹宁
Original assignee: 阿里巴巴集团控股有限公司
Priority date: 2019-01-23
Filing date: 2020-01-14
Publication date: 2020-07-30
Also published as: TW202109356A; CN111476253B; CN111476253A

Abstract

Embodiments of the present application provide clothing image classification and image classification methods and apparatuses, and a device. The clothing image classification method comprises: obtaining multiple clothing images; for any clothing image, dividing the clothing image into multiple region images; respectively extracting color histogram features of the multiple region images; performing weighting processing on the color histogram features of the multiple region images, and obtaining image features of the clothing image; clustering the multiple clothing images on the basis of the image features of the multiple clothing images, and obtaining multiple image sets; and determining the number of images of the multiple image sets respectively corresponding to at least one clothing image. The technical solution provided by the embodiments of the present application improves the classification accuracy of clothing.

Description

Clothing image classification, image classification method, device and equipment

This application claims the priority of the Chinese patent application with the application number 201910063759.7 and the invention title of "Clothing Image Classification, Image Classification Method, Apparatus and Equipment" filed on January 23, 2019, and the entire contents of which are incorporated into this application by reference .

Technical field

The embodiments of the present application relate to the field of computer application technology, and in particular to a clothing image classification, an image classification method, device, and equipment.

Background technique

At present, in order to facilitate management and improve work efficiency, existing garment processing factories use more than one production unit to handle only one fixed job, so as to realize the mass garment production process, which is the assembly line work method in the usual sense.

Since garment processing factories may process different garments at the same time, in order to obtain the processing quantity of different garments, it is necessary to count the number of different garments processed. Generally, a camera is used to collect images of clothing, and the sift feature points, textures, edge contours and other features of the image are extracted as image features. Then the images can be clustered based on the image features to obtain images of different types, and then confirm a certain type The image corresponds to the processing of the garment.

However, due to the diversification of clothing types, the collected clothing images may contain clothing images with richer textures, such as images collected for clothing with richer patterns, or clothing images with relatively simple textures, for example, for pure colors. The texture and contour features of these clothing images are too complex or too simple. When using the texture and other features of these clothing images to classify clothing images, the results are inaccurate and classification errors occur.

Summary of the invention

The embodiments of the present application provide a clothing image classification, an image classification method, device, and equipment to solve the technical problem of insufficient accuracy of multiple image classification results in the prior art.

In the first aspect, an embodiment of the present application provides a clothing image classification method, including:

Obtain multiple clothing images;

For any clothing image, divide the clothing image into multiple regional images;

Extracting the color histogram features of the multiple regional images respectively;

Weighting the color histogram features of the multiple regional images to obtain the image features of the clothing image;

Clustering the multiple clothing images based on the image features of the multiple clothing images to obtain multiple image sets;

It is determined that the plurality of image sets respectively correspond to the number of images of at least one clothing image.

In the second aspect, an image classification method is provided in an embodiment of the present application, including:

Acquire multiple images to be processed;

For any image to be processed, dividing the image to be processed into multiple regional images;

Weighting the color histogram features of the multiple regional images to obtain the image feature of the image to be processed;

Classifying the multiple to-be-processed images based on the image features of the multiple to-be-processed images to obtain multiple image sets;

Determine the image type to which at least one image to be processed respectively corresponding to the multiple image sets belongs;

Determine the number of images of at least one image to be processed corresponding to each image category.

In the third aspect, an embodiment of the present application provides a clothing image classification device, including:

The first acquisition module is used to acquire multiple clothing images;

The first division module is configured to divide the clothing image into multiple regional images for any clothing image;

The first extraction module is configured to extract the color histogram features of the multiple regional images respectively;

The first weighting module is configured to perform weighting processing on the color histogram features of the multiple regional images to obtain the image features of the clothing image;

The first clustering module is configured to cluster the multiple clothing images based on the image features of the multiple clothing images to obtain multiple image sets;

The first determining module is configured to determine the number of images corresponding to at least one clothing image in each of the plurality of image sets.

In a fourth aspect, an image classification device is provided in an embodiment of the present application, including:

The second acquisition module is used to acquire multiple images to be processed;

The second division module is configured to divide the to-be-processed image into multiple regional images for any one to-be-processed image;

The second extraction module is used to extract the color histogram features of the multiple regional images respectively;

The second weighting module is configured to perform weighting processing on the color histogram features of the multiple regional images to obtain the image features of the image to be processed;

The second clustering module is configured to classify the multiple to-be-processed images based on the image features of the multiple to-be-processed images to obtain multiple image sets;

The second determining module is configured to determine the image type to which at least one image to be processed respectively corresponding to the multiple image sets belongs;

The third determining module is used to determine the image quantity of at least one image to be processed corresponding to each image category.

In a fifth aspect, an embodiment of the present application provides a clothing image classification device, including: a storage component and a processing component, the storage component stores one or more computer instructions, and the one or more computer instructions are used for the processing The component is called and executed;

The processing component is used for:

Acquire multiple clothing images; for any clothing image, divide the clothing image into multiple regional images; extract the color histogram features of the multiple regional images respectively; combine the color histogram features of the multiple regional images Perform weighting processing to obtain image features of the clothing image; cluster the multiple clothing images based on the image features of the multiple clothing images to obtain multiple image sets; determine that the multiple image sets correspond to each other The number of images of at least one clothing image.

In a sixth aspect, an embodiment of the present application provides an image classification device, including: a storage component and a processing component, the storage component stores one or more computer instructions, and the one or more computer instructions are provided to the processing component Call and execute

The processing component is used for:

Acquire multiple images to be processed; for any image to be processed, divide the image to be processed into multiple regional images; extract the color histogram features of the multiple regional images respectively; The histogram features are weighted to obtain the image features of the image to be processed; based on the image features of the multiple images to be processed, the multiple images to be processed are classified to obtain multiple image sets; The image types to which at least one image to be processed respectively correspond to each image set; determine the number of images of at least one image to be processed corresponding to each image category.

In the embodiment of the present application, after extracting multiple clothing images, the clothing image can be divided into multiple regional images for any clothing image, and then the color histogram feature of each regional image can be extracted separately. The color histogram characteristics of the image can reflect the local color characteristics of the clothing image in the region; the color histograms of multiple regional images are weighted to generate the image characteristics of the clothing image, and the obtained image characteristics can realize the overall color characteristics of the product image In addition, because the texture of the image is actually produced by the change of pixel value, the histogram feature of each area image is actually obtained by counting the color histogram of each pixel. Therefore, the color histogram of multiple area images is weighted The image characteristics obtained by processing can also characterize the characteristics of the image from the texture. Furthermore, based on the image features of multiple clothing images, when multiple clothing images are clustered, accurate classification results can be obtained. Since the image features of each clothing image can represent the color and texture features of the clothing image locally and as a whole, the classification error of multiple image sets obtained is small, and it can be accurately determined that multiple image sets correspond to at least one clothing image. The number of images can be used to accurately monitor the clothing production process through the number of clothing images.

These and other aspects of the present application will be more concise and understandable in the description of the following embodiments.

Description of the drawings

In order to more clearly describe the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the drawings in the following description These are some embodiments of the application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative work.

Fig. 1 shows a flowchart of an embodiment of a clothing image classification method provided by the present application;

Figure 2 shows a flowchart of another embodiment of a clothing image classification method provided by the present application;

Fig. 3 shows a flowchart of an embodiment of a picture classification method provided by the present application;

Fig. 4 shows a schematic structural diagram of an embodiment of a clothing image classification device provided by the present application;

FIG. 5 shows a schematic structural diagram of an embodiment of a clothing image classification device provided by the present application;

FIG. 6 shows a schematic structural diagram of an embodiment of a picture classification device provided by this application;

Fig. 7 shows a schematic structural diagram of an embodiment of a picture classification device provided by the present application.

detailed description

In order to enable those skilled in the art to better understand the solutions of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely in conjunction with the drawings in the embodiments of the present application.

In some of the procedures described in the specification and claims of this application and the above-mentioned drawings, multiple operations appearing in a specific order are included, but it should be clearly understood that these operations may not be in the order in which they appear in this document. Execution or parallel execution, the sequence numbers of operations such as 101, 102, etc., are only used to distinguish different operations, and the sequence numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions of "first" and "second" in this article are used to distinguish different messages, devices, modules, etc., and do not represent a sequence, nor do they limit the "first" and "second" Are different types.

The embodiments of the present invention can be applied to digital factory management, through the intelligent monitoring of the clothing production process, so as to know the production progress of the clothing at any time and improve production efficiency.

In the prior art, a camera can be used to collect images of clothing, and image features such as texture and contour of the image can be extracted, and the collected multiple images can be classified based on the image features, so as to obtain images belonging to different types, and then determine different The production schedule of the garment corresponding to the type of image. However, when the image is characterized by features such as texture and contour, the result of image classification is not accurate enough, and classification errors are prone to occur. The inventor found that when extracting image features such as texture and contour of an image, because the entire image is directly extracted, the obtained image features cannot well represent local features, and the features extracted in this way are easily affected by the shape, The influence of factors such as color and pattern is also easily affected by factors such as light and exposure, and the extracted features are not accurate enough, leading to classification errors.

Based on this, the inventor proposed the technical solution of the present application. In the embodiment of the present invention, after extracting multiple clothing images, the clothing image can be divided into multiple regional images for any clothing image, and then can be extracted separately The color histogram features of each regional image are weighted to obtain the image features of the clothing image. The clothing image is divided into multiple regional images, and the color histogram of each regional image is extracted to extract the local features of the clothing image. Because the color histogram features of multiple regional images are weighted, the clothing image is realized The overall characteristics of the clothing image can be fully characterized from a spatial perspective; in addition, the histogram feature can represent the characteristics of the image from the color and texture, and is not easily affected by the clothing itself and the image collection environment. Furthermore, the weighted processing of the color histograms of the multiple region images is used to generate the image features of the clothing images. When the multiple clothing images are clustered, accurate classification results can be obtained. The obtained multiple image sets have relatively small classification errors, and the number of images corresponding to at least one clothing image in each of the multiple image sets can be accurately determined, and the clothing production process can be accurately monitored through the number of clothing images.

The following will clearly and completely describe the technical solutions in the embodiments of the present application in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative work fall within the protection scope of this application.

As shown in FIG. 1, it is a flowchart of an embodiment of a clothing image classification method provided by an embodiment of the present invention. The method may include the following steps:

101: Acquire multiple clothing images.

Among them, each clothing image can correspond to one clothing, and the clothing corresponding to each image is different.

Optionally, in order to obtain an accurate production quantity of garments to perform production scheduling of different types of garments, garment images can be collected for garments that have been produced and are in the quality inspection program.

In practical applications, since a factory may contain multiple production lines, images collected by different cameras may have different sizes. In order to classify and process the same image, multiple images collected by different cameras can be normalized To obtain multiple clothing images. When the multiple images are normalized, the aspect ratio of the original image can be maintained, and then multiple clothing images with the same aspect ratio and normalized size can be obtained.

102: For any clothing image, divide the clothing image into multiple regional images.

It is possible to select any clothing image from a plurality of clothing images, and divide the selected clothing image into multiple regional images.

In some embodiments, in order to obtain a balanced division result, for any clothing image, the clothing image may be divided into multiple area images according to the preset area position and the preset area size.

Wherein, the area position may refer to the center position of the area image, and the area size may refer to the image size of the area image, specifically it may refer to the number of horizontal pixels and the number of vertical pixels of the area image.

In a possible implementation manner, in order to make the clothing image evenly divided and to collect color features at different positions, any clothing image can be divided into multiple regional images according to the number of N*M. N can be equal to M, and the number of horizontal and vertical divisions of the clothing image is the same. Therefore, the area size of each area image is the same. Among them, N is greater than 1, less than the number of pixels in the clothing image. In practical applications, N can take a value of 3, that is, the clothing image can be evenly divided into 9 regional images of the same size.

103: Extract the color histogram features of the multiple regional images respectively.

The clothing image is divided into blocks to obtain the color histogram features of each region image respectively, so as to extract the local features of the clothing image and pay attention to the local features of the clothing image.

In order to improve statistical efficiency, in practical applications, you can directly perform histogram statistics on multiple pixels of the regional image to obtain the color histogram characteristics of the regional image. Extract the color histogram features of the multiple area images respectively, and obtain the color histogram features corresponding to the multiple area images respectively. 104: Perform weighting processing on the color histogram features of multiple regional images to obtain image features of the clothing image.

Among them, the color histogram feature of each region image can be expressed in the form of a color feature matrix.

As a possible implementation manner, the color histogram features of multiple regional images can be weighted according to a predetermined weighting rule to obtain the image features of the clothing image. The predetermined weighting rule may actually mean that the color histogram feature of each regional image is weighted according to the feature weight and weighted position corresponding to each regional image to obtain the image feature of the clothing image.

In some embodiments, in order to balance the features of each region on the image features of the clothing image, so that the obtained image features are more accurate, the weight of the color histogram feature of each region image can be set to 1. When the color histogram feature of each region image is a color feature matrix, performing weighting processing on the color histogram features of the multiple region images to obtain the image features of the clothing image may include: according to each region image in the clothing image Determine the weighting order of the color histogram features of each regional image; combine multiple color feature matrices in horizontal or vertical matrix according to their corresponding weighting order, and the obtained feature matrix is the image feature of the clothing image.

In practical applications, in order to ensure that the color histogram features of each region image have the same local impact on the image features of the clothing image and generate accurate image features, the feature weight of each region image can be set the same, for example, set to 1. At this time, the color histogram features corresponding to the multiple regional images can be directly stitched to obtain the image features of the clothing image. For example, assuming that the color histogram features corresponding to the three adjacent area images A1, A2, and A3 are B1, B2, and B3, the clothing image corresponding to the area images A1, A2, and A3 may have the feature B1B2B3.

105: Based on the image features of the multiple clothing images, cluster the multiple clothing images to obtain multiple image sets.

Each clothing image can represent a clothing, and the image feature of each clothing image can represent the feature of the clothing. If the clothing image is the same, it means that the clothing corresponding to the clothing image is the same. For the same type of clothing, the image features of the clothing image are the same. Based on the respective image features of multiple clothing images, clustering multiple clothing images to obtain multiple image sets can specifically be to divide the respective image features of multiple clothing images and classify clothing images with higher feature similarity into the same category Image, get multiple image collections.

Based on the respective image features of multiple clothing images, a clustering algorithm is used to cluster multiple clothing images to obtain multiple image sets.

In some embodiments, a clustering algorithm may be used to perform feature clustering on respective image features of multiple clothing images, and then clothing images corresponding to feature classes with the same image features are used as an image set to obtain multiple image sets.

Wherein, each image set may correspond to at least one clothing image, and the clothing corresponding to the at least one clothing image is of the same type of clothing.

106: Determine the number of images in which the multiple image sets respectively correspond to at least one clothing image.

Each image collection can correspond to at least one clothing image, and the number of at least one clothing image can be counted to obtain the number of images, that is, to obtain the number of images corresponding to each image collection.

In some embodiments, after determining the number of images corresponding to at least one clothing image in each of the multiple image sets, the multiple image sets and the number of images corresponding to each image set may also be output to facilitate the user to view.

In the embodiment of the present invention, after extracting multiple clothing images, the clothing image can be divided into multiple regional images for any clothing image, and then the color histogram features of each regional image can be extracted separately. The color histogram characteristics of the image can reflect the local color characteristics of the clothing image in the region; the color histograms of multiple regional images are weighted to generate the image characteristics of the clothing image, and the obtained image characteristics can realize the overall color characteristics of the product image s concern.

In addition, because the texture of the image is actually produced by the change in pixel value, the histogram feature of each area image is actually obtained by counting the color histogram of each pixel. Therefore, the color histogram of multiple area images is weighted. Image features can also characterize image characteristics from texture. Furthermore, based on the image features of multiple clothing images, when multiple clothing images are clustered, accurate classification results can be obtained. Since the image features of each clothing image can represent the color and texture features of the clothing image locally and as a whole, the classification error of multiple image sets obtained is small, and it can be accurately determined that multiple image sets correspond to at least one clothing image. The number of images can be used to accurately monitor the clothing production process through the number of clothing images.

In practical applications, the color histogram feature may include a color feature matrix, that is, the color histogram feature expresses the characteristics of the image in the form of a color feature matrix.

The performing weighting processing on the color histogram features of the multiple regional images to obtain the image features of the clothing image may include:

Perform horizontal or vertical matrix stitching processing on the respective color feature matrices of the multiple regional images to obtain image features of the clothing image.

The horizontal splicing processing of any two color feature matrices may specifically refer to splicing any row of the first color feature matrix and a row corresponding to the second color feature matrix to obtain a horizontal splicing matrix of two color feature matrices. For example, suppose the first color feature matrix is [101; 100], the second color feature matrix is [202; 200], and the horizontal mosaic matrix obtained by horizontally splicing two color feature matrices is [101202; 100200].

The vertical splicing processing of any two color feature matrices may specifically refer to longitudinal splicing of the first color feature matrix and the second color feature matrix based on the last row of the first color feature matrix and the first row of the second color feature matrix to obtain Vertical mosaic matrix of two color feature matrices. For example, assuming that the first color feature matrix is [101; 100] and the second color feature matrix is [202; 200], the vertical stitching matrix obtained by vertically stitching the two color feature matrices may be [101; 100; 202; 200].

As shown in FIG. 2, it is a flowchart of another embodiment of a clothing image classification method provided by an embodiment of the present invention. The difference from the embodiment shown in FIG. 1 lies in that it is determined in step 106 that the multiple After the image sets respectively correspond to the number of images of at least one clothing image, the method may further include:

201: Determine the clothing type where at least one clothing image corresponding to each of the multiple image sets is located.

Each image collection can correspond to at least one clothing image, and clothing images belonging to the same image belong to the same clothing type. Therefore, at least one clothing image corresponding to each image collection belongs to the same clothing type. It can be determined that multiple image collections correspond to each other. At least one clothing type of clothing image.

As an embodiment, the determining the clothing type of at least one clothing image corresponding to each of the multiple image sets may include:

Determine at least one training image and the image characteristics of each training image; wherein each training image corresponds to a clothing type;

Determining a training image matching each image set based on the image features of at least one clothing image corresponding to the multiple image sets;

The clothing types of the training images respectively corresponding to the multiple image sets are used as clothing types where at least one clothing image corresponding to the multiple image sets is located.

In order to make the feature expression modes of the images the same to achieve feature matching of different images, the image feature acquisition method of the training image is the same as the feature acquisition method of the embodiment shown in FIGS. 1 to 2.

Therefore, the image features of each training image can be obtained in the following ways:

For any training image, dividing the training image into multiple training region images;

Extracting the color histogram features of the multiple training region images respectively;

The color histogram features of the multiple training region images are weighted to obtain the image features of the multiple training images.

The regional position of each regional image is different. In order to make the image feature representation of different clothing images the same, the color histogram feature of the regional image can be weighted according to the regional location to obtain the image feature of the clothing image. As another embodiment, the above step 104: weighting the color histogram features of the multiple region images, and obtaining the image features of the clothing image may include:

The color histogram features of the multiple regional images are weighted according to the regional positions of the multiple regional images to obtain the image features of the clothing image.

Performing weighting processing according to the regional positions of multiple regional images can specifically refer to determining the weighted position of the regional image according to the regional position of any regional image, and according to the corresponding weighted position of the regional image, the corresponding color histogram feature The feature stitching is performed to obtain the image features of the clothing image when the feature stitching of the color histogram features of the multiple regional images is completed. When the color histogram feature is a color feature matrix, the feature splicing of the color histogram feature can refer to the matrix splicing of the color histogram matrix.

Since the method of simply counting the pixel value of each pixel can represent the overall color characteristics of different pixels, but cannot display the characteristics of different pixels in different color spaces, as another embodiment, the step 103 extracts The color histogram features of the multiple regional images may include:

Extracting the color histogram features of the multiple region images in multiple color channels respectively.

The color histogram features of each area image in multiple color channels can be extracted sequentially. Based on multiple color channels, the color histogram features of multiple regional images can be extracted respectively.

Among them, each color channel can represent different color characteristics, and the characteristics of different pixels in different color spaces can be obtained by converting to the corresponding color channel.

In order to obtain accurate color histogram features, as yet another embodiment, the respectively extracting the color histogram features of the multiple region images in multiple color channels may include:

Extracting channel histogram features corresponding to each of the multiple color channels of the regional image for any one of the multiple regional images;

The channel histogram features corresponding to the multiple color channels are weighted to obtain the color histogram feature of the regional image.

The channel histogram feature can refer to the channel feature matrix. Weighting the channel histogram features corresponding to multiple color channels to obtain the color histogram feature of the regional image may include: performing horizontal or vertical matrix stitching processing on the multiple channel feature matrices to obtain the color histogram feature of the regional image . It is also possible to determine the respective weighting order of multiple color channels through the channel order of multiple color channels, and perform horizontal or vertical splicing of multiple channel feature matrices according to their respective weighting order. The feature matrix obtained is the color histogram of the regional image. Figure features.

The color histogram features of the regional image are obtained by extracting the channel histogram features corresponding to the multiple color histogram channels of the regional image, and weighting the multiple channel histogram features. In practical applications, for the weighting processing of channel histogram features corresponding to multiple color channels, each channel histogram feature may be weighted to generate the color histogram of the regional image.

In some embodiments, for any one of the multiple regional images, extracting the channel histogram features corresponding to the multiple color channels of the regional image includes:

For any one of the multiple regional images, assign each pixel of the regional image to the corresponding channel value of the multiple color channels;

Divide each color channel into multiple histogram statistical intervals;

For any color channel, count the channel values of the multiple pixels of the region image in the color channel, and the number of distributions corresponding to multiple histogram statistical intervals in the color channel;

Based on the respective distribution numbers of the multiple histogram statistical intervals, the channel histogram characteristics of the region image in the color channel are determined.

In some embodiments, the clothing image classification method described above may further include:

Multiple color channels formed by three color channels corresponding to at least one color space model are determined.

At least one color space module can be set in advance, and multiple color channels can be obtained through three color channels corresponding to at least one color space model. The determination of multiple color channels may specifically be to obtain three corresponding to at least one color space model set in advance. Multiple color channels formed by color channels, thereby improving processing efficiency.

For any regional image, the channel histogram corresponding to the multiple color channels of the regional image can be extracted. Specifically, the respective pixel values of multiple pixels of the regional image can be converted to multiple color channels to obtain the corresponding multiple of each pixel. The channel value of each color channel, and for each color channel, the channel value corresponding to multiple pixels in each area image can be obtained, and then channel histogram statistics can be performed for any color channel.

Due to the large number of channels, if the data corresponding to each channel is counted one by one, the calculation process needs to be increased, and the data dimension of the obtained channel histogram feature will also be very high. Therefore, each color channel can be divided into histogram statistical intervals, so that different color channels can be counted according to the histogram statistical intervals.

In some possible designs, the at least one color space includes RGB color space, HSV color space, and YcbCr color space;

Wherein, the RGB color space corresponds to the R color channel, the G color channel, and the B color channel, the HSV color space corresponds to the H color channel, the S color channel, and the V color channel, and the YcbCr color space corresponds to the Y color channel, Cb color channel, and Cv color. aisle.

For different color channels, according to their channel characteristics, each color channel can be divided into multiple histogram statistical intervals according to the number of divisions.

As a possible implementation, the R color channel can be divided into 24 histogram statistical intervals, the G color channel can be divided into 24 histogram statistical intervals, and the B color channel can be divided into 24 histogram statistical intervals. The HSV color space corresponds to The H color channel is divided into 32 histogram statistics intervals, the S color channel is divided into 32 histogram statistics intervals, the V color channel corresponds to 8 histogram statistics intervals, and the YcbCr color space corresponds to the Y color channel is divided into 24 histogram statistics intervals. , The Cb color channel is divided into 16 histogram statistical intervals, and the Cv color channel is divided into 16 statistical intervals.

Furthermore, it is possible to count the number of occurrences in each statistical interval of the channel values of the multiple pixel points corresponding to the multiple color channels, and then obtain the histogram characteristic value corresponding to each statistical interval.

Taking the above G color channel division as an example, after the G color channel is divided into 24 histogram statistical intervals, the channel values of multiple pixels in the G color channel can be counted, the number of occurrences in each statistical interval, and then G The distribution numbers of the color channels corresponding to each of the 24 statistical intervals, the channel histogram feature of the G color channel is obtained, and the feature is 1*24 dimensional data. The histogram features corresponding to color channels such as R, B, H, S, V, Y, Cb, Cr can also be obtained, and the dimensions of their features are 1*24, 1*24, 1*32, 1*, respectively 32, 1*8, 1*24, 1*16, 1*16.

After that, the channel histogram features corresponding to the multiple color channels can be weighted to obtain the color histogram features of the regional image, and the feature weighting can be performed in this way, for example, the channel histogram features corresponding to each channel Perform weighting directly to obtain a 1*200-dimensional color feature histogram. For example, each clothing channel is divided into 9 regional images on average, and the corresponding color histogram features of multiple regional images are subjected to feature weighting processing, such as simple feature stitching, to obtain 1*1800 dimensional image features of clothing images.

After clustering multiple clothing images based on image features, the clustering results of multiple clothing images can be output to facilitate users to view. As yet another embodiment, after the determining the number of images of the plurality of image sets corresponding to at least one clothing image, the method further includes:

Sort the number of images to determine the output order of each image set;

At least one clothing image corresponding to the multiple image sets is sequentially outputted according to the respective output order of the multiple image sets.

Optionally, sorting the number of images to determine the output order of each image collection may specifically refer to ascending or descending order of the number of images according to the number of images to determine the output order of each image collection.

As shown in FIG. 3, it is a flowchart of another embodiment of an image classification method provided by an embodiment of the present invention. The method may include the following steps:

301: Acquire multiple images to be processed.

302: For any image to be processed, divide the image to be processed into multiple regional images.

303: Extract the color histogram features of the multiple regional images respectively.

304: Perform weighting processing on the color histogram features of the multiple regional images to obtain the image features of the image to be processed.

305: Based on the image features of the multiple to-be-processed images, classify the multiple to-be-processed images to obtain multiple image sets.

306: Determine the image type to which at least one image to be processed respectively corresponding to the multiple image sets belongs.

307: Determine the image quantity of at least one image to be processed corresponding to each image category.

The image classification method provided by the embodiment of the present invention is the same as the image processing method used in the clothing image classification method shown in FIG. 1 to FIG. 2, and will not be repeated here.

In the embodiment of the present invention, after extracting multiple images to be processed, the image to be processed can be divided into multiple regional images for any image to be processed, and then the color histogram feature of each regional image can be extracted separately. The color histogram features of multiple regional images are weighted to obtain the image features of the image to be processed. The color histogram features of the regional image can reflect the local features of the image to be processed. Weighting the color histogram features of multiple regional images can realize the attention to the overall features of the image to be processed; in addition, the histogram features can be based on color and texture. The above characterizes the characteristics of the image, and is not easily affected by the clothing itself and the image collection environment. Furthermore, the weighted processing of the color histograms of the multiple regional images is used to generate the image characteristics of the image to be processed, and when the multiple images to be processed are clustered, accurate classification results can be obtained. The classification error of the obtained multiple image sets is small, and the number of images of the multiple image sets corresponding to at least one image to be processed can be accurately determined, and the clothing production process can be accurately monitored through the number of images to be processed.

As shown in FIG. 4, it is a schematic structural diagram of another embodiment of a clothing image classification apparatus provided by an embodiment of the present invention. The apparatus may include the following modules:

The first acquiring module 401 is used to acquire multiple clothing images;

The first division module 402 is configured to divide the clothing image into multiple regional images for any clothing image;

The first extraction module 403 is configured to extract the color histogram features of the multiple regional images respectively;

The first weighting module 404 is configured to perform weighting processing on the color histogram features of the multiple regional images to obtain the image features of the clothing image;

The first clustering module 405 is configured to cluster the multiple clothing images based on the image features of the multiple clothing images to obtain multiple image sets;

The first determining module 406 is configured to determine the number of images corresponding to at least one clothing image in each of the multiple image sets.

In the embodiments of this application, the color histogram features of the regional image can reflect the local features of the clothing image, and weighting the color histogram features of multiple regional images can realize the attention to the overall features of the clothing image; in addition, the histogram features The characteristics of the image can be characterized from the color and texture, and it is not easily affected by the clothing itself and the image collection environment. Furthermore, the weighted processing of the color histograms of the multiple region images is used to generate the image features of the clothing images. When the multiple clothing images are clustered, accurate classification results can be obtained. The classification error of the obtained multiple image sets is small, and the number of images corresponding to at least one clothing image in the multiple image sets can be accurately determined, and the clothing production process can be accurately monitored through the number of clothing images.

As an embodiment, the device shown in FIG. 4 may further include:

The fourth determining module is used to determine the clothing type of at least one clothing image corresponding to the multiple image sets.

As yet another embodiment, the device shown in FIG. 4 may further include:

A sorting module, used to sort the number of images to determine the output order of each image set;

The output module is configured to sequentially output at least one clothing image corresponding to each of the multiple image sets according to the respective output order of the multiple image sets.

The regional position of each regional image is different. In order to make the image feature representation of different clothing images the same, the color histogram feature of the regional image can be weighted according to its regional location to obtain the image features of the clothing image as another implementation For example, the first weighting module includes:

The first weighting unit is configured to perform weighting processing on the color histogram features of the multiple regional images according to the regional positions of the multiple regional images to obtain the image features of the clothing image.

Since the method of simply counting the pixel value of each pixel can represent the overall color characteristics of different pixels, but cannot display the characteristics of different pixels in different color spaces, therefore, as another embodiment, the first extraction module include:

The first extraction unit is configured to extract the color histogram features of the multiple regional images in multiple color channels.

In order to obtain accurate color histogram features, as another embodiment, the first extraction unit includes:

The first extraction subunit is configured to extract the channel histogram features corresponding to each of the multiple color channels of the regional image for any one of the multiple regional images;

The first weighting subunit is used for weighting the channel histogram features corresponding to the multiple color channels to obtain the color histogram feature of the regional image.

In some embodiments, the first extraction subunit may specifically include:

Divide each color channel into multiple histogram statistical intervals;

In some embodiments, it further includes:

The channel determining module is used to determine multiple color channels formed by three color channels corresponding to at least one color space model.

As an embodiment, the color histogram feature includes a color feature matrix;

The first weighting module may include:

The second weighting module is configured to perform horizontal or vertical matrix splicing processing on the respective color feature matrices of the multiple regional images to obtain the image features of the clothing image.

The device described in FIG. 4 can execute the clothing image classification method described in the embodiments shown in FIG. 1 to FIG. 2, and its implementation principles and technical effects will not be described in detail. The specific manners of performing operations of the various modules and units of the clothing image classification device in the foregoing embodiment have been described in detail in the embodiment of the method, and detailed description will not be given here.

In addition, an embodiment of the present application also provides a computer-readable storage medium that stores a computer program, and when the computer program is executed by a computer, the clothing image classification method of the embodiments shown in FIGS. 1 to 2 can be implemented.

As shown in Figure 5, it is a schematic structural diagram of another embodiment of a clothing image classification device provided by an embodiment of the present invention. The device may include a storage component 501 and a processing component 502. The storage component 501 stores one or Multiple computer instructions, the one or more computer instructions are for the processing component 502 to call and execute;

The processing component 502 can be used for:

In the embodiments of this application, the color histogram features of the regional image can reflect the local features of the clothing image, and weighting the color histogram features of multiple regional images can realize the attention to the overall features of the clothing image; in addition, the histogram features The characteristics of the image can be characterized from the color and texture, and it is not easily affected by the clothing itself and the image collection environment. Furthermore, the weighted processing of the color histograms of the multiple region images is used to generate the image features of the clothing images. When the multiple clothing images are clustered, accurate classification results can be obtained. The obtained multiple image sets have relatively small classification errors, and the number of images corresponding to at least one clothing image in each of the multiple image sets can be accurately determined, and the clothing production process can be accurately monitored through the number of clothing images.

As an embodiment, the processing component 501 may also be used for:

Determine the clothing type where at least one clothing image corresponding to each of the multiple image sets is located.

As yet another embodiment, the processing component 501 may also be used for:

Sort the number of images to determine the output order of each image set;

Since the method of simply counting the pixel value of each pixel can represent the overall color characteristics of different pixels, but cannot display the characteristics of different pixels in different color spaces, therefore, as another embodiment, the processing component will The color histogram features of the multiple regional images are weighted, and the image features of the clothing image can be obtained by:

In order to obtain accurate color histogram features, as another embodiment, the processing component separately extracts the color histogram features of the multiple regional images, specifically:

In some embodiments, the processing component separately extracting the color histogram features of the multiple regional images in the multiple color channels may specifically be: for any one of the multiple regional images, extracting the Channel histogram characteristics corresponding to multiple color channels of the regional image;

In some embodiments, for any one of the multiple regional images, the processing component extracts the channel histogram features corresponding to the multiple color channels of the regional image, specifically:

Divide each color channel into multiple histogram statistical intervals;

For any color channel, count the channel values of the multiple pixels of the regional image in the color channel, and the number of distributions corresponding to multiple histogram statistical intervals in the color channel;

In some embodiments, the processing component 501 may also be used to:

In some possible designs, the at least one color space determined by the processing component includes an RGB color space, an HSV color space, and a YcbCr color space;

The device described in FIG. 5 can execute the clothing image classification method described in the embodiments shown in FIG. 1 to FIG. 2, and its implementation principles and technical effects will not be repeated. The specific methods of processing components in the clothing image classification device in the foregoing embodiment have been described in detail in the embodiment of the method, and detailed description will not be given here.

As shown in FIG. 6, it is a schematic structural diagram of another embodiment of an image classification apparatus provided by an embodiment of the present invention. The apparatus may include the following modules:

The second acquisition module 601 is used to acquire multiple images to be processed;

The second dividing module 602 is configured to divide the to-be-processed image into multiple regional images for any one to-be-processed image;

The second extraction module 603 is configured to extract the color histogram features of the multiple regional images respectively;

The second weighting module 604 is configured to perform weighting processing on the color histogram features of the multiple regional images to obtain the image features of the image to be processed;

The second clustering module 605 is configured to classify the multiple to-be-processed images based on the image features of the multiple to-be-processed images to obtain multiple image sets;

The second determining module 606 is configured to determine the image type to which at least one image to be processed respectively corresponding to the multiple image sets belongs;

The third determining module 607 is configured to determine the image quantity of at least one image to be processed corresponding to each image category.

In the embodiment of the present invention, the color histogram features of the regional image can reflect the local features of the image to be processed, and weighting the color histogram features of multiple regional images can realize the focus on the overall features of the image to be processed; in addition, the histogram Features can characterize image characteristics from color and texture, and are not easily affected by the clothing itself and the image collection environment. Furthermore, the weighted processing of the color histograms of the multiple regional images is used to generate the image characteristics of the image to be processed, and when the multiple images to be processed are clustered, accurate classification results can be obtained. The classification error of the obtained multiple image sets is small, and the number of images of the multiple image sets corresponding to at least one image to be processed can be accurately determined, and the clothing production process can be accurately monitored through the number of images to be processed.

The device described in FIG. 6 can execute the image classification method described in the embodiment shown in FIG. 3, and its implementation principles and technical effects will not be repeated. The specific methods for performing operations of the various modules and units of the image classification device in the foregoing embodiment have been described in detail in the embodiment of the method, and detailed description will not be given here.

In addition, an embodiment of the present application also provides a computer-readable storage medium that stores a computer program, and when the computer program is executed by a computer, the image classification method of the embodiment shown in FIG. 3 can be implemented.

As shown in FIG. 7, it is a schematic structural diagram of another embodiment of an image classification device provided by an embodiment of the present invention. The device may include a storage component 701 and a processing component 702. The storage component 701 stores one or more Computer instructions, the one or more computer instructions are for the processing component 702 to call and execute;

The processing component 702 is used to:

The device described in FIG. 7 can execute the image classification method described in the embodiment shown in FIG. 3, and its implementation principles and technical effects will not be repeated. The specific manners of processing components in the image classification device in the foregoing embodiment have been described in detail in the embodiment of the method, and detailed description will not be given here.

Those skilled in the art can clearly understand that, for the convenience and conciseness of the description, the specific working process of the system, device and unit described above can refer to the corresponding process in the foregoing method embodiment, which will not be repeated here.

The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, they may be located in One place, or it can be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the objectives of the solutions of the embodiments. Those of ordinary skill in the art can understand and implement without creative work.

Through the description of the above implementation manners, those skilled in the art can clearly understand that each implementation manner can be implemented by software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution essentially or the part that contributes to the existing technology can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM/RAM, magnetic A disc, an optical disc, etc., include a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiment.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the application, not to limit them; although the application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand: The technical solutions recorded in the foregoing embodiments are modified, or some of the technical features thereof are equivalently replaced; these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

A clothing image classification method, characterized in that it includes:

Obtain multiple clothing images;

For any clothing image, divide the clothing image into multiple regional images;

Extracting the color histogram features of the multiple regional images respectively;

Weighting the color histogram features of the multiple regional images to obtain the image features of the clothing image;

Clustering the multiple clothing images based on the image features of the multiple clothing images to obtain multiple image sets;

It is determined that the plurality of image sets respectively correspond to the number of images of at least one clothing image.
The method according to claim 1, further comprising:

Determine the clothing type where at least one clothing image corresponding to each of the multiple image sets is located.
The method according to claim 1, wherein after determining the number of images of the plurality of image sets corresponding to at least one clothing image, the method further comprises:

Sort the number of images to determine the output order of each image set;

At least one clothing image corresponding to the multiple image sets is sequentially outputted according to the respective output order of the multiple image sets.
The method according to claim 1, wherein said weighting the color histogram features of the multiple regional images to obtain the image features of the clothing image comprises:

The color histogram features of the multiple regional images are weighted according to the regional positions of the multiple regional images to obtain the image features of the clothing image.
The method according to claim 1, wherein said separately extracting the color histogram features of the plurality of regional images comprises:

Extracting the color histogram features of the multiple region images in multiple color channels respectively.
The method according to claim 5, wherein the extracting the color histogram features of the multiple region images in multiple color channels respectively comprises:

Extracting channel histogram features corresponding to each of the multiple color channels of the regional image for any one of the multiple regional images;

The channel histogram features corresponding to the multiple color channels are weighted to obtain the color histogram feature of the regional image.
The method according to claim 6, wherein, for any one of the multiple regional images, extracting the channel histogram features corresponding to the multiple color channels of the regional image respectively comprises:

For any one of the multiple regional images, assign each pixel of the regional image to the corresponding channel value of the multiple color channels;

Divide each color channel into multiple histogram statistical intervals;

For any color channel, count the channel values of the multiple pixels of the regional image in the color channel, and the number of distributions corresponding to multiple histogram statistical intervals in the color channel;

Based on the respective distribution numbers of the multiple histogram statistical intervals, the channel histogram characteristics of the region image in the color channel are determined.
The method according to claim 7, further comprising:

Multiple color channels formed by three color channels corresponding to at least one color space model are determined.
The method according to claim 8, wherein the at least one color space includes an RGB color space, an HSV color space, and a YcbCr color space;

Wherein, the RGB color space corresponds to the R color channel, the G color channel, and the B color channel, the HSV color space corresponds to the H color channel, the S color channel, and the V color channel, and the YcbCr color space corresponds to the Y color channel, Cb color channel, and Cv color. aisle.
The method of claim 1, wherein the color histogram feature includes a color feature matrix;

The performing weighting processing on the color histogram features of the multiple regional images to obtain the image features of the clothing image includes:

Perform horizontal or vertical matrix stitching processing on the respective color feature matrices of the multiple regional images to obtain the image features of the clothing image.
An image classification method, characterized in that it includes:

Acquire multiple images to be processed;

For any image to be processed, dividing the image to be processed into multiple regional images;

Extracting the color histogram features of the multiple regional images respectively;

Weighting the color histogram features of the multiple regional images to obtain the image feature of the image to be processed;

Classifying the multiple to-be-processed images based on the image features of the multiple to-be-processed images to obtain multiple image sets;

Determine the image type to which at least one image to be processed respectively corresponding to the multiple image sets belongs;

Determine the number of images of at least one image to be processed corresponding to each image category.
A clothing image classification device, characterized in that it comprises:

The first acquisition module is used to acquire multiple clothing images;

The first division module is configured to divide the clothing image into multiple regional images for any clothing image;

The first extraction module is configured to extract the color histogram features of the multiple regional images respectively;

The first weighting module is configured to perform weighting processing on the color histogram features of the multiple regional images to obtain the image features of the clothing image;

The first clustering module is configured to cluster the multiple clothing images based on the image features of the multiple clothing images to obtain multiple image sets;

The first determining module is configured to determine the number of images corresponding to at least one clothing image in each of the plurality of image sets.
An image classification device, characterized by comprising:

The second acquisition module is used to acquire multiple images to be processed;

The second division module is configured to divide the to-be-processed image into multiple regional images for any one to-be-processed image;

The second extraction module is used to extract the color histogram features of the multiple regional images respectively;

The second weighting module is configured to perform weighting processing on the color histogram features of the multiple regional images to obtain the image features of the image to be processed;

The second clustering module is configured to classify the multiple to-be-processed images based on the image features of the multiple to-be-processed images to obtain multiple image sets;

The second determining module is configured to determine the image type to which at least one image to be processed respectively corresponding to the multiple image sets belongs;

The third determining module is used to determine the image quantity of at least one image to be processed corresponding to each image category.
A clothing image classification device, characterized by comprising: a storage component and a processing component, the storage component stores one or more computer instructions, and the one or more computer instructions are called and executed by the processing component;

The processing component is used for:

Acquire multiple clothing images; for any clothing image, divide the clothing image into multiple regional images; extract the color histogram features of the multiple regional images respectively; combine the color histogram features of the multiple regional images Perform weighting processing to obtain image features of the clothing image; cluster the multiple clothing images based on the image features of the multiple clothing images to obtain multiple image sets; determine that the multiple image sets correspond to each other The number of images of at least one clothing image.
An image classification device, characterized by comprising: a storage component and a processing component, the storage component stores one or more computer instructions, and the one or more computer instructions are called and executed by the processing component;

The processing component is used for:

Acquire multiple images to be processed; for any image to be processed, divide the image to be processed into multiple regional images; extract the color histogram features of the multiple regional images respectively; The histogram features are weighted to obtain the image features of the image to be processed; based on the image features of the multiple images to be processed, the multiple images to be processed are classified to obtain multiple image sets; The image types to which at least one image to be processed respectively correspond to each image set; determine the number of images of at least one image to be processed corresponding to each image category.