A barbell weight recognition method, device, equipment and storage medium

Abstract

The application discloses a barbell weight recognition method, device and equipment and a storage medium, and is applied to the field of artificial intelligence. An image collected by an image collection device is input into a barbell detection model to perform barbell detection, and a target image containing a barbell is determined. The target image is input into a barbell plate image segmentation model to obtain an image segmentation result containing each barbell plate. Color features of each image segmentation result are determined, the color features are matched with standard color features in a standard color feature set, and a matching result is determined. The weight of the barbell plate in the image segmentation result is determined based on the matching result, and the weight of the barbell in the target image is determined based on the weight of the barbell plate. The application performs target recognition and image segmentation on the image containing the barbell, performs feature matching based on the color features of the segmentation result, determines the weight of the barbell plate based on the feature matching result, and further determines the total weight of the barbell in the image, thereby improving the accuracy and recognition efficiency of the barbell weight recognition.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence, and in particular to a barbell weight recognition method, a barbell weight recognition device, an electronic device, and a computer-readable storage medium. Background Technology

[0002] A barbell is a type of exercise training equipment. A standard barbell generally consists of three parts: the barbell bar, weight plates, and clamps. The intensity of strength training is adjusted by controlling the number of weight plates. Therefore, it is necessary to accurately record the weight of the weight plates used during training. Currently, the main methods for recording the weight of weight plates are manual recording or electronic tags attached to the weight plates. However, manual recording may be inaccurate, and electronic tags are prone to damage or confusion, leading to inaccurate weight statistics, affecting training results, and potentially causing physical injuries to the trainee. Summary of the Invention

[0003] The purpose of this invention is to provide a barbell weight recognition method, a barbell weight recognition device, an electronic device, and a computer-readable storage medium, which are applied in the field of artificial intelligence. This method improves the accuracy and efficiency of barbell weight recognition by performing target recognition and image segmentation on an image containing a barbell, performing feature matching based on the color features of the segmentation results, determining the weight of the barbell plates based on the feature matching results, and then determining the total weight of the barbell in the image.

[0004] To solve the above-mentioned technical problems, the present invention provides a barbell weight recognition method, comprising:

[0005] An image acquisition device is set up in the training area, and the image acquired by the image acquisition device is input into the barbell detection model to perform barbell detection and determine the target image containing the barbell.

[0006] The target image is input into the barbell plate image segmentation model to obtain the image segmentation result containing each barbell plate;

[0007] Construct a standard color feature set for the barbell and determine the color features of each image segmentation result;

[0008] The color feature is matched with the standard color features in the standard color feature set to determine the matching result;

[0009] The weight of the barbell plate in the image segmentation result is determined based on the matching result, and the weight of the barbell in the target image is determined based on the weight of the barbell plate.

[0010] Optionally, a standard color feature set for the barbell is constructed to determine the color features of each image segmentation result, including:

[0011] Determine the target color space, and convert the image segmentation result from the current color space to the target color space to obtain the target image segmentation result;

[0012] Construct a set of target standard color features for the barbell in the target color space, and determine the target color features of each target image segmentation result;

[0013] Accordingly, the color feature is matched with the standard color features in the standard color feature set to determine the matching result, including:

[0014] The target color feature is matched with the standard color features in the target standard color feature set to determine the matching result.

[0015] Optionally, determining the barbell weight in the target image based on the weight of the barbell plates includes:

[0016] The difference in weight between the barbell plates on the left and right sides of the barbell bar in the target image is determined based on the Euclidean norm.

[0017] When the difference is less than the symmetry threshold and the modulo operation result of the total weight of the barbell plates and the standard maximum weight is 0, the weight of the barbell in the target image is determined based on the weight of the barbell plates.

[0018] Optionally, the target image is input into a barbell plate image segmentation model to obtain image segmentation results containing each barbell plate, including:

[0019] The target image is input into the barbell plate image segmentation model to obtain an initial image segmentation result containing each of the barbell plates;

[0020] The area of ​​the barbell plate in the initial image segmentation result is determined, and the area of ​​the barbell plate is input into the validity detection model to determine the validity of the initial image segmentation result;

[0021] If the initial image segmentation result is a valid segmentation result, then the initial image segmentation result is determined as the image segmentation result.

[0022] Optionally, the color feature is matched with the standard color features in the standard color feature set to determine the matching result, including:

[0023] Determine the Gaussian distribution results of the color feature and each of the standard color features in the standard color feature set;

[0024] Based on the Gaussian distribution result, the standard color feature that matches the color feature is determined, and the matching standard color feature is determined as the matching result.

[0025] Optionally, the method further includes:

[0026] The weight of the barbell in each target image within a preset time period is obtained, and the weight of the barbell in each target image is smoothed based on a time-series filter to obtain the target barbell weight.

[0027] Optionally, the target image is input into a barbell plate image segmentation model to obtain image segmentation results containing each barbell plate, including:

[0028] Acquire sample images of various types of barbells at different distances and angles from the image acquisition device;

[0029] The initial barbell plate image segmentation model is trained based on the sample images to obtain the trained barbell plate image segmentation model.

[0030] The target image is input into the barbell plate image segmentation model to obtain the image segmentation result containing each of the barbell plates.

[0031] To solve the above-mentioned technical problems, the present invention provides a barbell weight recognition device, comprising:

[0032] The first module is used to set up an image acquisition device in the training area, input the image acquired by the image acquisition device into the barbell detection model for barbell detection, and determine the target image containing the barbell.

[0033] The second module is used to input the target image into the barbell plate image segmentation model to obtain the image segmentation result containing each barbell plate;

[0034] The third module is used to construct a standard color feature set for the barbell and determine the color features of each image segmentation result.

[0035] The fourth module is used to match the color features with the standard color features in the standard color feature set to determine the matching result;

[0036] The fifth module is used to determine the weight of the barbell plates in the image segmentation result based on the matching result, and to determine the weight of the barbell in the target image based on the weight of the barbell plates.

[0037] To solve the above-mentioned technical problems, the present invention provides an electronic device, comprising:

[0038] Memory, used to store computer programs;

[0039] A processor is used to implement the barbell weight recognition method described above when executing the computer program.

[0040] To address the aforementioned technical problems, the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the aforementioned barbell weight recognition method.

[0041] As can be seen, this invention sets up an image acquisition device in the training area, inputs the image acquired by the image acquisition device into a barbell detection model to perform barbell detection, and determines the target image containing the barbell; inputs the target image into a barbell plate image segmentation model to obtain image segmentation results containing each barbell plate; constructs a standard color feature set for the barbell, and determines the color features of each image segmentation result; matches the color features with the standard color features in the standard color feature set to determine the matching result; determines the weight of the barbell plates in the image segmentation result based on the matching result, and determines the weight of the barbell in the target image based on the weight of the barbell plates.

[0042] This invention improves the accuracy and efficiency of barbell weight recognition by performing target recognition and image segmentation on images containing barbells, performing feature matching based on the color features of the segmentation results, determining the weight of the barbell plates based on the feature matching results, and thus determining the total weight of the barbell in the image. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0044] Figure 1 A flowchart of a barbell weight recognition method provided in an embodiment of the present invention;

[0045] Figure 2 This is a structural block diagram of a barbell weight recognition device provided in an embodiment of the present invention. Detailed Implementation

[0046] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] When performing free weight training with a barbell, it is necessary to accurately record the weight used. Currently, this is mainly done manually or by attaching electronic tags to the weight plates. However, this method suffers from problems such as inaccurate recording and easy damage to the tags. While there is a weight recognition technology based on computer vision, the recognition accuracy is not high due to factors such as similar colors of the weight plates, changes in lighting, and the atmosphere of the scene.

[0048] This invention employs a multi-stage detection and verification mechanism, obtaining the accurate barbell weight within the training period through time-series processing. An adaptive area threshold calculation method is used to determine the validity of the barbell recognition region based on its area. A symmetry-based weight verification method compares the weight balance on both sides to determine weight validity. In the mask optimization scheme for image segmentation, precise recognition regions are obtained by optimizing masking and edge optimization algorithms.

[0049] The following combination Figure 1 , Figure 1 A flowchart of a barbell weight recognition method provided in an embodiment of the present invention, the method may include:

[0050] S101: Set up an image acquisition device in the training area, input the image acquired by the image acquisition device into the barbell detection model to perform barbell detection, and determine the target image containing the barbell.

[0051] This embodiment allows the installation of an image acquisition device in the training area to capture images of that area. This embodiment does not limit the type of image acquisition device or the format of the acquired images; these can be configured based on the specific application. Furthermore, this embodiment does not limit the operating time or image acquisition frequency of the image acquisition device; these can also be configured based on the actual application.

[0052] Since there may be scenarios where no trainees are training in the training area during the operation of the image acquisition device, or where trainees are training using non-barbell equipment, this embodiment can first input the images acquired by the image acquisition device into the barbell detection model for barbell detection to identify target images containing barbells.

[0053] This embodiment does not limit the specific type of barbell detection model, nor does it limit the training method of the barbell detection model. Generally, the barbell detection model can be a CNN (Convolutional Neural Networks) model, which extracts the features of the target in the image through operations such as convolution and pooling to identify the barbell.

[0054] Considering the issues of lighting and angle, this embodiment can acquire training images of trainees under different lighting conditions, different angles, and different distances from the image acquisition device as a sample set for training the barbell detection model.

[0055] In this embodiment, the image acquisition device can be positioned directly in front of the trainee to capture a frontal image of the trainee using a barbell during training. In one example, the image acquisition device can be a movable tracking image acquisition device, used to follow the trainee and identify the frontal training image to avoid the problem of the barbell obstructing the training image.

[0056] S102: Input the target image into the barbell plate image segmentation model to obtain the image segmentation result containing each barbell plate.

[0057] In this embodiment, the target image can be input into the barbell plate image segmentation model to obtain the image segmentation result containing each barbell plate.

[0058] This embodiment does not limit the specific type and training method of the image segmentation model. Similar to the barbell detection model, considering lighting and angle issues, this embodiment can acquire sample images of various types of barbells at different distances and angles from the image acquisition device. Based on these sample images, the initial barbell plate image segmentation model is trained to obtain a trained barbell plate image segmentation model. The target image is then input into the barbell plate image segmentation model to obtain image segmentation results containing each barbell plate. Each barbell plate in the training images in the sample set is labeled.

[0059] In this embodiment, the region masking function of the image segmentation model can be defined as:

[0060] M(x,y)=πR i (x,y);

[0061] In the formula, (x,y) are the center coordinates of the detected target, M(x,y) is the region masking function, and R... i (x,y) is the indicator function for the i-th masking layer region.

[0062] The masking region is parameterized as follows: corner_mask = λ1 * min(W, H), edge_mask = [λ2 * W, λ3 * H], where W and H are the width and height of the training image, respectively; λ1 is the first scaling factor; λ2 is the second scaling factor; λ3 is the third scaling factor; corner_mask is the corner mask parameter; and edge_mask is the edge mask parameter. During the training of the image segmentation model, the values ​​of each scaling factor can be updated until convergence.

[0063] By performing a dot product operation between the region masking function and the input image, the edges and corners of the image can be masked, background interference can be reduced, and the barbell plates can be segmented. In this embodiment, the number of image segmentation results corresponds to the number of barbell plates in the image.

[0064] In this embodiment, the validity of the image segmentation result can be verified. Specifically, the target image is input into the barbell plate image segmentation model to obtain an initial image segmentation result containing each barbell plate; the area of ​​the barbell plates in the initial image segmentation result is determined, and the area of ​​the barbell plates is input into the validity detection model to determine the validity of the initial image segmentation result; if the initial image segmentation result is a valid segmentation result, then the initial image segmentation result is determined as the image segmentation result.

[0065] The expression for the validity detection model in this embodiment can be:

[0066] A = a * S * β(r);

[0067] In the formula, A is the area of ​​the barbell plate, S is the total area of ​​the image, β(r) is the resolution adaptive function, and a is the scaling factor.

[0068] S103: Construct a standard color feature set for the barbell and determine the color features of each image segmentation result.

[0069] S104: Match the color features with the standard color features in the standard color feature set to determine the matching result.

[0070] This embodiment can construct a standard color feature set for a barbell and determine the color features of each image segmentation result. The color features are then matched with the standard color features in the standard color feature set to determine the matching result.

[0071] In this embodiment, each color of barbell plate corresponds to a specific weight, and different colors of barbell plates correspond to different weights. Therefore, a standard color feature set for the barbell can be pre-constructed based on its color, containing standard color features for each barbell color. This embodiment does not limit the method of color feature extraction; extraction can be performed based on the actual application.

[0072] This embodiment can determine the color features of each image segmentation result by matching the color features with standard color features in a standard color feature set. That is, it finds the color feature most similar to the color features of the image segmentation result from the standard color features. Based on the correspondence between barbell weight and color, the weight of the barbell plates in each image segmentation result can be determined through the matching result.

[0073] Since the effects of image color analysis differ under different color spaces, this embodiment can determine the target color space, convert the image segmentation result from the current color space to the target color space, and obtain the target image segmentation result; construct the target standard color feature set of the barbell in the target color space, and determine the target color features of each target image segmentation result.

[0074] This embodiment does not limit the specific type of the target color space. When the format of the image acquired by the image acquisition device (i.e., the current color space) is RGB (Red Green Blue), the target color space can be HSV (Hue Saturation Value) format.

[0075] Therefore, this embodiment can convert the RGB format image segmentation result from the current color space to the target color space to obtain the HSV format target image segmentation result. Specifically, a non-linear mapping from RGB to HSV can be established, and the format conversion can be performed through the non-linear mapping:

[0076] H=g1(R,G,B);

[0077] S=g2(R,G,B);

[0078] V=g3(R,G,B);

[0079] Where R is the pixel value of the red channel of the image, G is the pixel value of the green channel of the image, B is the pixel value of the blue channel of the image, H is the hue of the image, S is the saturation of the image, V is the brightness of the image, g1 is the first mapping function, g2 is the second mapping function, and g3 is the third mapping function.

[0080] This embodiment can construct feature vector representations for each barbell color:

[0081] c i =[H i ,S i V i ] T ;

[0082] In the formula, c i H represents the standard color characteristic of the i-th barbell color. i Let S be the hue of the i-th barbell color. i V represents the saturation of the i-th barbell color. i Let be the lightness of the i-th barbell color, and T be the transpose of the matrix. Based on the standard color features of each barbell color, a target standard color feature set C={ In this embodiment, a standard color feature set can be constructed in the same manner.

[0083] Accordingly, the color features are matched with the standard color features in the standard color feature set to determine the matching result. Specifically, the target color features are matched with the standard color features in the target standard color feature set to determine the matching result.

[0084] This embodiment does not limit the specific matching method. Generally, Gaussian distribution can be used for matching to determine the Gaussian distribution results of the target color feature and each standard color feature in the target standard color feature set. Based on the Gaussian distribution results, the standard color features in the target standard color feature set that match the target color feature are determined, and the matched standard color features are determined as the matching results. Correspondingly, the standard color features that match the color feature in the standard color feature set can be determined in the same way.

[0085] Specifically, the standard color feature c of the image segmentation result can be determined, along with the various standard color features c. i Gaussian distribution between p(c|c) i )=N(μ i ,Σ i ), where N(μ) i ,Σ i ) represents the normal distribution of color features, μ i Let Σ be the corresponding mean vector. i Let be the corresponding covariance matrix.

[0086] The similarity between the target color feature and the standard color feature in the target standard color feature set can be determined by the Gaussian distribution results. Therefore, this embodiment can determine the standard color feature that matches the target color feature based on the Gaussian distribution results.

[0087] S105: Determine the weight of the barbell plates in the image segmentation result based on the matching result, and determine the weight of the barbell in the target image based on the weight of the barbell plates.

[0088] This embodiment can determine the weight of the barbell plates in the image segmentation result based on the matching result, and determine the weight of the barbell in the target image based on the weight of the barbell plates.

[0089] Specifically, each image segmentation result can correspond to a matching result. Based on the correspondence between barbell weight and color, the weight of the barbell plates in each image segmentation result can be determined through the matching result. The total weight of the barbell plates can be obtained by adding up all the weights of the barbell plates in the image segmentation results obtained from the same target image.

[0090] Since the barbell includes standard components of uniform weight, such as the barbell bar, in addition to the weight plates, this embodiment can add the total weight of the weight plates to the weight of the standard components to obtain the weight of the barbell in the target image.

[0091] To ensure the effectiveness of weight recognition throughout the entire movement process, temporal smoothing is required, and the weight value that appears most frequently is used as the final estimated weight. Specifically, the barbell weight of each target image within a preset time period is obtained, and the barbell weight of each target image is smoothed based on a temporal filter to obtain the target barbell weight.

[0092] In this embodiment, the preset time period can be the time for a training cycle. This embodiment does not limit the parameters of the timing filter and can be set based on the actual application.

[0093] Since the weight of the weight plates on both sides of the barbell needs to be consistent in a normal training scenario, this embodiment can verify the symmetry of the weight of the weight plates on both sides of the barbell to ensure the accuracy of the identified barbell weight.

[0094] Specifically, the difference in weight between the barbell plates on the left and right sides of the barbell in the target image can be determined first based on the Euclidean norm. When the difference is less than the symmetry threshold and the modulo operation result of the total weight of the barbell plates and the standard maximum weight is 0, the weight of the barbell in the target image can be determined based on the weight of the barbell plates.

[0095] In this embodiment, the difference D between the weights of the barbell plates on the left and right sides of the barbell bar is determined, and the difference D needs to be less than the symmetry threshold ε.

[0096] Furthermore, the total weight of the barbell plates must satisfy: mod(Weight,k)=0, where mod is the modulo operation and k is the standard maximum weight.

[0097] When both of the above judgment conditions are met at the same time, the recognition result of the target image can be considered valid, and the recognized barbell weight is a valid value.

[0098] Based on the above embodiments, the present invention improves the accuracy and efficiency of barbell weight recognition by performing target recognition and image segmentation on an image containing a barbell, performing feature matching based on the color features of the segmentation results, determining the weight of the barbell plates based on the feature matching results, and thus determining the total weight of the barbell in the image.

[0099] The following combination Figure 2 , Figure 2 This is a structural block diagram of a barbell weight recognition device provided in an embodiment of the present invention. The device may include:

[0100] The first module 100 is used to set up an image acquisition device in the training area, input the image acquired by the image acquisition device into the barbell detection model to perform barbell detection, and determine the target image containing the barbell.

[0101] The second module 200 is used to input the target image into the barbell plate image segmentation model to obtain the image segmentation result containing each barbell plate;

[0102] The third module 300 is used to construct a standard color feature set of the barbell and determine the color features of each image segmentation result.

[0103] The fourth module 400 is used to match the color feature with the standard color feature in the standard color feature set to determine the matching result;

[0104] The fifth module 500 is used to determine the weight of the barbell plate in the image segmentation result based on the matching result, and to determine the weight of the barbell in the target image based on the weight of the barbell plate.

[0105] Based on the above embodiments, the present invention improves the accuracy and efficiency of barbell weight recognition by performing target recognition and image segmentation on an image containing a barbell, performing feature matching based on the color features of the segmentation results, determining the weight of the barbell plates based on the feature matching results, and thus determining the total weight of the barbell in the image.

[0106] Based on the above embodiments, the third module 300 may include:

[0107] The first unit is used to determine the target color space and convert the image segmentation result from the current color space to the target color space to obtain the target image segmentation result;

[0108] The second unit is used to construct a set of target standard color features of the barbell in the target color space and to determine the target color features of each target image segmentation result.

[0109] Correspondingly, the fourth module 400 includes:

[0110] The third unit is used to match the target color feature with the standard color feature in the target standard color feature set to determine the matching result.

[0111] Based on the above embodiments, the fifth module 500 may include:

[0112] The fourth unit is used to determine the difference in weight of the barbell plates on the left and right sides of the barbell bar in the target image based on the Euclidean norm.

[0113] The fifth unit is used to determine the weight of the barbell in the target image based on the weight of the barbell plates when the difference is less than the symmetry threshold and the modulo operation result of the total weight of the barbell plates and the standard maximum weight is 0.

[0114] Based on the above embodiments, the second module 200 may include:

[0115] The sixth unit is used to input the target image into the barbell plate image segmentation model to obtain an initial image segmentation result containing each of the barbell plates;

[0116] The seventh unit is used to determine the area of ​​the barbell plate in the initial image segmentation result, and input the area of ​​the barbell plate into the validity detection model to determine the validity of the initial image segmentation result;

[0117] The eighth unit is used to determine the initial image segmentation result as the image segmentation result if the initial image segmentation result is a valid segmentation result.

[0118] Based on the above embodiments, the fourth module 400 may include:

[0119] The ninth unit is used to determine the Gaussian distribution results of the color feature and each of the standard color features in the standard color feature set;

[0120] The tenth unit is used to determine the standard color feature that matches the color feature based on the Gaussian distribution result, and to determine the matched standard color feature as the matching result.

[0121] Based on the above embodiments, the device may further include:

[0122] The sixth module is used to obtain the barbell weight of each target image within a preset time period, and to smooth the barbell weight of each target image based on a time-series filter to obtain the target barbell weight.

[0123] Based on the above embodiments, the second module 200 may include:

[0124] The eleventh unit is used to acquire sample images of various types of barbells at different distances and angles from the image acquisition device;

[0125] The twelfth unit is used to train the initial barbell plate image segmentation model based on the sample images to obtain the trained barbell plate image segmentation model.

[0126] The thirteenth unit is used to input the target image into the barbell plate image segmentation model to obtain the image segmentation result containing each of the barbell plates.

[0127] Based on the above embodiments, the present invention also provides an electronic device, which may include a memory and a processor. The memory stores a computer program, and when the processor calls the computer program in the memory, it can implement the steps provided in the above embodiments. Of course, the device may also include various necessary network interfaces, a power supply, and other components.

[0128] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by an execution terminal or processor, can implement the method provided in the embodiments of the present invention; the storage medium may include various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0129] In this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, without necessarily requiring or implying any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

Claims

1. A method for identifying barbell weight, characterized in that, include: An image acquisition device is set up in the training area, and the image acquired by the image acquisition device is input into the barbell detection model to perform barbell detection and determine the target image containing the barbell. The target image is input into the barbell plate image segmentation model to obtain the image segmentation result containing each barbell plate; Construct a standard color feature set for the barbell and determine the color features of each image segmentation result; The color feature is matched with the standard color features in the standard color feature set to determine the matching result; The weight of the barbell plate in the image segmentation result is determined based on the matching result, and the weight of the barbell in the target image is determined based on the weight of the barbell plate.

2. The barbell weight recognition method according to claim 1, characterized in that, Construct a standard color feature set for the barbell and determine the color features of each image segmentation result, including: Determine the target color space, and convert the image segmentation result from the current color space to the target color space to obtain the target image segmentation result; Construct a set of target standard color features for the barbell in the target color space, and determine the target color features of each target image segmentation result; Accordingly, the color feature is matched with the standard color features in the standard color feature set to determine the matching result, including: The target color feature is matched with the standard color features in the target standard color feature set to determine the matching result.

3. The barbell weight recognition method according to claim 1, characterized in that, Determining the weight of the barbell in the target image based on the weight of the barbell plates includes: The difference in weight between the barbell plates on the left and right sides of the barbell bar in the target image is determined based on the Euclidean norm. When the difference is less than the symmetry threshold and the modulo operation result of the total weight of the barbell plates and the standard maximum weight is 0, the weight of the barbell in the target image is determined based on the weight of the barbell plates.

4. The barbell weight recognition method according to claim 1, characterized in that, The target image is input into the barbell plate image segmentation model to obtain image segmentation results containing each barbell plate, including: The target image is input into the barbell plate image segmentation model to obtain an initial image segmentation result containing each of the barbell plates; The area of ​​the barbell plate in the initial image segmentation result is determined, and the area of ​​the barbell plate is input into the validity detection model to determine the validity of the initial image segmentation result; If the initial image segmentation result is a valid segmentation result, then the initial image segmentation result is determined as the image segmentation result.

5. The barbell weight recognition method according to claim 1, characterized in that, Matching the color feature with the standard color features in the standard color feature set to determine the matching result includes: Determine the Gaussian distribution results of the color feature and each of the standard color features in the standard color feature set; Based on the Gaussian distribution result, the standard color feature that matches the color feature is determined, and the matching standard color feature is determined as the matching result.

6. The barbell weight recognition method according to claim 1, characterized in that, Also includes: The weight of the barbell in each target image within a preset time period is obtained, and the weight of the barbell in each target image is smoothed based on a time-series filter to obtain the target barbell weight.

7. The barbell weight recognition method according to claim 1, characterized in that, The target image is input into the barbell plate image segmentation model to obtain image segmentation results containing each barbell plate, including: Acquire sample images of various types of barbells at different distances and angles from the image acquisition device; The initial barbell plate image segmentation model is trained based on the sample images to obtain the trained barbell plate image segmentation model. The target image is input into the barbell plate image segmentation model to obtain the image segmentation result containing each of the barbell plates.

8. A barbell weight recognition device, characterized in that, include: The first module is used to set up an image acquisition device in the training area, input the image acquired by the image acquisition device into the barbell detection model to perform barbell detection, and determine the target image containing the barbell. The second module is used to input the target image into the barbell plate image segmentation model to obtain the image segmentation result containing each barbell plate; The third module is used to construct a standard color feature set for the barbell and determine the color features of each image segmentation result. The fourth module is used to match the color features with the standard color features in the standard color feature set to determine the matching result; The fifth module is used to determine the weight of the barbell plates in the image segmentation result based on the matching result, and to determine the weight of the barbell in the target image based on the weight of the barbell plates.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the barbell weight recognition method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, implement the barbell weight recognition method as described in any one of claims 1 to 7.

Images