Image classification method and related equipment

Abstract

The invention discloses an image classification method and related equipment, and aims to solve the problem of low claim settlement processing efficiency caused by long time consumption of manual image classification. The image classification method comprises the steps of obtaining an image of a breeding object applying for claim settlement; based on a pre-trained target detection model, detecting each target part of the breeding object in the image and the position of each target part in the target image; and determining a classification result of the target image based on the relative position relationship between the first part and the second part. The method provided by the embodiment of the invention is realized through the image classification device, compared with a manual classification mode, the classification time is shortened, the processing efficiency of the claim settlement case is improved, and particularly, the processing efficiency can be remarkably improved when a large number of images are classified.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to an image classification method and related equipment. Background Art

[0002] When the insured pig dies within the scope of insurance liability during the insurance period, the insured can apply for compensation. When applying for compensation, you need to provide at least an uncut ear image (or unprocessed image) and a cut ear image (or processed image) of the dead pig. Among them, the uncut ear image refers to the complete pig body image taken of the dead pig; the cut ear image refers to the dead pig's ears being cut off, the cut ears are placed on the dead pig's abdomen, and then the whole body image of the pig is taken.

[0003] Currently, during the claims processing, manual image classification is required for the received uncut and chopped ear images, that is, to distinguish whether the current image is a chopped ear image or an uncut ear image. Manual image classification is time-consuming and leads to low claims processing efficiency. This problem is more prominent when a large number of images to be classified are received at the same time.

[0004] Similarly, similar situations also occur when processing claims for other farmed animals besides pigs, such as cattle and sheep. For example, when applying for claims for the death of a live cattle, it is necessary to provide images of the dead cattle with ears cut off (or other body parts cut off) and images of the ears not cut off. Manual image classification also takes a long time, resulting in low claims processing efficiency. Summary of the invention

[0005] The embodiments of the present application provide an image classification method and related equipment to solve the problem that manual image classification is time-consuming and leads to low claims processing efficiency.

[0006] The present application embodiment adopts the following technical solutions:

[0007] An image classification method, comprising:

[0008] Obtain images of the breeding objects for which claims are being applied for;

[0009] Based on a pre-trained target detection model, detecting each target part of the farmed object in the image and the position of each target part in the target image; each target part includes: a first part and a second part of the body of the farmed object;

[0010] Based on the relative position relationship between the first part and the second part, a classification result of the target image is determined.

[0011] An image classification device, comprising:

[0012] An image acquisition unit, used to acquire an image of the breeding object for which compensation is applied;

[0013] A target part detection unit detects target parts of the farmed object in the image and positions of the target parts in the target image based on a pre-trained target detection model; the target parts include: a first part and a second part of the farmed object's body;

[0014] The image classification unit determines a classification result of the target image based on the relative position relationship between the first part and the second part.

[0015] A computing device, comprising: a memory and a processor, wherein:

[0016] The memory is used to store computer programs;

[0017] The processor is coupled to the memory and is used to execute the computer program stored in the memory to perform the above method.

[0018] A computer-readable storage medium storing a computer program, wherein the computer program can implement the above method when executed by a computer.

[0019] At least one of the above technical solutions adopted in the embodiments of the present application can achieve the following beneficial effects:

[0020] According to the method of the embodiment of the present application, each target part of the farmed object in the image and the position of each target part in the target image are detected by the target detection model; based on the relative position relationship of the two parts, the classification result of the target image is determined. Because in the normal living state of the farmed object, the relative position of the two target parts in the body follows a certain rule; in the claim stage, in order to show that the farmed object meets the claim conditions, one of its target parts will be processed, thereby destroying the relative position relationship between the target objects. Therefore, based on the relative position relationship between the two target objects, the classification result of the target image can be determined. The method of the embodiment of the present application is implemented by an image classification device. Compared with the manual classification method, it shortens the classification time and improves the processing efficiency of claims cases. In particular, when processing the classification of a large number of images, it can significantly improve the processing efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0022] Figure 1 A specific implementation flow chart of an image classification method provided in an embodiment of the present application;

[0023] Figure 2a This is a schematic diagram of the detection result of an uncut ear image identified by a target detection model in a specific implementation of the image classification method of the present application;

[0024] Figure 2b This is a schematic diagram of the detection result of a cut-ear image identified by a target detection model in a specific implementation of the image classification method of the present application;

[0025] Figure 3 This is a specific implementation flow chart of the orientation judgment of the breeding object in a specific implementation of the image classification method of the present application;

[0026] Figure 4 This is a specific implementation flow chart of image classification in a specific implementation of the image classification method of the present application;

[0027] Figure 5 A schematic diagram of the specific structure of an image classification device provided in an embodiment of the present application;

[0028] Figure 6 A schematic diagram of the specific structure of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0030] Those skilled in the art may appreciate that, with the development of technology and the emergence of new scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0031] The terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and need not be used to describe a specific order or sequential order. It should be understood that the terms used in this way can be interchangeable under appropriate circumstances, which is only to describe the distinction mode adopted by the objects of the same attributes when describing in the embodiments of the present application. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, so that the process, method, system, product or equipment comprising a series of units need not be limited to those units, but may include other units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0032] In order to solve the problem that manual image classification is time-consuming and leads to low efficiency in claims processing, an embodiment of the present application provides an image classification method.

[0033] The executor of the method may be any computing device that can implement the method, such as a server, a mobile phone, a personal computer, a smart wearable device, a smart robot, and the like.

[0034] In addition, the execution order of different steps is not limited in the embodiment of the present application. When using the method provided in the embodiment of the present application, the execution order of different steps can be adjusted according to actual needs.

[0035] For ease of description, the method provided in the embodiment of the present application is described in detail below by taking an image classification device as the executor of the method as an example.

[0036] like Figure 1 As shown, it is a specific implementation flow chart of an image classification method provided in an embodiment of the present application, including the following steps 11 to 13:

[0037] Step 11, obtaining an image of the breeding object for which compensation is applied.

[0038] The breeding objects in the embodiments of the present application refer to animals raised artificially (by breeders), such as pigs, cattle, sheep, etc.

[0039] The images include unprocessed images and processed images.

[0040] Among them, the unprocessed image refers to the image obtained by taking a photo of the complete body of the breeding object; usually after the breeding object meets the conditions for applying for compensation, such as the death of the breeding object, the complete body of the breeding object is photographed to obtain the unprocessed image.

[0041] A processed image refers to a processed image obtained by taking a picture of the entire body of a farmed object, performing a specific treatment on a certain part of the body of the farmed object, and photographing the entire body of the farmed object (including the treated part) after taking an image of the complete body of the farmed object. Taking a pig as an example, if the certain part of the farmed object is a pig ear, the processed image may be an image of the pig body taken after the pig ear is cut off (corresponding to the above-mentioned specific treatment) and placed in the abdomen of a dead pig (or other selected location).

[0042] Step 12: Detect target parts of the farmed object in the image and the positions of the target parts in the target image based on a pre-trained target detection model.

[0043] The object detection model is a deep learning model in the field of computer vision. It aims to identify and locate the category and location of the target object from an image or video. Commonly used object detection models include the Faster R-CNN (Faster Region-based Convolutional Network) detection model, the YOLO (You Only Look Once) detection model, and the CenterNet (center point network) detection model, etc.

[0044] In an embodiment of the present application, one of the above-mentioned target detection models can be selected to detect the target parts of the breeding object and the positions of each target part in the target image.

[0045] By performing supervised training on the target detection model, a trained target detection model is obtained. During the training process, images with labeled body parts of the farmed object can be used as training samples.

[0046] When using training samples to train the target detection model, images with labeled farmed objects are used as inputs of the target detection model, and the body parts of the farmed objects and the locations of the parts in the same image are used as detection results of the target detection model.

[0047] In each round of iterative training in the specific training process, images of the breeding objects are input into the target detection model to be trained, and the value of the loss function is determined based on the detection results of the target detection model and the above-mentioned true detection results; based on the principle of minimizing the value of the loss function, the parameters of the target detection model are adjusted until the target detection model converges.

[0048] In this step, before the acquired image is input into the target detection model, the acquired image is also oriented. Specifically, it includes:

[0049] The orientation of the image is determined based on the image's aspect ratio. If the image is horizontal, the horizontal image is rotated 90 degrees clockwise (or counterclockwise).

[0050] In this embodiment, the target parts detected based on the target detection model include: a first part and a second part of the body of the farmed object. The first part and the second part are different parts of the farmed object that are visible on the outside, such as two of the parts of the pig body, pig head, pig ear, pig leg, pig head, etc.

[0051] The target detection model uses a bounding box to frame each target part of the identified aquaculture object to locate the location of the target part; at the same time, the target detection model also marks the confidence of the target part corresponding to the current bounding box on each bounding box. In the target detection task, the confidence is usually used to indicate the degree of certainty of the target detection model on whether the detected target actually exists. Among them, the bounding box is identified by the target detection model and is used to locate the location of each target part of the aquaculture object in the target image.

[0052] Typically, the target detection model will give at least one corresponding candidate box for the same detection target, and mark different confidence levels for different candidate boxes.

[0053] In this embodiment, based on the pre-trained target detection model, detecting each target part of the breeding object in the image and the position of each target part in the target image includes:

[0054] Based on a pre-trained target detection model, at least one candidate frame of a single target part of the breeding object in the target image is obtained; based on a preset selection rule, a bounding box for locating the current target part is selected from the at least one candidate frame.

[0055] The selection rules include:

[0056] Rule 1: If the target detection model fails to identify the breeding object, the candidate box with the largest area is selected from the identified candidate boxes containing the current target part as the bounding box for locating the current target part.

[0057] Rule 2: If the target detection model recognizes a breeding object, then from the identified candidate frames containing the current target part, the candidate frame with the largest overlapping area with the bounding box where the breeding object is located is selected, and the ratio of the overlapping area to the area of ​​the candidate frame containing the current target part is greater than the fifth threshold, as the bounding box for locating the current target part.

[0058] In the above rule 2, the value of the fifth threshold is 0.9. This value is selected according to the positional relationship between the cultured object and each part in the sample. If the overlapping area is less than 0.9 of the target part area, it is considered to be another part of the cultured object or a misjudgment.

[0059] The above rules 1 and 2 apply to the unprocessed body parts of the breeding object. The above selected bounding box is 1. If more than 1 bounding box is selected for the same body part according to the above rules, the bounding box with the largest confidence is selected from the multiple selected bounding boxes as the bounding box of the positioning target position and retained.

[0060] For the target part of the breeding object that has been processed, the following selection rule 3 is used to select the bounding box: if the breeding object body is not detected, all the identified candidate frames containing the target part are used as the bounding box for locating the target part; if the target detection model identifies the breeding object body, all the candidate frames within the range of the bounding box where the breeding object body is located are selected from the identified candidate frames containing the target part as the bounding box for locating the target part.

[0061] Taking pigs as breeding objects and ear cutting as treatment method as an example, the detection results of uncut ear images are determined ( Figure 2a ) and the detection results of the cut ear image ( Figure 2b shown).

[0062] Figure 2a In the figure, the target detection model identified the target parts such as the pig, head, ear, and leg, and the corresponding confidence levels were 0.98, 0.95, 0.67, (0.90, 0.88, 0.89, and 0.92), respectively; among them, 0.90, 0.88, 0.89, and 0.92 are the confidence levels of the four legs, respectively. Figure 2b In the figure, the target detection model identified the target parts such as the pig, head, ear, and leg, and the corresponding confidence levels were 0.97, 0.89, (0.95, 0.68), and (0.87, 0.93, 0.90, and 0.91), respectively. Among them, 0.87, 0.93, 0.90, and 0.91 are the confidence levels of the four legs, respectively.

[0063] The embodiment of the present application further includes: determining the orientation of the farmed object in the target image based on the position of each target part in the target image. The orientation of the farmed object may be the orientation of the head of the farmed object.

[0064] In one implementation, the orientation of the farming object includes the following steps 121 to 123:

[0065] Step 121, obtaining the center point of the bounding box where each target part is located;

[0066] Step 122: If there is a head in the target part, the vertical orientation of the head is determined based on the center point of the bounding box where the head is located and the center point of the bounding box where the breeding object body is located.

[0067] Specifically, if the ordinate of the center point of the bounding box where the head is located is smaller than the ordinate of the center point of the bounding box where the breeding object is located, the head is at the bottom, otherwise, the head is at the top.

[0068] Step 123, determining the orientation of the farming object based on the orientation of the head in the vertical direction, the center point of the bounding box where the head is located, and the center point of the bounding box where the legs are located in the target part.

[0069] Specifically, when the head is on top, if the horizontal coordinate of the center point of the bounding box where the head is located is greater than the horizontal coordinate of the center point of the bounding box where the legs are located, the breeding object body faces the left; otherwise, the breeding object body faces the right.

[0070] When the head is at the bottom, if the horizontal coordinate of the center point of the bounding box where the head is located is smaller than the horizontal coordinate of the center point of the bounding box where the legs are located, the breeding object body faces the left side; otherwise, the breeding object body faces the right side.

[0071] In addition, if the body of the farmed object is not detected in the target object, the coordinate system is determined with the detected head as the origin. If the position of the legs is detected in the first or third quadrant, the body of the farmed object is facing left; otherwise, the body of the farmed object is facing right.

[0072] When the head is not detected, if the legs are detected on the left side of the target image, the body of the cultivated object is facing the left; otherwise, the body of the cultivated object is facing the right side.

[0073] It should be noted that when judging the orientation of the breeding object, the judged images have been direction-corrected, that is, there is no horizontal image in the image.

[0074] In another implementation, Figure 3 As shown, based on the positions of the target parts in the target image, determining the orientation of the breeding object in the target image includes the following steps 31 to 33:

[0075] Step 31, determining the relative position of the first part in the target image based on the center point of the bounding box where the breeding object is located and the center point of the bounding box where the first part is located;

[0076] Step 32, determining the relative position of the second part in the target image based on the center point of the bounding box where the breeding object is located and the center point of the bounding box where the second part is located;

[0077] Step 33, determining the orientation of the farmed object in the target image based on the relative position of the first part in the target image and the relative position of the second part in the target image.

[0078] Step 13: Determine a classification result of the target image based on the relative position relationship between the first part and the second part.

[0079] When the farmed object is in a normal living state, the relative position relationship between the first part and the second part follows a certain growth law. For example, under normal breeding conditions, the distance between the ear and other body parts, such as the tail, of a 5-month-old pig is within a certain range. If the relative position relationship between the first part and the second part in the target image does not meet the conventional growth law, that is, the position relationship between the two is not within the certain range, it means that the farmed object in the target image has been processed. Therefore, the target image can be classified based on the relative position relationship between the first part and the second part.

[0080] In a specific implementation, the first part detected by the target detection model is the part where the farmed object is processed, such as the ears and tail of a pig; processing the farmed object indicates that the farmed object has the necessary conditions for applying for compensation, and the processed farmed object meets the claim requirements.

[0081] Treatment of the breeding objects, such as ear cutting, tail cutting, etc. In this embodiment, the necessary conditions for applying for compensation can be met, which can be the death of the breeding object.

[0082] When the first part is a part of the farmed object that has been processed, the relative position relationship between the first part and the second part after processing has changed significantly compared to the relative position relationship between the first part and the second part under the normal survival state of the farmed object. Therefore, it is easier to achieve classification of the target image.

[0083] In order to more efficiently determine the classification result of the target image, in one implementation, the target part also includes a third part of the body of the breeding object. Then, based on the relative positional relationship between the first part and the second part, determining the classification result of the target image includes:

[0084] Based on the detected existence state of the third part and the relative position relationship between the first part and the second part, a classification result of the target image is determined.

[0085] like Figure 4 As shown, based on the detected existence state of the third part and the relative position relationship between the first part and the second part, the classification result of the target image is determined, and the specific implementation includes the following steps 131 to 133:

[0086] Step 131, determining a first relative position between the first portion and the second portion, and a second relative position between the first portion and the second portion.

[0087] The first relative position is the difference between the first portion and the second portion in one coordinate direction, and the second relative position is the difference between the first portion and the second portion in another coordinate direction.

[0088] Specifically, taking the case where the animal is facing left and its head is upward, the first relative position may be the difference between the ordinate of the center point of the bounding box where the first part is located and the right boundary of the bounding box where the second part is located. diff The second relative position may be the difference Y between the horizontal coordinate of the center point of the bounding box where the first part is located and the upper boundary of the bounding box where the second part is located diff .

[0089] Step 132, if the third part is not detected in the target image, determine the classification result of the target image based on the first relative position, the second relative position, the size of the bounding box of the first part in the other coordinate direction and the size of the bounding box of the second part in the one coordinate direction.

[0090] Specifically, take the third part as the head, the first part as the ear, and the second part as the body of the breeding object as an example:

[0091] The specific implementation of step 132 includes the following steps 1321 to 1323:

[0092] Step 1321: Determine a first ratio based on the size of the bounding box of the first relative position and the second part in one of the coordinate directions.

[0093] Specifically: Based on X diff and the width of the cultured object to determine a first ratio.

[0094] Step 1322: Determine a second ratio based on the second relative position and the size of the bounding box of the first part in the other coordinate direction.

[0095] Specifically: Based on Y diff and the height of the ears, determine the second ratio.

[0096] Step 1323: if the first ratio is less than a first threshold, or the second ratio is less than a second threshold, the target image is an unprocessed image, otherwise it is a processed image.

[0097] The first threshold is set to 0.05, and the second threshold is set to 0.1.

[0098] Step 133, if a third part is detected in the target image, a classification result of the target image is determined based on the first relative position, the size of the bounding box of the first part in the one coordinate direction, and the size of the bounding box of the second part in the one coordinate direction.

[0099] Specifically, take the third part as the head, the first part as the ear, and the second part as the body of the breeding object as an example:

[0100] The specific implementation of step 133 includes the following steps 1331 to 1333:

[0101] Step 1331 : determining a third ratio based on the first relative position and the size of the bounding box of the first part in the one coordinate direction.

[0102] Specifically: Based on X diff and the width of the ear, determine the third ratio.

[0103] Step 1332: Determine a fourth ratio based on the first relative position and the size of the bounding box of the second part in the one coordinate direction.

[0104] Specifically: Based on X diff and the width of the cultured object to determine the fourth ratio.

[0105] Step 1333: If the third ratio is less than the third threshold value, and the fourth ratio is less than the fourth threshold value, the target image is an unprocessed image, otherwise it is a processed image.

[0106] The third threshold is set to 0.1, and the fourth threshold is set to 0.05.

[0107] The first threshold, the second threshold, the third threshold and the fourth threshold are obtained based on the statistical value of the relative position relationship of each target part in the image for which the classification result has been determined.

[0108] According to the method of the embodiment of the present application, each target part of the farmed object in the image and the position of each target part in the target image are detected by the target detection model; based on the relative position relationship of the two parts, the classification result of the target image is determined. Because in the normal living state of the farmed object, the relative position of the two target parts in the body follows a certain rule; in the claim stage, in order to show that the farmed object meets the claim conditions, one of its target parts will be processed, thereby destroying the relative position relationship between the target objects. Therefore, based on the relative position relationship between the two target objects, the classification result of the target image can be determined. The method of the embodiment of the present application is implemented by an image classification device. Compared with the manual classification method, it shortens the classification time and improves the processing efficiency of claims cases. In particular, when processing the classification of a large number of images, it can significantly improve the processing efficiency.

[0109] In order to solve the problem that manual image classification is time-consuming and leads to low claims processing efficiency, based on the same inventive concept as the above-mentioned embodiment of the present application, an image classification device is provided in the embodiment of the present application.

[0110] The specific structural diagram of the device is shown in Figure 5 As shown, it includes the following functional units 51-53:

[0111] The image acquisition unit 51 is used to acquire the image of the breeding object for which the claim is applied.

[0112] The target part detection unit 52 detects the target parts of the farmed object in the image and the positions of the target parts in the target image based on a pre-trained target detection model; the target parts include: a first part and a second part of the body of the farmed object.

[0113] Among them, the first part is the part where the breeding object is processed; processing of the breeding object indicates that the breeding object has the necessary conditions for applying for compensation, and the processed breeding object meets the claim requirements.

[0114] When the first part is a part where the farmed object is processed, the first part and the position of the first part in the target image are identified based on the target detection model.

[0115] The target part detection unit 52 is specifically used for: if the target detection model fails to identify the breeding object, then all the identified candidate frames containing the first part are used as the bounding boxes for locating the first part; if the target detection model identifies the breeding object, then all the candidate frames within the range of the bounding box where the breeding object is located are selected from the identified candidate frames containing the first part as the bounding boxes for locating the first part.

[0116] For other parts except the first part, the target part detection unit 52 is specifically used for:

[0117] Based on a pre-trained target detection model, at least one candidate frame of a single target part of the breeding object in the target image is obtained; based on a preset selection rule, a bounding box for locating the current target part is selected from the at least one candidate frame.

[0118] The preset selection rules include:

[0119] Rule 1: If the target detection model fails to identify the breeding object, the candidate box with the largest area is selected from the identified candidate boxes containing the current target part as the bounding box for locating the position of the current target part.

[0120] Rule 2: If the target detection model recognizes a breeding object, then from the identified candidate frames containing the current target part, the candidate frame with the largest overlapping area with the bounding box where the breeding object is located is selected, and the ratio of the overlapping area to the area of ​​the candidate frame containing the current target part is greater than the fifth threshold, as the bounding box for locating the current target part.

[0121] The image classification unit 53 determines a classification result of the target image based on the relative position relationship between the first part and the second part.

[0122] In one implementation, the target part also includes a third part, and the image classification unit 53 is specifically used to determine the classification result of the target image based on the detected existence state of the third part and the relative position relationship between the first part and the second part.

[0123] In this implementation, the image classification unit 53 further includes subunits 531 - 533 .

[0124] The relative position determination subunit 531 is used to determine a first relative position between a first part and a second part, and a second relative position between the first part and the second part; the first relative position is the difference between the first part and the second part in one coordinate direction, and the second relative position is the difference between the first part and the second part in another coordinate direction.

[0125] The first classification result determination subunit 532 determines the classification result of the target image based on the first relative position, the second relative position, the size of the bounding box of the first part in the other coordinate direction and the size of the bounding box of the second part in the one coordinate direction if the third part is not detected in the target image.

[0126] Specifically used for: determining a first ratio based on the first relative position and the size of the bounding box of the second part in one coordinate direction; determining a second ratio based on the second relative position and the size of the bounding box of the first part in the other coordinate direction; if the first ratio is less than a first threshold, or the second ratio is less than a second threshold, the target image is an unprocessed image, otherwise it is a processed image.

[0127] The second classification result determination subunit 533 determines the classification result of the target image based on the first relative position, the size of the bounding box of the first part in one of the coordinate directions, and the size of the bounding box of the second part in one of the coordinate directions if the third part is detected in the target image.

[0128] Specifically used for: determining a third ratio based on the size of the bounding box of the first relative position and the first part in one of the coordinate directions; determining a fourth ratio based on the size of the bounding box of the first relative position and the second part in one of the coordinate directions; if the third ratio is less than a third threshold value, and the fourth ratio is less than a fourth threshold value, the target image is an unprocessed image, otherwise it is a processed image.

[0129] According to the image classification device of the embodiment of the present application, each target part of the farmed object in the image and the position of each target part in the target image are detected through the target detection model; based on the relative position relationship of the two parts, the classification result of the target image is determined. Because under the normal living state of the farmed object, the relative position of the two target parts in the body follows a certain rule; in the claim stage, in order to show that the farmed object meets the claim conditions, one of its target parts will be processed, thereby destroying the relative position relationship between the target objects. Therefore, based on the relative position relationship between the two target objects, the classification result of the target image can be determined. The method of the embodiment of the present application is implemented by an image classification device. Compared with the manual classification method, it shortens the classification time and improves the processing efficiency of claims cases. In particular, when processing the classification of a large number of images, it can significantly improve the processing efficiency.

[0130] Based on the same inventive concept as the aforementioned embodiments of the present application, the embodiments of the present application also provide a computing device.

[0131] like Figure 6 As shown, the computing device includes: a memory 61 and a processor 62. The memory 61 can be configured to store various other data to support operations on the electronic device. Examples of these data include instructions for any application or method for operating on the electronic device. The memory 61 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk or an optical disk.

[0132] The processor 62 is coupled to the memory 61 and is used to execute the computer program stored in the memory 61 to execute an image classification method described in the above embodiment.

[0133] When the processor 62 executes the computer program to execute the image classification method, the target parts of the farmed object in the image and the positions of the target parts in the target image are detected by the target detection model; based on the relative position relationship of the two parts, the classification result of the target image is determined. Since the relative positions of the two target parts in the body of the farmed object in the normal living state follow certain rules; in the claim stage, in order to show that the farmed object meets the claim conditions, one of its target parts will be processed, thereby destroying the relative position relationship between the target objects. Therefore, based on the relative position relationship between the two target objects, the classification result of the target image can be determined. The method of the embodiment of the present application is implemented by an image classification device. Compared with the manual classification method, it shortens the classification time and improves the processing efficiency of claim cases. In particular, when processing the classification of a large number of images, it can significantly improve the processing efficiency.

[0134] When the processor 62 executes the computer program in the memory 61, in addition to the above functions, it can also realize other functions, and the details can be found in the description of the previous embodiments.

[0135] Further, such as Figure 6 As shown, the computing device also includes: a display 64, a communication component 63, a power component 65, an audio component 66 and other components. Figure 6 Only some components are shown schematically, and it does not mean that the computing device only includes Figure 6 Components shown.

[0136] Accordingly, an embodiment of the present application further provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a computer, the methods provided in the above embodiments can be implemented.

[0137] An embodiment of the present application also provides a computer program product, including a computer program, which implements the methods provided in the above embodiments when executed by a processor.

[0138] The computer program product includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions. When the program instructions are executed by a processor, the methods provided in the above embodiments are implemented.

[0139] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0140] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of 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 is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0141] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An image classification method, characterized in that: include: Obtain images of the breeding objects for which claims are being applied; Based on a pre-trained target detection model, detecting each target part of the breeding object in the image and the position of each target part in the target image; The target parts include: a first part and a second part of the body of the cultured object; Based on the relative position relationship between the first part and the second part, a classification result of the target image is determined.

2. The method according to claim 1, characterized in that The first part is the part where the cultured object is processed; The processing of the breeding objects indicates that the breeding objects meet the necessary conditions for applying for compensation, and the processed breeding objects meet the compensation requirements.

3. The method according to claim 2, characterized in that Based on the pre-trained target detection model, detecting each target part of the breeding object in the image and the position of each target part in the target image includes: If the target detection model fails to identify the breeding object, each identified candidate frame containing the first part is used as a bounding box for locating the position of the first part; If the target detection model recognizes the farming object, all candidate frames that are within the range of the bounding box where the farming object is located are selected from the identified candidate frames containing the first part as the bounding box for locating the position of the first part; The bounding box is identified by the target detection model and is used to locate the positions of target parts of the breeding object in the target image.

4. The method according to claim 1, characterized in that The target part also includes a third part of the body of the cultured subject, then, Determining a classification result of the target image based on a relative positional relationship between the first part and the second part includes: Based on the detected existence state of the third part and the relative position relationship between the first part and the second part, a classification result of the target image is determined.

5. The method according to claim 4, characterized in that Determining a classification result of the target image based on the detected existence state of the third part and the relative position relationship between the first part and the second part includes: Determine a first relative position between the first part and the second part, and a second relative position between the first part and the second part; the first relative position is the difference between the first part and the second part in one coordinate direction, and the second relative position is the difference between the first part and the second part in another coordinate direction; If the third part is not detected in the target image, determining a classification result of the target image based on the first relative position, the second relative position, the size of the bounding box of the first part in the other coordinate direction, and the size of the bounding box of the second part in the one coordinate direction; If a third part is detected in the target image, a classification result of the target image is determined based on the first relative position, the size of the bounding box of the first part in the one coordinate direction, and the size of the bounding box of the second part in the one coordinate direction.

6. The method according to claim 5, characterized in that Determining a classification result of the target image based on the first relative position, the second relative position, the size of the bounding box of the first part in the other coordinate direction, and the size of the bounding box of the second part in the one coordinate direction, includes: determining a first ratio based on the first relative position and the size of the bounding box of the second part in the one coordinate direction; determining a second ratio based on the second relative position and the size of the bounding box of the first part in the other coordinate direction; If the first ratio is smaller than a first threshold, or the second ratio is smaller than a second threshold, the target image is an unprocessed image, otherwise it is a processed image.

7. The method according to claim 5, characterized in that Determining a classification result of the target image based on the first relative position, the size of the bounding box of the first part in the one coordinate direction, and the size of the bounding box of the second part in the one coordinate direction, includes: determining a third ratio based on the first relative position and a size of the bounding box of the first portion in the one coordinate direction; determining a fourth ratio based on the first relative position and the size of the bounding box of the second part in the one coordinate direction; If the third ratio is smaller than the third threshold, and the fourth ratio is smaller than the fourth threshold, the target image is an unprocessed image, otherwise it is a processed image.

8. The method according to claim 1, characterized in that The method further comprises: Based on the positions of the target parts in the target image, the orientation of the farmed object in the target image is determined.

9. The method according to claim 8, characterized in that Based on the positions of the target parts in the target image, determining the orientation of the farmed object in the target image includes: Determine the relative position of the first part in the target image based on the center point of the bounding box where the breeding object is located and the center point of the bounding box where the first part is located; Determine the relative position of the second part in the target image based on the center point of the bounding box where the cultured object is located and the center point of the bounding box where the second part is located; Based on the relative position of the first part in the target image and the relative position of the second part in the target image, the orientation of the farmed object in the target image is determined.

10. The method according to claim 1, characterized in that Based on a pre-trained target detection model, detecting each target part of the breeding object in the image and the position of each target part in the target image, including: Based on a pre-trained target detection model, obtaining at least one candidate frame containing a single target part of the breeding object in the target image; Based on a preset selection rule, a bounding box is selected from the at least one candidate box as a bounding box for locating the position of the current target part.

11. The method according to claim 10, characterized in that The preset selection rules include: If the target detection model fails to identify the breeding object, a candidate frame with the largest area is selected from the identified candidate frames containing the current target part as a bounding box for locating the position of the current target part; If the target detection model identifies a breeding object, then from the identified candidate frames containing the current target part, the candidate frame with the largest overlapping area with the bounding box where the breeding object is located is selected, and the ratio of the overlapping area to the area of ​​the candidate frame containing the current target part is greater than the fifth threshold, as the bounding box for locating the current target part.

12. An image classification device, characterized in that: include: An image acquisition unit, used to acquire an image of the breeding object for which compensation is applied; A target part detection unit, based on a pre-trained target detection model, detects each target part of the breeding object in the image and the position of each target part in the target image; The target parts include: a first part and a second part of the body of the cultured object; The image classification unit determines a classification result of the target image based on the relative position relationship between the first part and the second part.

13. A computing device, characterized in that: include: Memory and processor, wherein: The memory is used to store computer programs; The processor is coupled to the memory and is used to execute the computer program stored in the memory to execute the method according to any one of claims 1 to 11.

14. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a computer, the method described in any one of claims 1 to 11 can be implemented.

15. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to any one of claims 1 to 11 when being executed by a processor.

Images