CN114648668A - Method and apparatus for classifying attributes of target object, and computer-readable storage medium - Google Patents
Method and apparatus for classifying attributes of target object, and computer-readable storage medium Download PDFInfo
- Publication number
- CN114648668A CN114648668A CN202210539076.6A CN202210539076A CN114648668A CN 114648668 A CN114648668 A CN 114648668A CN 202210539076 A CN202210539076 A CN 202210539076A CN 114648668 A CN114648668 A CN 114648668A
- Authority
- CN
- China
- Prior art keywords
- component
- image
- target object
- attribute
- processed
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 56
- 238000003860 storage Methods 0.000 title claims abstract description 12
- 230000011218 segmentation Effects 0.000 claims abstract description 109
- 238000012549 training Methods 0.000 claims description 21
- 238000000605 extraction Methods 0.000 claims description 13
- 230000009471 action Effects 0.000 description 18
- 210000000245 forearm Anatomy 0.000 description 18
- 238000010586 diagram Methods 0.000 description 16
- 241001465754 Metazoa Species 0.000 description 8
- 238000013145 classification model Methods 0.000 description 5
- 238000009826 distribution Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 210000002414 leg Anatomy 0.000 description 5
- 210000000689 upper leg Anatomy 0.000 description 5
- 230000008569 process Effects 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 206010000117 Abnormal behaviour Diseases 0.000 description 3
- 244000309466 calf Species 0.000 description 3
- 238000004891 communication Methods 0.000 description 3
- 238000010276 construction Methods 0.000 description 3
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 241000282898 Sus scrofa Species 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 241000283690 Bos taurus Species 0.000 description 1
- 241001494479 Pecora Species 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000005452 bending Methods 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The application discloses a method and equipment for classifying attributes of target objects and a computer-readable storage medium. The method comprises the following steps: performing component segmentation on a target object in an image to be processed to obtain component segmentation information of the target object, wherein the component segmentation information represents position information of at least one component contained in the target object; obtaining an image area of a reference component in the image to be processed based on the component segmentation information, wherein the reference component is a component related to the attribute to be classified; and performing attribute classification based on the image area of the reference component to obtain the attribute category to be classified. By the method, the accuracy and convenience of attribute classification of the target object can be improved.
Description
Technical Field
The present application relates to the field of artificial intelligence technologies, and in particular, to a method and an apparatus for classifying attributes of a target object, and a computer-readable storage medium.
Background
And (4) performing attribute classification, namely performing attribute classification on the target object based on the image of the target object to obtain attribute category information of the target object. Attribute classification is required in many application scenarios. For example, workplaces of many industries (hygiene, financial, energy, logistics, and construction, etc.) have requirements for wear, and the wear attributes of a worker may be determined by attribute classification, and whether the wear of the worker is in compliance based on the wear attributes. In another example, to achieve better physical training, human actions are scored. The motion attribute of the human body can be determined through attribute classification, and the human body motion is scored based on the motion attribute.
However, the current method for classifying the attributes of the target object is complex to implement.
Disclosure of Invention
The application provides a method and equipment for classifying attributes of a target object and a computer-readable storage medium, which can improve the accuracy and convenience of attribute classification of the target object.
In order to solve the technical problem, the application adopts a technical scheme that: provided is a method for classifying attributes of a target object. The method comprises the following steps: performing component segmentation on a target object in an image to be processed to obtain component segmentation information of the target object, wherein the component segmentation information represents position information of at least one component contained in the target object; obtaining an image area of a reference component in the image to be processed based on the component segmentation information, wherein the reference component is a component related to the attribute to be classified; and performing attribute classification based on the image area of the reference component to obtain the attribute category to be classified.
In order to solve the technical problem, the application adopts a technical scheme that: provided is an attribute classification device for a target object. The attribute classification device includes: the device comprises a component segmentation module, a region determination module and an attribute classification module, wherein the component segmentation module is used for performing component segmentation on a target object in an image to be processed to obtain component segmentation information of the target object, and the component segmentation information represents position information of at least one component contained in the target object; the region determining module is used for obtaining an image region of a reference component in the image to be processed based on the component segmentation information, wherein the reference component is a component related to the attribute to be classified; and the attribute classification module is used for performing attribute classification based on the image area of the reference component to obtain the attribute class to be classified of the target object.
In order to solve the above technical problem, another technical solution adopted by the present application is: the attribute classification device of the target object comprises a processor and a memory connected with the processor, wherein the memory stores program instructions; the processor is configured to execute the program instructions stored by the memory to implement the above-described method.
In order to solve the above technical problem, the present application adopts another technical solution: there is provided a computer readable storage medium storing program instructions that when executed are capable of implementing the above method.
By the method, the attribute classification is not directly performed on the to-be-processed image of the target object, but the region of the reference component (component related to the to-be-classified attribute) of the target object in the to-be-processed image is analyzed, and the attribute classification is performed on the image region of the reference component to obtain the to-be-classified attribute category of the target object. Therefore, on one hand, the attribute classification method provided by the application is high in universality and applicable to any part of a target object, different attribute classification methods do not need to be set for realizing the attribute tasks to be classified related to different parts, and convenience of attribute classification is improved. On the other hand, because the image area of the reference component is obtained by analyzing the image to be processed based on the component segmentation, the interference caused by the background area except the target object in the image to be processed can be eliminated, the accuracy of the image area of the reference component is high, and the accuracy of the attribute category to be classified obtained by classifying the attribute of the image area of the reference component is high.
Drawings
FIG. 1 is a flowchart illustrating an embodiment of a method for classifying attributes of a target object according to the present application;
FIG. 2 is a schematic illustration of a to-be-processed image of a human body;
FIG. 3 is a schematic diagram of a split network;
FIG. 4 is a schematic diagram of part segmentation information for a human body;
FIG. 5 is a schematic diagram of a classification model;
FIG. 6 is a schematic illustration of image areas of the left and right forearm;
FIG. 7 is a schematic illustration of image regions of a left lower leg and a right lower leg;
FIG. 8 is a schematic diagram of an attribute classification system;
FIG. 9 is a schematic diagram illustrating an embodiment of an apparatus for classifying attributes of target objects according to the present application;
FIG. 10 is a schematic structural diagram of another embodiment of an attribute classification apparatus of a target object of the present application;
FIG. 11 is a schematic structural diagram of an embodiment of a computer-readable storage medium of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
The terms "first", "second" and "third" in this application are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implying any indication of the number of technical features indicated. Thus, a feature defined as "first," "second," or "third" may explicitly or implicitly include at least one of the feature. In the description of the present application, "plurality" means at least two, e.g., two, three, etc., unless explicitly specifically limited otherwise.
Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. Those skilled in the art will explicitly and implicitly appreciate that the embodiments described herein may be combined with other embodiments without conflict.
Fig. 1 is a flowchart illustrating an embodiment of a method for classifying attributes of a target object according to the present application. It should be noted that, if the result is substantially the same, the flow sequence shown in fig. 1 is not limited in this embodiment. As shown in fig. 1, the present embodiment may include:
s11: and carrying out component segmentation on the target object in the image to be processed to obtain component segmentation information of the target object.
The component segmentation information characterizes position information of at least one component comprised by the target object.
The execution subject of the embodiment of the method of the present application may be an attribute classification device, where the attribute classification device may be a terminal device (such as a computer or a smart camera), and the attribute classification device may also be a server (such as but not limited to a physical server, a cloud server, or a distributed server). The target object may be any one of a human body and an animal (pig, cow, sheep, etc.). The image to be processed may be an image including a target object, and the image to be processed is acquired by shooting an area where the target object is located through a camera.
The method comprises the steps of extracting features of an image to be processed to obtain image features; and performing component segmentation on the image features to obtain component segmentation information. The image features may be obtained by at least one of semantic feature extraction and spatial feature extraction. Namely, semantic feature extraction can be carried out on the image to be processed to obtain semantic features; and taking the semantic features as image features. Or, extracting spatial features of the image to be processed to obtain spatial features; and taking the spatial features as image features. Or, semantic feature extraction and spatial feature extraction can be carried out on the image to be processed, and semantic features and spatial features are correspondingly obtained; and fusing the semantic features and the spatial features to obtain image features. Under the condition that the semantic features and the spatial features are fused to obtain the image features, the obtained image features can better express the target object and are applied to component segmentation, the accuracy of component segmentation information can be improved, and the robustness of semantic segmentation to different application scenes is improved.
The spatial features include a shallow level of overall positional information of the components of the target object in the image to be processed. The semantic features include deeper level of detail location information, such as edge information, of each component of the target object in the image to be processed, relative to the spatial features. Referring to fig. 2, fig. 2 is a schematic diagram of a to-be-processed image of a human body, as shown in fig. 2, the head of the human body is provided with a safety helmet and a mask, the spatial features include overall position information of the head, and the semantic features include detailed position information of the head, such as edge information of the safety helmet and edge information of the mask. The semantic features can be extracted on the basis of the spatial features; alternatively, semantic features and spatial features are extracted using different feature extraction branches of the segmented network.
As illustrated with reference to fig. 3, fig. 3 is a schematic diagram of a segmentation network, and as shown in fig. 3, the segmentation network includes a spatial feature extraction branch, a semantic feature extraction branch, a feature fusion layer, and a component segmentation layer, where the spatial feature extraction branch sequentially uses three successive layers of convolution, batch normalization, and a ReLU activation function to process an image to be processed, so as to obtain spatial features whose sizes may be, but are not limited to, 1/8, 16/1 of the image to be processed. And after the semantic feature extraction branch performs light-weight feature extraction on the image to be processed, further processing the image by using a global average pooling layer to obtain semantic features. The feature fusion layer fuses the spatial features and the semantic features to obtain image features; the component division layer performs component division on the image features to obtain component division information.
The component segmentation information of the target object may include probability information of each pixel point belonging to each component in at least one component of the target object, for example, the component segmentation information of the target object may include probability distribution of each pixel point in the image to be processed, and the probability distribution of each pixel point is probability distribution of each component of the target object to which each pixel point belongs.
The component segmentation information of the target object may characterize a position of at least one component of the target object in the image to be processed. For each pixel point, the category of the pixel point is a component corresponding to the maximum value of the probability, for example, at least one component includes a component a1, a component a2 and a component a3, the probability that the pixel point a belongs to the component a1 is 0.3, the probability that the pixel point a belongs to the component a2 is 0.1, and the probability that the pixel point a belongs to the component a3 is 0.6, so that the pixel point belongs to the component a 3. The pixel points belonging to the part a1 commonly form the position of the part a1, the pixel points belonging to the part a2 commonly form the position of the part a2, and the pixel points belonging to the part a3 commonly form the position of the part a 3.
In some embodiments, the at least one component includes respective base components of the target object, and accordingly, the component segmentation information is the base component segmentation information. Specifically, in S11, the component segmentation information may be obtained by performing component segmentation on each base component of the target object in the image to be processed. The basic components comprise all components obtained by dividing the target object, and the component segmentation information represents the position of each basic component in the image to be processed.
Wherein the target object may be divided into a plurality of portions, each portion being a base part. For example, in one implementation, the human body may be divided into an upper body and a lower body according to the structure of the human body, and the upper body, the lower body, and the background may be a base member, respectively. For another example, the human body can be divided into 14 parts of a human head 1, a left hand 2, a left lower arm 3, a left upper arm 4, a trunk 5, a right hand 6, a right lower arm 7, a right upper arm 8, a left foot 9, a left lower leg 10, a left thigh 11, a right foot 12, a right lower leg 13, and a right thigh 14 according to the human body configuration with smaller fine granularity, and the 14 parts and the background 0 are respectively a basic part. Fig. 4 is a schematic diagram of the basic part segmentation information of the human body corresponding to fig. 2, and as shown in fig. 4, the part segmentation information of the human body represents the positions of the 15 basic parts in the image to be processed.
In some embodiments, the image to be processed is acquired for a target scene, the at least one component includes each specific component of the target object, and accordingly, the component segmentation information is specific component segmentation information. Specifically, in S11, each specific part of the target object in the image to be processed may be subjected to part segmentation to obtain specific part segmentation information. The specific component is a component associated with the target scene, the specific component comprising at least part of the base component. The target scene is associated with one or more attributes and, accordingly, the particular component includes one or more attribute-associated components. The specific part segmentation information characterizes the position of each specific part in the image to be processed.
Before the segmented network is applied to component segmentation, it may be trained until the trained segmented network meets network performance requirements. Specifically, a training image including a training object may be acquired, the training image being labeled with real part segmentation information; extracting the features of the training images to obtain the features of the training images; performing component segmentation on the training image features to obtain training component segmentation information; parameters of the segmentation network are adjusted based on a difference between the training component segmentation information and the real component segmentation information. The relevant description of the training component segmentation information and the real component segmentation information is similar to the component segmentation information of the application process, and is not repeated herein.
The training object and the target object are the same class of objects or similar class of objects. A loss function may be constructed based on a difference between the training component segmentation information and the real component segmentation information, and parameters of the segmentation network may be adjusted based on the loss function. For example, the loss function is a cross-entropy loss function
：
Wherein p isiAnd q isiRespectively representing the real probability distribution and the prediction probability distribution of the ith pixel point in the image to be processed.
S12: and obtaining an image area of the reference part in the image to be processed based on the part segmentation information.
The reference component is a component related to the attribute to be classified.
The attribute to be classified may be, but is not limited to, a wearing attribute (at least one of a clothing color attribute of wearing, a clothing style attribute of wearing, a manner attribute of wearing clothing), an action attribute (at least one of a leg action attribute, a head action attribute). If the attribute to be classified is a wearing attribute, the reference component may be a component related to the wearing attribute. If the attribute to be classified is an action attribute, the reference component may be a component of the target object related to the action attribute. For example, the image to be processed is collected for a target scene, the attributes associated with the target scene include a wearing attribute and an action attribute, and the attributes to be classified include a wearing attribute and an action attribute.
If the component division information is the base component information in S11, a reference component may be determined from among the base components based on the base component division information before S12. Specifically, the identification information of the reference component is known, and the reference component can be determined from the respective base components based on the known identification information. Or, the attribute associated with each basic component is known, the identification information of the attribute to be classified is known, and the component with the associated attribute as the attribute to be classified can be determined from each basic component as the reference component based on the identification information of the attribute to be classified.
If the component division information is the specific component information in S11, a reference component may be determined from among the specific components before S12. The method comprises the following steps that when a target scene to which a specific component belongs only corresponds to one attribute and the attribute is an attribute to be classified, the specific component can be directly determined as a reference component; in the case where the target scene to which the specific component belongs corresponds to a plurality of attributes, the reference component related to each attribute to be classified may be determined from each specific component based on the identification information of the reference component or the identification information of the attribute to be classified.
In S12, the position of the reference component characterized by the component segmentation information may be in a corresponding connected domain in the image to be processed as an image region of the reference component. Or, based on the part segmentation information, determining a rectangular image region containing the reference part from the image to be processed; the pixel values outside the reference part in the rectangular image area are set to preset pixel values (e.g., 0) to obtain an image area of the reference part.
S13: and performing attribute classification based on the image area of the reference component to obtain the attribute class to be classified of the target object.
The attributes involved in S13 are classified into classes for the attributes to be classified, and the attribute class to be classified is a class of the attribute to be classified of the target object.
The attributes to be classified may correspond to a plurality of categories. For example, if the attribute to be classified is a color attribute of the worn garment, the attribute corresponds to a plurality of color categories (e.g., at least two of white, black, orange, red, green, or white). If the attribute to be classified is the clothing style attribute of wearing, corresponding to a plurality of style categories (such as at least two of formal clothing, non-formal clothing, skirt, trousers, short sleeves and the like); if the attribute to be classified is the mode attribute of the clothes to be worn, corresponding to a plurality of mode categories (such as sleeve drawing and sleeve not drawing); if the attribute to be classified is a leg motion attribute, the attribute corresponds to a plurality of leg motion categories (such as bending and unbending).
It can be understood that if the attribute classification is directly performed on the image to be processed of the target object, in order to ensure the accuracy of the attribute classification result, different attribute classification methods are required to be implemented for the task of classifying the attributes to be classified related to different components.
Through the implementation of the embodiment, the attribute classification is not directly performed on the to-be-processed image of the target object, but the image area of the reference component (component related to the to-be-classified attribute) of the target object in the to-be-processed image is analyzed, and the attribute classification is performed on the image area of the reference component to obtain the to-be-classified attribute category of the target object. Therefore, on one hand, the attribute classification method provided by the application is high in universality and applicable to any part of the target object, different attribute classification methods do not need to be set for realizing the attribute tasks to be classified related to different parts, and convenience in attribute classification of the target object is improved. On the other hand, because the image area of the reference component is obtained by analyzing the image to be processed based on the component segmentation, the interference caused by the background area except the target object in the image to be processed can be eliminated, the accuracy of the image area of the reference component is high, and the accuracy of the attribute category to be classified obtained by classifying the attribute of the image area of the reference component is high.
Further, the method for classifying the attributes of the target object provided by the application can be realized according to a trained classification model. Referring to fig. 5 in combination, the classification model includes a segmentation network for component segmentation and an attribute classification network for attribute classification of the image region of the reference component. Based on the method, a segmentation network can be utilized to segment the target object in the image to be processed to obtain segment segmentation information; determining an image area of a reference part in the image to be processed based on the part segmentation information; and performing attribute classification on the image area of the reference component by using an attribute classification network to obtain the attribute category to be classified of the target object.
It can be understood that, in the case that the attribute classification method of the target object is implemented by means of the classification model, it is not necessary to deploy different classification models in order to implement the tasks of the attributes to be classified related to different components, thereby reducing the complexity of attribute classification.
For convenience of understanding, several application scenarios of the attribute classification method of the target object of the present application are described in detail as follows:
application scenario 1: and identifying the human body wearing compliance. In places with wearing requirements, a human body needs to wear the compliance. For example, in workplaces of the health industry, catering workers need to wear white work clothes, cleaning workers need to wear orange work clothes, in workplaces of the financial industry, banking workers need to wear formal dresses, and in workplaces of the energy industry, the logistics industry, and the construction industry, workers also need to wear corresponding work clothes. The wearing compliance may include a color compliance of the garment being worn, a style compliance of the garment being worn, a manner compliance of the garment being worn (e.g., failure to roll off a sleeve, pull on a leg, etc. causing an area of skin to be exposed). Therefore, the attribute classification can be performed on the human body according to the wearing attributes (the color attribute of the worn clothes, the style attribute of the worn clothes, and the mode attribute of the worn clothes), and the wearing category of the human body, namely, what the color of the worn clothes is, what the style of the worn clothes is, and what the mode of the worn clothes is, can be determined; determining whether the wearing of the human body is in compliance based on the attribute classification result.
Application scenario 2: and (4) scoring the human body action. In the process of practicing sports such as dancing and playing, the actions of the human body are scored. The attribute classification can be carried out on the human body according to the action attribute, and the action category of the human body is determined; the motion of the human body is scored based on the motion category of the human body.
Application scenario 3: and (4) detecting abnormal behaviors of the animals. Without human supervision, animal habitation areas are prone to abnormal behavior, such as swine bites. Therefore, in the animal social region, the attributes of the animals can be classified according to the action attributes, and the action types of the animals are determined; determining whether the animal has abnormal behavior based on the action category of the animal.
Based on the enumerated application scenes, in the case that the image to be processed is collected for the target scene, after determining the attribute category to be classified of the target object through S11-S13, determining whether the attribute to be classified of the target object conforms to the attribute specification associated with the target scene based on the attribute category to be classified.
For example, the image to be processed is acquired in a construction scene, the wearing specification includes wearing a safety helmet and wearing a dark blue garment, the target object is a human body, the attribute to be classified is a wearing attribute, and whether the wearing of the human body is in compliance (wearing the safety helmet and wearing the dark blue garment) is determined based on the wearing attribute category. For another example, the to-be-processed image is acquired during the process of performing the sports, the target object is a human body, the to-be-classified attribute is an action attribute, and whether the action attribute is in compliance is determined based on the action attribute category.
The method for classifying the attributes of the target object provided by the application is explained in detail below in combination with an application scenario of whether the mode of wearing the clothes by the human body is compliant or not:
when the manner of wearing the garment is in compliance, the skin area of the human body is not exposed, and the garment appears as a body without rolling up the sleeves and the legs. The body parts associated with the manner of wearing the garment are therefore the left forearm, the right forearm, the left calf, the left thigh.
1) An image to be processed of a human body is acquired, and an example of the image to be processed is shown in fig. 2.
2) The human body in the image to be processed is subjected to part segmentation to obtain part segmentation information (a background 0, a head 1, a left hand 2, a left forearm 3, a left forearm 4, a torso 5, a right hand 6, a right forearm 7, a right forearm 8, a left foot 9, a left shank 10, a left thigh 11, a right foot 12, a right shank 13 and a right thigh 14) of the human body, wherein the left forearm 3, the left forearm 4, the left shank 10 and the right shank 13 are reference parts, and an example of the part segmentation information is shown in fig. 4.
3) Rectangular image areas of the left forearm 3, the right forearm 7, the left lower leg 10 and the right lower leg 13 in the image to be processed are determined and are respectively used as image areas of the left forearm 3, the right forearm 7, the left lower leg 10 and the right lower leg 13.
4) The image areas of the left forearm 7, the left lower leg 10 and the right lower leg 13 are filled with black pixels at positions other than the left forearm 3 in the image area of the left forearm 3. Fig. 6 is a schematic diagram showing the image region of the left lower arm 3 after filling, and the image region of the right lower arm 7 after filling on the left side, and fig. 7 is a schematic diagram showing the image region of the left lower leg 10 after filling on the left side, and the image region of the right lower leg 13 after filling on the right side.
5) The image regions of the left forearm 3, right forearm 7, left calf 10, right calf 13 are respectively subjected to attribute classification to determine whether the wearing is compliant.
The attribute classification method of the target object is realized according to an attribute classification system. The attribute classification system will be described below with reference to an application scenario in which a manner of wearing a garment by a human body is compliant or not.
Fig. 8 is a schematic diagram of an attribute classification system, which may include an image pickup end and an attribute classification device/apparatus of a target object, as shown in fig. 8. The camera end can be an independent camera or equipment containing the camera, and the attribute classification equipment/device can be any equipment/device with attribute classification capability. And the camera terminal and the attribute classification equipment/device establish communication connection. The camera is arranged in a workplace with wearing requirements on a human body and used for acquiring an image to be processed of the human body; the attribute classification equipment/device is used for acquiring the image to be processed from the camera and performing attribute classification so as to determine the attribute category to be classified of the target object.
Fig. 9 is a schematic structural diagram of an embodiment of an attribute classification apparatus of a target object of the present application. As shown in fig. 9, the attribute classifying apparatus includes a component dividing module 11, a region determining module 12, and an attribute classifying module 13.
The component segmentation module 11 may be configured to perform component segmentation on a target object in an image to be processed to obtain component segmentation information of the target object, where the component segmentation information represents position information of at least one component included in the target object. The region determining module 12 may be configured to obtain an image region of a reference component in the image to be processed based on the component segmentation information, where the reference component is a component related to the attribute to be classified; the attribute classification module 13 may be configured to perform attribute classification based on the image region of the reference component, so as to obtain an attribute class to be classified of the target object.
By implementing the embodiment, the method and the device for classifying the attributes of the target object do not directly classify the attributes of the target object to be processed, but analyze the area of the reference component (component related to the attributes to be classified) of the target object in the target object to be processed by using the component segmentation module and the area determination module, and classify the attributes of the image area of the reference component by using the attribute classification module to obtain the attribute category to be classified of the target object. Therefore, on one hand, the attribute classification method of the attribute classification device is high in universality and applicable to any part of a target object, different attribute classification methods do not need to be set for realizing the attribute tasks to be classified related to different parts, and the complexity of attribute classification is reduced. On the other hand, because the image area of the reference component is obtained by analyzing the image to be processed based on the component segmentation, the interference caused by the background area except the target object in the image to be processed can be eliminated, the accuracy of the image area of the reference component is high, and the accuracy of the attribute category to be classified obtained by classifying the attribute of the image area of the reference component is high.
Fig. 10 is a schematic structural diagram of another embodiment of the attribute classification device of the target object of the present application. As shown in fig. 10, the attribute classification apparatus includes a processor 21, and a memory 22 coupled to the processor 21.
Wherein the memory 22 stores program instructions for implementing the method of any of the above embodiments; processor 21 is operative to execute program instructions stored by memory 22 to implement the steps of the above-described method embodiments. The processor 21 may also be referred to as a CPU (Central Processing Unit). The processor 21 may be an integrated circuit chip having signal processing capabilities. The processor 21 may also be a general purpose processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
FIG. 11 is a schematic structural diagram of an embodiment of a computer-readable storage medium of the present application. As shown in fig. 11, the computer-readable storage medium 30 of the embodiment of the present application stores program instructions 31, and when executed, the program instructions 31 implement the method provided by the above-mentioned embodiment of the present application. The program instructions 31 may form a program file stored in the computer-readable storage medium 30 in the form of a software product, so as to enable a computer device (which may be a personal computer, a server, or a network device) or a processor (processor) to execute all or part of the steps of the methods according to the embodiments of the present application. And the aforementioned computer-readable storage medium 30 includes: various media capable of storing program codes, such as a usb disk, a mobile hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, or terminal devices, such as a computer, a server, a mobile phone, and a tablet.
In the several embodiments provided in the present application, it should be understood that the disclosed system, apparatus and method may be implemented in other manners. For example, the above-described apparatus embodiments are merely illustrative, and for example, a division of a unit is merely a logical division, and an actual implementation may have another division, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit may be implemented in the form of hardware, or may also be implemented in the form of a software functional unit. The above embodiments are merely examples and are not intended to limit the scope of the present disclosure, and all modifications, equivalents, and flow charts using the contents of the specification and drawings of the present disclosure or those directly or indirectly applied to other related technical fields are intended to be included in the scope of the present disclosure.
Claims (12)
1. A method for classifying attributes of a target object, comprising:
performing component segmentation on a target object in an image to be processed to obtain component segmentation information of the target object, wherein the component segmentation information represents position information of at least one component contained in the target object;
obtaining an image area of a reference component in the image to be processed based on the component segmentation information, wherein the reference component is a component related to the attribute to be classified;
and performing attribute classification based on the image area of the reference component to obtain the attribute class to be classified of the target object.
2. The method according to claim 1, wherein performing component segmentation on a target object in the image to be processed to obtain component segmentation information of the target object comprises:
performing feature extraction on the image to be processed to obtain image features;
and carrying out component segmentation on the image features to obtain the component segmentation information.
3. The method according to claim 2, wherein the performing feature extraction on the image to be processed to obtain image features comprises:
respectively extracting semantic features and spatial features of the image to be processed to correspondingly obtain the semantic features and the spatial features;
and fusing the semantic features and the spatial features to obtain the image features.
4. The method according to claim 2, wherein the component segmentation of the target object in the image to be processed to obtain the component segmentation information of the target is implemented based on a segmentation network, and the training of the segmentation network comprises:
acquiring a training image containing the training object, wherein the training image is marked with real part segmentation information;
extracting the features of the training images to obtain the features of the training images;
performing component segmentation on the training image features to obtain training component segmentation information;
adjusting parameters of the segmentation network based on a difference between the training component segmentation information and the real component segmentation information.
5. The method according to claim 1, wherein performing component segmentation on a target object in the image to be processed to obtain component segmentation information of the target object comprises:
performing component segmentation on each basic component of the target object in the image to be processed to obtain basic component segmentation information;
before obtaining the image region of the reference component in the image to be processed based on the component segmentation information, the method further includes:
the reference component is determined from the respective base components based on the base component division information.
6. The method of claim 1, wherein the image to be processed is acquired for a target scene; the method for performing component segmentation on the target object in the image to be processed to obtain component segmentation information of the target object includes:
performing component segmentation on each specific component of the target object in the image to be processed to obtain specific component segmentation information, wherein the specific component is a component associated with the target scene;
before obtaining the image area of the reference component in the image to be processed based on the component segmentation information, the method further includes:
the reference component is determined from the specific components based on the specific component division information.
7. The method according to claim 1, wherein the deriving an image region of a reference component in the target image based on the component segmentation information comprises:
determining a rectangular image region including the reference component from the image to be processed based on the component division information;
and setting pixel values outside the reference component in the rectangular image area as preset pixel values to obtain an image area of the reference component.
8. The method of claim 1, wherein the image to be processed is captured for a target scene, and wherein the attribute to be classified comprises at least one of a wear attribute and a motion attribute associated with the target scene.
9. The method of claim 1, wherein the image to be processed is acquired for a target scene, and after the attribute classification based on the image region of the reference component is performed to obtain the attribute class to be classified of the target object, the method further comprises:
and determining whether the attribute to be classified of the target object meets the attribute specification associated with the target scene or not based on the attribute category to be classified.
10. An apparatus for classifying a property of a target object, comprising:
the device comprises a component segmentation module, a component analysis module and a component analysis module, wherein the component segmentation module is used for performing component segmentation on a target object in an image to be processed to obtain component segmentation information of the target object, and the component segmentation information represents position information of at least one component contained in the target object;
the region determining module is used for obtaining an image region of a reference component in the image to be processed based on the component segmentation information, wherein the reference component is a component related to the attribute to be classified;
and the attribute classification module is used for performing attribute classification based on the image area of the reference component to obtain the attribute category to be classified of the target object.
11. An apparatus for classifying attributes of a target object, comprising a processor, a memory coupled to the processor, wherein,
the memory stores program instructions;
the processor is configured to execute the program instructions stored by the memory to implement the method of any of claims 1-9.
12. A computer-readable storage medium, characterized in that it stores program instructions executable by a processor, which when executed, implement the method of any one of claims 1-9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210539076.6A CN114648668A (en) | 2022-05-18 | 2022-05-18 | Method and apparatus for classifying attributes of target object, and computer-readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210539076.6A CN114648668A (en) | 2022-05-18 | 2022-05-18 | Method and apparatus for classifying attributes of target object, and computer-readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114648668A true CN114648668A (en) | 2022-06-21 |
Family
ID=81997257
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210539076.6A Pending CN114648668A (en) | 2022-05-18 | 2022-05-18 | Method and apparatus for classifying attributes of target object, and computer-readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114648668A (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109614925A (en) * | 2017-12-07 | 2019-04-12 | 深圳市商汤科技有限公司 | Dress ornament attribute recognition approach and device, electronic equipment, storage medium |
| CN110188817A (en) * | 2019-05-28 | 2019-08-30 | 厦门大学 | A kind of real-time high-performance street view image semantic segmentation method based on deep learning |
| CN110544217A (en) * | 2019-08-30 | 2019-12-06 | 深圳市商汤科技有限公司 | image processing method and device, electronic equipment and storage medium |
| CN110930409A (en) * | 2019-10-18 | 2020-03-27 | 电子科技大学 | Salt body semantic segmentation method based on deep learning and semantic segmentation model |
| CN112906706A (en) * | 2021-03-31 | 2021-06-04 | 西南科技大学 | Improved image semantic segmentation method based on coder-decoder |
| CN113449735A (en) * | 2021-07-15 | 2021-09-28 | 北京科技大学 | Semantic segmentation method and device for superpixel segmentation |
| CN113658200A (en) * | 2021-07-29 | 2021-11-16 | 东北大学 | Edge perception image semantic segmentation method based on self-adaptive feature fusion |
-
2022
- 2022-05-18 CN CN202210539076.6A patent/CN114648668A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109614925A (en) * | 2017-12-07 | 2019-04-12 | 深圳市商汤科技有限公司 | Dress ornament attribute recognition approach and device, electronic equipment, storage medium |
| CN110188817A (en) * | 2019-05-28 | 2019-08-30 | 厦门大学 | A kind of real-time high-performance street view image semantic segmentation method based on deep learning |
| CN110544217A (en) * | 2019-08-30 | 2019-12-06 | 深圳市商汤科技有限公司 | image processing method and device, electronic equipment and storage medium |
| CN110930409A (en) * | 2019-10-18 | 2020-03-27 | 电子科技大学 | Salt body semantic segmentation method based on deep learning and semantic segmentation model |
| CN112906706A (en) * | 2021-03-31 | 2021-06-04 | 西南科技大学 | Improved image semantic segmentation method based on coder-decoder |
| CN113449735A (en) * | 2021-07-15 | 2021-09-28 | 北京科技大学 | Semantic segmentation method and device for superpixel segmentation |
| CN113658200A (en) * | 2021-07-29 | 2021-11-16 | 东北大学 | Edge perception image semantic segmentation method based on self-adaptive feature fusion |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP7132387B2 (en) | Image processing device, image processing method and program | |
| CN107679448B (en) | Eyeball action-analysing method, device and storage medium | |
| US10423848B2 (en) | Method, system, and computer-readable recording medium for long-distance person identification | |
| US9020250B2 (en) | Methods and systems for building a universal dress style learner | |
| CN109299658B (en) | Face detection method, face image rendering device and storage medium | |
| US20160092726A1 (en) | Using gestures to train hand detection in ego-centric video | |
| CN107679449B (en) | Lip motion method for catching, device and storage medium | |
| WO2021174941A1 (en) | Physical attribute recognition method, system, computer device, and storage medium | |
| CN111666920B (en) | Target article wearing detection method and device, storage medium and electronic device | |
| CN107633206B (en) | Eyeball motion capture method, device and storage medium | |
| US11908157B2 (en) | Image processing device, image processing method, and recording medium in which program is stored | |
| JP2010262425A (en) | Computer execution method for recognizing and classifying clothes | |
| JP2015106300A (en) | Image search device, control method for image search device, and program | |
| CN112434562A (en) | Method and device for detecting wearing state of mask, electronic device and storage medium | |
| CN110069983A (en) | Vivo identification method, device, terminal and readable medium based on display medium | |
| CN113657195A (en) | Face image recognition method, face image recognition equipment, electronic device and storage medium | |
| CN112507978B (en) | Person attribute identification method, device, equipment and medium | |
| CN108875654A (en) | A kind of face characteristic acquisition method and device | |
| CN113392741A (en) | Video clip extraction method and device, electronic equipment and storage medium | |
| CN111159476B (en) | Target object searching method and device, computer equipment and storage medium | |
| WO2022111271A1 (en) | Clothing standardization detection method and apparatus | |
| Sudha et al. | A fast and robust emotion recognition system for real-world mobile phone data | |
| CN111126102A (en) | Personnel searching method and device and image processing equipment | |
| CN114648668A (en) | Method and apparatus for classifying attributes of target object, and computer-readable storage medium | |
| JP6090162B2 (en) | Site separation position extraction apparatus, program, and method |
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: 20220621 |