A gesture label calibration system, method and device

Through multiple image acquisition devices, the gesture display area is captured from different angles, and the display device is used to obtain the tags input by the user, and the gesture recognition model is trained, which solves the problem of low label acquisition efficiency in gesture recognition model training, and realizes efficient gesture calibration and Accurate identification results.

CN113515974BActive Publication Date: 2024-11-01MOMENTA (SUZHOU) TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202010277092.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-04-10
Publication Date
2024-11-01
Estimated Expiration
2040-04-10

AI Technical Summary

Technical Problem

In the prior art, training of gesture recognition models requires a large number of sample images with gesture labels, and how to efficiently obtain these labels is a key issue.

Method used

Through multiple image acquisition devices, the processor determines the gesture label based on the acquired image and preset gesture recognition method, and uses the display device to obtain the tag input by the user, and trains the initial and pre-trained gesture recognition model. Until the convergence condition is reached, the calibration of gesture labels is achieved.

Benefits of technology

It realizes efficient calibration of gestures in the image, reduces the calibration workload, improves the accuracy of the gesture recognition model and the accuracy of the recognition results, and can handle gesture label calibration in occlusion situations.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113515974B_ABST
    Figure CN113515974B_ABST
Patent Text Reader

Abstract

The embodiment of the present invention discloses a system, method and device for calibrating gesture labels. The system includes a processor and multiple image acquisition devices, wherein the processor is connected to the multiple image acquisition devices; the multiple image acquisition devices are configured to shoot a gesture display area from different angles; the processor is configured to obtain images captured by the multiple image acquisition devices for the gesture display area within the same acquisition cycle as images to be recognized; and based on the images to be recognized captured by the multiple image acquisition devices and a preset gesture recognition method, determine the gesture label corresponding to the gesture to be recognized displayed in the gesture display area, and calibrate the gesture label corresponding to the image to be recognized, wherein the image to be recognized includes: the gesture to be recognized displayed in the gesture display area, so as to realize the calibration of the gesture label corresponding to the gesture contained in the image.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of image recognition technology, and in particular to a system, method and device for calibrating gesture labels. Background Art

[0002] With the development of science and technology, various intelligent control devices have emerged to improve the quality of life of users. For example, in the fields of smart home and smart driving, the corresponding functions of the controlled devices are controlled by the user's gestures to improve the user experience.

[0003] In the current technology of controlling the corresponding functions of the controlled device through gestures, it is first necessary to identify the meaning represented by the gesture, and then control the corresponding function of the controlled device based on the meaning represented by the recognized gesture. In the related technology, the meaning represented by the gesture in the image generally needs to be recognized by a neural network model. In order to train a gesture recognition model that can be used to identify the gesture labels of the gestures contained in the image, it is necessary to obtain a large number of sample images calibrated with gesture labels to train the gesture recognition model. Accordingly, how to obtain sample images calibrated with gesture labels is crucial, that is, how to calibrate gesture labels for the gestures contained in the image is crucial. Summary of the invention

[0004] The present invention provides a gesture label calibration system, method and device to achieve the calibration of gesture labels corresponding to gestures contained in an image. The specific technical solution is as follows:

[0005] In a first aspect, an embodiment of the present invention provides a gesture label calibration system, the system comprising a processor and a plurality of image acquisition devices, the processor being connected to the plurality of image acquisition devices;

[0006] The multiple image acquisition devices are configured to photograph the gesture display area from different angles;

[0007] The processor is configured to obtain images captured by the multiple image acquisition devices for the gesture display area within the same acquisition cycle as images to be recognized; and based on the images to be recognized captured by the multiple image acquisition devices and a preset gesture recognition method, determine the gesture label corresponding to the gesture to be recognized displayed in the gesture display area, and calibrate the gesture label corresponding to the image to be recognized, wherein the image to be recognized includes: the gesture to be recognized displayed in the gesture display area.

[0008] Optionally, the system further comprises a display device;

[0009] The processor is specifically configured to: determine the gesture label corresponding to the gesture to be recognized displayed in the gesture display area based on the images to be recognized acquired by the multiple image acquisition devices and the preset gesture recognition method, and mark the gesture label corresponding to the image to be recognized.

[0010] Sending the image to be recognized acquired by each of the plurality of image acquisition devices to the display device;

[0011] The display device is configured to display the image to be recognized captured by each of the multiple image acquisition devices, so that the user can input a gesture label corresponding to the gesture to be recognized contained in the target image to be recognized based on the gesture to be recognized contained in the target image to be recognized among the images to be recognized;

[0012] The processor obtains the gesture label input by the user for the target image to be recognized; determines the gesture label as a gesture label corresponding to the gesture to be recognized contained in each image in the image to be recognized, and calibrates it.

[0013] Optionally, the processor is further configured to: after determining the gesture label as a gesture label corresponding to the gesture contained in each image in the to-be-recognized image and marking the gesture label, obtain an initial first gesture recognition model;

[0014] Obtain multiple groups of sample gesture images containing sample gestures, and calibrated gesture labels corresponding to each group of sample gesture images, wherein each group of sample gesture images includes: images captured by the multiple image acquisition devices for the gesture display area within the same acquisition cycle, and the calibrated gesture labels corresponding to each group of sample gesture images are: gesture labels corresponding to the sample gestures displayed in the gesture display area determined by the processor based on the group of sample gesture images and the preset gesture recognition method;

[0015] The initial first gesture recognition model is trained using the multiple groups of sample gesture images and the calibrated gesture labels corresponding to each group of sample gesture images, as well as the images to be recognized and the gesture labels corresponding to them, until the initial first gesture recognition model reaches a preset first convergence condition, thereby obtaining a target gesture recognition model for detecting the gesture labels corresponding to the gestures contained in the images.

[0016] Optionally, the process in which the processor determines, based on the images to be recognized acquired by the multiple image acquisition devices and a preset gesture recognition method, a gesture label corresponding to the gesture to be recognized displayed in the gesture display area, and calibrates the gesture label corresponding to the image to be recognized is specifically configured as follows:

[0017] Based on a pre-trained gesture recognition model and the images to be recognized captured by the multiple image acquisition devices, the gesture label corresponding to the gesture to be recognized displayed in the gesture display area is determined, and the gesture label corresponding to the image to be recognized is calibrated, wherein the pre-trained gesture recognition model is: a model trained based on multiple groups of sample images containing sample gestures and sample gesture labels corresponding to each group of sample images, each group of sample images includes: images captured by multiple sample image acquisition devices for sample gestures from different angles in the same acquisition cycle.

[0018] Optionally, the processor is further configured to, before determining the gesture label corresponding to the gesture to be recognized displayed in the gesture display area based on the pre-trained gesture recognition model and the images to be recognized acquired by the multiple image acquisition devices, and calibrating the gesture label corresponding to the image to be recognized, train the pre-trained gesture recognition model to obtain the process, wherein the process includes:

[0019] Obtaining an initial second gesture recognition model;

[0020] Obtain multiple groups of sample images and sample gesture labels corresponding to each group of sample images, wherein each group of sample images includes: images captured by multiple sample image acquisition devices for sample gestures from the different angles in the same acquisition cycle;

[0021] The initial second gesture recognition model is trained using each group of sample images and the sample gesture labels corresponding to each group of sample images until the initial second gesture recognition model reaches a preset second convergence condition, thereby obtaining a pre-trained gesture recognition model for detecting gesture labels corresponding to gestures contained in the image.

[0022] In a second aspect, an embodiment of the present invention provides a method for calibrating a gesture label, which is applied to a processor of a gesture recognition system, wherein the gesture recognition system further includes a plurality of image acquisition devices for photographing the gesture display area from different angles; the method includes:

[0023] Obtaining images captured by multiple image acquisition devices in the same acquisition cycle for the gesture display area as images to be recognized, wherein the images to be recognized include: the gesture to be recognized displayed in the gesture display area;

[0024] Based on the images to be recognized captured by the multiple image acquisition devices and a preset gesture recognition method, a gesture label corresponding to the gesture to be recognized displayed in the gesture display area is determined, and the gesture label corresponding to the image to be recognized is calibrated.

[0025] Optionally, the step of determining a gesture label corresponding to the gesture to be recognized displayed in the gesture display area based on the images to be recognized acquired by the multiple image acquisition devices and a preset gesture recognition method, and calibrating the gesture label corresponding to the image to be recognized includes:

[0026] Sending the image to be recognized captured by each of the multiple image acquisition devices to a display device, so that the display device displays the image to be recognized captured by each of the multiple image acquisition devices, so that the user can input a gesture label corresponding to the gesture to be recognized contained in the target image to be recognized based on the gesture to be recognized contained in the target image to be recognized;

[0027] Obtaining the gesture label input by the user for the target image to be recognized;

[0028] The gesture label is determined as the gesture label corresponding to the gesture to be recognized contained in each image in the images to be recognized, and calibration is performed.

[0029] Optionally, after the step of determining the gesture label as a gesture label corresponding to the gesture to be recognized contained in each image in the image to be recognized and performing calibration, the method further includes:

[0030] Obtaining an initial first gesture recognition model;

[0031] Obtain multiple groups of sample gesture images containing sample gestures, and calibrated gesture labels corresponding to each group of sample gesture images, wherein each group of sample gesture images includes: images captured by the multiple image acquisition devices for the gesture display area within the same acquisition cycle, and the calibrated gesture labels corresponding to each group of sample gesture images are: gesture labels corresponding to the sample gestures displayed in the gesture display area determined by the processor based on the group of sample gesture images and the preset gesture recognition method;

[0032] The initial first gesture recognition model is trained using the multiple groups of sample gesture images and the calibrated gesture labels corresponding to each group of sample gesture images, as well as the images to be recognized and the gesture labels corresponding to them, until the initial first gesture recognition model reaches a preset first convergence condition, thereby obtaining a target gesture recognition model for detecting the gesture labels corresponding to the gestures contained in the images.

[0033] Optionally, the step of determining a gesture label corresponding to the gesture to be recognized displayed in the gesture display area based on the images to be recognized acquired by the multiple image acquisition devices and a preset gesture recognition method, and calibrating the gesture label corresponding to the image to be recognized includes:

[0034] Based on a pre-trained gesture recognition model and the images to be recognized captured by the multiple image acquisition devices, the gesture label corresponding to the gesture to be recognized displayed in the gesture display area is determined, and the gesture label corresponding to the image to be recognized is calibrated, wherein the pre-trained gesture recognition model is: a model trained based on multiple groups of sample images containing sample gestures and sample gesture labels corresponding to each group of sample images, each group of sample images includes: images captured by multiple sample image acquisition devices for sample gestures from different angles in the same acquisition cycle.

[0035] Optionally, before the step of determining, based on the pre-trained gesture recognition model and the images to be recognized acquired by the multiple image acquisition devices, gesture labels corresponding to the gestures to be recognized displayed in the gesture display area, and calibrating the gesture labels corresponding to the images to be recognized, the method further includes:

[0036] The process of training and obtaining the pre-trained gesture recognition model includes:

[0037] Obtaining an initial second gesture recognition model;

[0038] Obtain multiple groups of sample images and sample gesture labels corresponding to each group of sample images, wherein each group of sample images includes: images captured by multiple sample image acquisition devices for sample gestures from the different angles in the same acquisition cycle;

[0039] The initial second gesture recognition model is trained using each group of sample images and the sample gesture labels corresponding to each group of sample images until the initial second gesture recognition model reaches a preset second convergence condition, thereby obtaining a pre-trained gesture recognition model for detecting gesture labels corresponding to gestures contained in the image.

[0040] In a third aspect, an embodiment of the present invention provides a gesture label calibration device, which is applied to a gesture recognition system, wherein the gesture recognition system further includes a plurality of image acquisition devices for photographing the gesture display area from different angles; the device includes:

[0041] A first acquisition module is configured to obtain images captured by multiple image acquisition devices in the same acquisition cycle for the gesture display area as images to be recognized, wherein the images to be recognized include: the gesture to be recognized displayed in the gesture display area;

[0042] A determination and calibration module is configured to determine the gesture label corresponding to the gesture to be recognized displayed in the gesture display area based on the images to be recognized acquired by the multiple image acquisition devices and a preset gesture recognition method, and calibrate the gesture label corresponding to the image to be recognized.

[0043] Optionally, the determining and calibration module includes:

[0044] a sending unit configured to send the image to be recognized captured by each of the multiple image acquisition devices to a display device, so that the display device displays the image to be recognized captured by each of the multiple image acquisition devices, so that a user can input a gesture label corresponding to the gesture to be recognized contained in the target image to be recognized based on the gesture to be recognized contained in the target image to be recognized;

[0045] A first obtaining unit is configured to obtain the gesture label input by the user for the target image to be recognized;

[0046] The determination and calibration unit is configured to determine the gesture label as the gesture label corresponding to the gesture to be recognized contained in each image in the images to be recognized, and perform calibration.

[0047] Optionally, the device further comprises:

[0048] A second obtaining module is configured to obtain an initial first gesture recognition model after determining the gesture label as the gesture label corresponding to the gesture to be recognized contained in each image in the image to be recognized and calibrating it;

[0049] The third acquisition module is configured to obtain multiple groups of sample gesture images containing sample gestures, and calibrated gesture labels corresponding to each group of sample gesture images, wherein each group of sample gesture images includes: images collected by the multiple image acquisition devices for the gesture display area within the same acquisition cycle, and the calibrated gesture labels corresponding to each group of sample gesture images are: gesture labels corresponding to the sample gestures displayed in the gesture display area determined by the processor based on the group of sample gesture images and the preset gesture recognition method;

[0050] The first training module is configured to train the initial first gesture recognition model using the multiple groups of sample gesture images and the calibrated gesture labels corresponding to each group of sample gesture images, as well as the images to be recognized and the gesture labels corresponding to them, until the initial first gesture recognition model reaches a preset first convergence condition, thereby obtaining a target gesture recognition model for detecting the gesture labels corresponding to the gestures contained in the images.

[0051] Optionally, the calibration module is specifically configured as follows:

[0052] Based on a pre-trained gesture recognition model and the images to be recognized captured by the multiple image acquisition devices, the gesture label corresponding to the gesture to be recognized displayed in the gesture display area is determined, and the gesture label corresponding to the image to be recognized is calibrated, wherein the pre-trained gesture recognition model is: a model trained based on multiple groups of sample images containing sample gestures and sample gesture labels corresponding to each group of sample images, each group of sample images includes: images captured by multiple sample image acquisition devices for sample gestures from different angles in the same acquisition cycle.

[0053] Optionally, the device further comprises:

[0054] The model training module is configured to train the pre-trained gesture recognition model before determining the gesture label corresponding to the gesture to be recognized displayed in the gesture display area and calibrating the gesture label corresponding to the image to be recognized based on the pre-trained gesture recognition model and the images to be recognized collected by the multiple image acquisition devices, wherein the model training module is specifically configured as follows:

[0055] Obtaining an initial second gesture recognition model;

[0056] Obtain multiple groups of sample images and sample gesture labels corresponding to each group of sample images, wherein each group of sample images includes: images captured by multiple sample image acquisition devices for sample gestures from the different angles in the same acquisition cycle;

[0057] The initial second gesture recognition model is trained using each group of sample images and the sample gesture labels corresponding to each group of sample images until the initial second gesture recognition model reaches a preset second convergence condition, thereby obtaining a pre-trained gesture recognition model for detecting gesture labels corresponding to gestures contained in the image.

[0058] From the above content, it can be seen that an embodiment of the present invention provides a system, method and device for calibrating gesture labels, the system includes a processor, and multiple image acquisition devices, the processor is connected to the multiple image acquisition devices; the multiple image acquisition devices are configured to shoot a gesture display area from different angles; the processor is configured to obtain images captured by the multiple image acquisition devices for the gesture display area within the same acquisition cycle as images to be recognized; and based on the images to be recognized captured by the multiple image acquisition devices and a preset gesture recognition method, determine the gesture label corresponding to the gesture to be recognized displayed in the gesture display area, and calibrate the gesture label corresponding to the image to be recognized, wherein the image to be recognized includes: the gesture to be recognized displayed in the gesture display area.

[0059] By applying the embodiments of the present invention, the gesture label of the gesture to be recognized displayed in the gesture display area can be determined based on the preset gesture recognition method and the image to be recognized obtained when the gesture to be recognized displayed in the gesture display area is photographed from different angles, that is, the gesture labels corresponding to the gestures to be recognized contained in multiple images to be recognized can be determined at one time, and the calibration of the gesture labels corresponding to the gestures contained in the images can be realized, and the calibration of the gesture labels corresponding to the gestures to be recognized in the images to be recognized where the gestures are occluded can be realized, which can reduce the workload of calibrating the gesture labels corresponding to the images to be recognized to a certain extent. Of course, it is not necessary to achieve all the advantages described above at the same time when implementing any product or method of the present invention.

[0060] The innovative features of the embodiments of the present invention include:

[0061] 1. Based on a preset gesture recognition method and the images to be recognized obtained when the gesture to be recognized displayed in the gesture display area is photographed from different angles, the gesture label of the gesture to be recognized displayed in the gesture display area can be determined, that is, the gesture labels corresponding to the gestures to be recognized contained in multiple images to be recognized can be determined at one time, so as to realize the calibration of the gesture labels corresponding to the gestures contained in the images, and the calibration of the gesture labels corresponding to the gestures to be recognized in the images to be recognized where the gestures are occluded can be realized, which can reduce the workload of calibrating the gesture labels corresponding to the images to be recognized to a certain extent.

[0062] 2. Display multiple images to be identified acquired by multiple image acquisition devices through a display device, so that the user can select target images to be identified with a more correct angle, that is, a target image to be identified with a more positive gesture to be identified, from multiple images to be identified based on the gesture to be identified contained in the target image to be identified in each image to be identified, determine the gesture label corresponding to the gesture to be identified, that is, determine the gesture label corresponding to each image to be identified, and then calibrate to ensure the accuracy of the gesture label corresponding to the determined image to be identified, and to reduce the difficulty of calibrating the gesture label corresponding to the gesture to be identified in the image with an incorrect angle, that is, the image containing an occluded gesture, to a certain extent.

[0063] 3. Using multiple groups of sample gesture images and the calibrated gesture labels corresponding to each group of sample gesture images, as well as the group of images to be recognized and their corresponding gesture labels, train the initial first gesture recognition model to obtain a target gesture recognition model for detecting gesture labels corresponding to gestures contained in the images, so that the target gesture recognition model learns the features of each angle of the sample gestures and the gestures to be recognized, thereby improving the target gesture recognition model's recognition of gestures in the image, determining the accuracy of gesture labels corresponding to gestures in the image, and improving the accuracy of the recognition results of the target gesture recognition model.

[0064] 4. Based on multiple groups of sample images containing sample gestures and sample gesture labels corresponding to each group of sample images, a pre-trained gesture recognition model is obtained, that is, a pre-trained gesture recognition model that has learned the features of each angle of each sample gesture. The gesture to be recognized in the image to be recognized is recognized, the gesture label corresponding to the gesture to be recognized displayed in the gesture display area is determined, and the gesture label corresponding to the image to be recognized is calibrated to obtain a more accurate gesture label corresponding to the image to be recognized. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings without creative work.

[0066] Figure 1 A schematic diagram of the structure of a gesture label calibration system provided by an embodiment of the present invention;

[0067] Figure 2 A schematic diagram of the structure of a gesture label calibration system provided by an embodiment of the present invention;

[0068] Figure 3 A schematic diagram of a flow chart of a method for calibrating a gesture label provided by an embodiment of the present invention;

[0069] Figure 4 A schematic diagram of the structure of a gesture label calibration device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0070] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0071] It should be noted that the terms "including" and "having" and any variations thereof in the embodiments of the present invention and the accompanying drawings are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device comprising a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products or devices.

[0072] The present invention provides a system, method and device for calibrating gesture labels to achieve calibration of gesture labels corresponding to gestures contained in an image. The embodiments of the present invention are described in detail below.

[0073] Figure 1 A structural diagram of a gesture label calibration system provided in an embodiment of the present invention. The system may include a processor 110 and multiple image acquisition devices 120, wherein the processor 110 is connected to the multiple image acquisition devices 120; wherein the connection may be direct or indirect, which is possible.

[0074] A plurality of image acquisition devices 120 are configured to capture the gesture display area from different angles;

[0075] The processor 110 is configured to obtain images captured by multiple image acquisition devices for the gesture display area within the same acquisition cycle as images to be recognized; and based on the images to be recognized captured by the multiple image acquisition devices and a preset gesture recognition method, determine the gesture label corresponding to the gesture to be recognized displayed in the gesture display area, and calibrate the gesture label corresponding to the image to be recognized.

[0076] The image to be recognized includes: the gesture to be recognized displayed in the gesture display area.

[0077] Among them, the multiple image acquisition devices 120 can be various types of image acquisition devices, for example, they can be infrared image acquisition devices, that is, image acquisition devices that can acquire infrared images, and accordingly, the image to be identified is an infrared image. The multiple image acquisition devices 120 can also be color image acquisition devices, that is, image acquisition devices that can acquire color images, and accordingly, the image to be identified is a color image. In one case, the multiple image acquisition devices 120 can be set inside the vehicle to shoot the gestures displayed by the people inside the vehicle in the gesture display area. The processor 110 can be set in the vehicle-mounted device of the vehicle, or it can not be set in the vehicle-mounted device of the vehicle, which is all possible. In another case, the multiple image acquisition devices 120 can also be set in the target room to shoot the gestures displayed by the people in the target room in the gesture display area.

[0078] The gesture display area may be an area that can be captured by the multiple image acquisition devices 120. The gesture display area may display gestures.

[0079] After multiple image acquisition devices 120 acquire images for the gesture display area in the same acquisition cycle, they can be sent to the processor 110, and the processor 110 obtains the corresponding images as images to be recognized, wherein each image to be recognized contains the gesture displayed in the gesture display area, which can be called a gesture to be recognized. Furthermore, based on the preset gesture recognition method and the images to be recognized acquired by each image acquisition device 120, the processor 110 determines the gesture label corresponding to the gesture to be recognized displayed in the gesture display area, determines the gesture label corresponding to the gesture to be recognized displayed in the determined gesture display area as the gesture label corresponding to each image to be recognized, and records the corresponding relationship between each image to be recognized and the determined gesture label.

[0080] The gesture labels can be set according to the actual needs of the user. For example, the gesture labels may include but are not limited to: scissor hand gesture, index finger and middle finger together gesture, "eight" gesture, "six" gesture, etc. Among them, the "eight" gesture refers to the displayed gesture representing "eight", and the "six" gesture refers to the displayed gesture representing "six".

[0081] By applying the embodiments of the present invention, the gesture label of the gesture to be recognized displayed in the gesture display area can be determined based on a preset gesture recognition method and the images to be recognized obtained when the gesture to be recognized displayed in the gesture display area is photographed from different angles. That is, the gesture labels corresponding to the gestures to be recognized contained in multiple images to be recognized can be determined at one time, so as to calibrate the gesture labels corresponding to the gestures contained in the images, and can also calibrate the gesture labels corresponding to the gestures to be recognized in the images to be recognized where the gestures are occluded, which can reduce the workload of calibrating the gesture labels corresponding to the images to be recognized to a certain extent.

[0082] In another embodiment of the present invention, Figure 2 As shown, the system may further include a display device 130;

[0083] In the process of determining the gesture label corresponding to the gesture to be recognized displayed in the gesture display area based on the images to be recognized acquired by multiple image acquisition devices and the preset gesture recognition method, and calibrating the gesture label corresponding to the image to be recognized, the processor 110 is specifically configured as follows:

[0084] Sending the image to be recognized acquired by each of the plurality of image acquisition devices to the display device 130;

[0085] The display device 130 is configured to display the image to be recognized captured by each of the multiple image acquisition devices, so that the user can input a gesture label corresponding to the gesture to be recognized contained in the target image to be recognized based on the gesture to be recognized contained in the target image to be recognized;

[0086] The processor 110 obtains a gesture label input by a user for a target image to be recognized; determines the gesture label as a gesture label corresponding to the gesture to be recognized contained in each image in the image to be recognized, and performs calibration.

[0087] In this embodiment, the preset gesture recognition method is a manual recognition method. Accordingly, the gesture label calibration system may further include a display device 130. After the processor 110 obtains the image to be recognized collected by each image acquisition device 120 in the plurality of image acquisition devices 120, the image to be recognized may be sent to the display device 130. The display device 130 displays the image to be recognized collected by each image acquisition device 120. The user may determine the image to be recognized with a relatively positive angle, that is, the image to be recognized containing the front gesture to be recognized, from the image to be recognized displayed by the display device 130 as the target image to be recognized, and determine the corresponding gesture label based on the gesture to be recognized in the target image to be recognized, and input the gesture label. The processor 110 obtains the gesture label input by the user for the target image to be recognized, and the gesture label is the gesture label corresponding to the gesture to be recognized displayed in the gesture display area. Subsequently, the gesture label input by the user is determined as the gesture label corresponding to the gesture to be recognized contained in each image in the image to be recognized, and is calibrated. That is, the corresponding relationship between each image to be recognized and the gesture label is recorded. This ensures the accuracy of the gesture labels corresponding to the image to be identified, and reduces the difficulty of calibrating the gesture labels corresponding to the gestures to be identified in images with incorrect angles, i.e., images containing occluded gestures, to a certain extent.

[0088] In one case, multiple image acquisition devices simultaneously shoot the gesture display area from different angles. Some image acquisition devices may have poor shooting angles for the gesture to be recognized displayed in the gesture display area, resulting in the image captured by such image acquisition devices only including part of the content of the gesture to be recognized. For example: the gesture to be recognized is a gesture with the thumb and little finger open, and the index finger, middle finger and ring finger bent, that is, the actual corresponding gesture label is "six", and the images captured by some image acquisition devices may only contain the content of the thumb open, and the state of the little finger is blocked by the bent index finger, middle finger and ring finger, or, the images captured by some image acquisition devices may only contain the content of the little finger open, and the state of the thumb is blocked by the bent index finger, middle finger and ring finger. If the gesture to be recognized displayed in the image captured by such image acquisition devices alone is easily mislabeled, or even cannot be labeled.

[0089] In the embodiment of the present invention, since multiple image acquisition devices can simultaneously take all-round photos of the gesture display area from different angles, there are image acquisition devices among the multiple image acquisition devices that directly capture the gesture to be recognized. The images captured by such image acquisition devices contain all the contents of the gesture to be recognized, that is, the target image to be recognized. The user can mark the gesture label corresponding to the gesture to be recognized based on the target image to be recognized that contains all the contents of the gesture to be recognized. Furthermore, the electronic device can use the gesture label corresponding to the gesture to be recognized marked based on the target image to be recognized as the gesture label corresponding to the gesture to be recognized in the images simultaneously captured by the multiple image acquisition devices, so as to realize the calibration of the gesture label corresponding to the gesture to be recognized in the image to be recognized where the gesture is occluded.

[0090] In another embodiment of the present invention, the processor 110 may also be configured to: after determining the gesture label as the gesture label corresponding to the gesture contained in each image to be recognized and marking it, obtain an initial first gesture recognition model;

[0091] Obtain multiple groups of sample gesture images containing sample gestures, and calibrated gesture labels corresponding to each group of sample gesture images, wherein each group of sample gesture images includes: images captured by multiple image acquisition devices 120 for a gesture display area within the same acquisition cycle, and the calibrated gesture labels corresponding to each group of sample gesture images are: gesture labels corresponding to sample gestures displayed in the gesture display area determined by the processor 110 based on the group of sample gesture images and a preset gesture recognition method;

[0092] An initial first gesture recognition model is trained using multiple groups of sample gesture images and calibrated gesture labels corresponding to each group of sample gesture images, as well as images to be recognized and their corresponding gesture labels, until the initial first gesture recognition model reaches a preset first convergence condition, thereby obtaining a target gesture recognition model for detecting gesture labels corresponding to gestures contained in the image.

[0093] In this embodiment, after the processor 110 determines the gesture labels corresponding to the multiple images to be identified, it can determine whether the number of image groups whose corresponding gesture labels are currently determined using the preset gesture recognition method exceeds a certain number, and the image group can be the images collected by the multiple image acquisition devices 120; when it is determined that the number of image groups whose corresponding gesture labels are currently determined exceeds a certain number, each image in the image group whose corresponding gesture labels are determined can be used as a sample gesture image, and the corresponding gesture label is called a calibrated gesture label. Each group of sample gesture images contains sample gestures, and different images in the same group of sample gesture images contain different acquisition angles corresponding to the sample gestures.

[0094] After the processor 110 obtains multiple groups of sample gesture images containing sample gestures and the calibrated gesture labels corresponding to each group of sample gesture images, the multiple groups of sample gesture images and the calibrated gesture labels corresponding to each group of sample gesture images, as well as the images to be recognized and their corresponding gesture labels, can be respectively input into an initial first gesture recognition model, and the initial first gesture recognition model can be trained using the multiple groups of sample gesture images and the calibrated gesture labels corresponding to each group of sample gesture images, as well as the images to be recognized and their corresponding gesture labels, until the initial first gesture recognition model reaches a preset first convergence condition, thereby obtaining a target gesture recognition model for detecting gesture labels corresponding to gestures contained in images.

[0095] In one implementation, the initial first gesture recognition model may be a neural network model based on a deep learning algorithm, such as a convolutional neural network model. The initial first gesture recognition model may include a first feature extraction layer and a first feature classification layer. Accordingly, the process of training the initial first gesture recognition model may be:

[0096] Each group of sample gesture images or images to be recognized is respectively used as a group of model input images; each group of model input images is input into the first feature extraction layer, and the current values ​​of the model parameters of the first feature extraction layer are used to extract image features for each group of model input images to obtain feature images corresponding to each group of model input images; the feature images corresponding to each group of model input images are input into the first feature classification layer; the predicted gesture labels corresponding to each group of model input images are determined using the current values ​​of the model parameters of the first feature classification layer and the feature images corresponding to each group of model input images; the updated values ​​of the model parameters of the first feature extraction layer and the first feature classification layer are determined using a preset optimization algorithm, the predicted gesture labels corresponding to each group of model input images, and the calibrated gesture labels; the current values ​​of the model parameters of the first feature extraction layer and the first feature classification layer are corrected to the updated values ​​of the model parameters of the first feature extraction layer and the first feature classification layer; until the initial first gesture recognition model reaches the preset first convergence condition, a target gesture recognition model for detecting gesture labels corresponding to gestures contained in the image is obtained.

[0097] The preset optimization algorithm may be a gradient descent method. The preset first convergence condition may be: based on the initial first gesture recognition model obtained after adjusting the parameters, each image in the verification set is recognized and detected, and the number of predicted gesture labels corresponding to each image in the verification set, which are consistent with the calibrated gesture labels corresponding to each image in the verification set, accounts for the number of all images in the verification set, and exceeds a preset ratio threshold; or, it may be: based on the initial first gesture recognition model obtained after adjusting the parameters, each image in the verification set is recognized and detected, and the number of predicted gesture labels corresponding to each image in the verification set, which are consistent with the calibrated gesture labels corresponding to each image in the verification set, exceeds a preset number threshold; or, it may be: the number of times the model parameters of the first feature extraction layer and the first feature classification layer of the first gesture recognition model are iteratively adjusted exceeds a preset number of times, etc.

[0098] By training the target gesture recognition model using sample gesture images containing sample gestures at various angles and their calibrated gesture labels, as well as the image to be recognized and its corresponding gesture labels, the features of the sample gestures and gestures to be recognized at various angles can be learned. Thus, when the target gesture recognition model recognizes the gestures in the image, the accuracy of the gesture labels corresponding to the gestures in the image can be improved, that is, the accuracy of the recognition results of the target gesture recognition model can be improved.

[0099] In another embodiment of the present invention, the processor 110, based on the images to be recognized acquired by multiple image acquisition devices and the preset gesture recognition method, determines the gesture label corresponding to the gesture to be recognized displayed in the gesture display area, and calibrates the gesture label corresponding to the image to be recognized, is specifically configured as follows:

[0100] Based on a pre-trained gesture recognition model and images to be recognized captured by multiple image acquisition devices 120, the gesture label corresponding to the gesture to be recognized displayed in the gesture display area is determined, and the gesture label corresponding to the image to be recognized is calibrated, wherein the pre-trained gesture recognition model is: a model trained based on multiple groups of sample images containing sample gestures and sample gesture labels corresponding to each group of sample images, each group of sample images includes: images captured by multiple sample image acquisition devices for sample gestures from different angles in the same acquisition cycle.

[0101] The multiple sample image acquisition devices may be the multiple image acquisition devices 120 mentioned above, or multiple other image acquisition devices of the same type with the same shooting angle as the multiple image acquisition devices 120 mentioned above. The same shooting angles may refer to: the angle error between the shooting angles is within a preset allowable range. The same type may refer to: the same image acquisition device for capturing infrared images, or the same image acquisition device for capturing color images.

[0102] In this embodiment, the processor 110 can directly use a pre-trained gesture recognition model that is trained based on multiple groups of sample images containing sample gestures and sample gesture labels corresponding to each group of sample images to perform recognition detection on images to be recognized acquired by multiple image acquisition devices, determine the gesture labels corresponding to the gestures to be recognized displayed in the gesture display area, and calibrate the gesture labels corresponding to the images to be recognized.

[0103] Specifically, the pre-trained gesture recognition model may include a second feature extraction layer and a second feature classification layer. The processor inputs a plurality of images to be recognized into the second feature extraction layer, and uses the values ​​of the model parameters of the second feature extraction layer to extract image features of the plurality of images to be recognized, and obtains feature images corresponding to the plurality of images to be recognized; inputs the feature images of the plurality of images to be recognized into the second feature classification layer, and uses the values ​​of the model parameters of the second feature classification layer and the feature images of the plurality of images to be recognized to determine the gesture labels corresponding to the plurality of images to be recognized, that is, the gesture labels corresponding to the gestures to be recognized displayed in the gesture display area contained in the images to be recognized. And the corresponding relationship between each image to be recognized and the gesture label is recorded.

[0104] In another embodiment of the present invention, the processor 110 is further configured to train a process of obtaining a pre-trained gesture recognition model before determining a gesture label corresponding to a gesture to be recognized displayed in a gesture display area based on the pre-trained gesture recognition model and the images to be recognized collected by multiple image acquisition devices, and calibrating the gesture label corresponding to the image to be recognized, wherein the process includes:

[0105] Obtaining an initial second gesture recognition model;

[0106] Obtain multiple groups of sample images and sample gesture labels corresponding to each group of sample images, wherein each group of sample images includes: images captured by multiple sample image acquisition devices from different angles for sample gestures in the same acquisition cycle;

[0107] An initial second gesture recognition model is trained using each group of sample images and the sample gesture labels corresponding to each group of sample images until the initial second gesture recognition model reaches a preset second convergence condition, thereby obtaining a pre-trained gesture recognition model for detecting gesture labels corresponding to gestures contained in the image.

[0108] In this embodiment, the initial second gesture recognition model may be a neural network model based on a deep learning algorithm, such as a convolutional neural network model.

[0109] Accordingly, the process of training the initial second gesture recognition model may be:

[0110] Input each group of sample images into the second feature extraction layer, and use the current values ​​of the model parameters of the second feature extraction layer to extract image features for each group of sample images to obtain feature images corresponding to each group of sample images; input the feature images corresponding to each group of sample images into the second feature classification layer; use the current values ​​of the model parameters of the second feature classification layer and the feature images corresponding to each group of sample images to determine the predicted gesture labels corresponding to each group of sample images; use the preset optimization algorithm, the predicted gesture labels corresponding to each group of sample images, and the sample gesture labels to determine the updated values ​​of the model parameters of the second feature extraction layer and the second feature classification layer; correct the current values ​​of the model parameters of the second feature extraction layer and the second feature classification layer to the updated values ​​of the model parameters of the second feature extraction layer and the second feature classification layer; until the initial second gesture recognition model reaches the preset second convergence condition, a pre-trained gesture recognition model for detecting gesture labels corresponding to gestures contained in the image is obtained.

[0111] The preset optimization algorithm may be a gradient descent method. The preset second convergence condition may be: based on the initial second gesture recognition model obtained after adjusting the parameters, each image in the verification set is recognized and detected, and the number of predicted gesture labels corresponding to each image in the verification set, which are consistent with the sample gesture labels corresponding to each image in the verification set, accounts for the number of all images in the verification set, and exceeds a preset ratio threshold; or, it may be: based on the initial second gesture recognition model obtained after adjusting the parameters, each image in the verification set is recognized and detected, and the number of predicted gesture labels corresponding to each image in the verification set, which are consistent with the sample gesture labels corresponding to each image in the verification set, exceeds a preset number threshold; or, it may be: the number of times the model parameters of the second feature extraction layer and the second feature classification layer of the second gesture recognition model are iteratively adjusted exceeds a preset number of times, etc.

[0112] Corresponding to the above system embodiment, the embodiment of the present invention provides a gesture label calibration method, which is applied to a processor of a gesture recognition system. The gesture recognition system may also include multiple image acquisition devices for photographing the gesture display area from different angles, and the multiple image acquisition devices are connected to the processor; Figure 3 As shown, the method may include the following steps:

[0113] S301: Obtain images captured by multiple image acquisition devices in the same acquisition cycle for the gesture display area as images to be recognized.

[0114] The image to be recognized includes: the gesture to be recognized displayed in the gesture display area;

[0115] S302: Based on the images to be recognized acquired by multiple image acquisition devices and a preset gesture recognition method, determine a gesture label corresponding to the gesture to be recognized displayed in the gesture display area, and calibrate the gesture label corresponding to the image to be recognized.

[0116] By applying the embodiments of the present invention, the gesture label of the gesture to be recognized displayed in the gesture display area can be determined based on a preset gesture recognition method and the images to be recognized obtained when the gesture to be recognized displayed in the gesture display area is photographed from different angles. That is, the gesture labels corresponding to the gestures to be recognized contained in multiple images to be recognized can be determined at one time, so as to calibrate the gesture labels corresponding to the gestures contained in the images, and can also calibrate the gesture labels corresponding to the gestures to be recognized in the images to be recognized where the gestures are occluded, which can reduce the workload of calibrating the gesture labels corresponding to the images to be recognized to a certain extent.

[0117] In another embodiment of the present invention, the S302 may include the following steps 011-013:

[0118] 011: Sending the image to be recognized captured by each of the multiple image acquisition devices to a display device, so that the display device displays the image to be recognized captured by each of the multiple image acquisition devices, so that the user can input a gesture label corresponding to the gesture to be recognized contained in the target image to be recognized based on the gesture to be recognized contained in the target image to be recognized;

[0119] 012: Obtaining the gesture label input by the user for the target image to be recognized;

[0120] 013: Determine the gesture label as the gesture label corresponding to the gesture to be recognized contained in each image in the images to be recognized, and calibrate it.

[0121] In another embodiment of the present invention, after step 013, the method may further include steps 021-023:

[0122] 021: Obtain an initial first gesture recognition model;

[0123] 022: obtaining multiple groups of sample gesture images including sample gestures, and calibrated gesture labels corresponding to each group of sample gesture images, wherein each group of sample gesture images includes: images captured by the multiple image acquisition devices for the gesture display area within the same acquisition cycle, and the calibrated gesture labels corresponding to each group of sample gesture images are: gesture labels corresponding to the sample gestures displayed in the gesture display area, determined by the processor based on the group of sample gesture images and the preset gesture recognition method;

[0124] 023: Using the multiple groups of sample gesture images and the calibrated gesture labels corresponding to each group of sample gesture images, as well as the image to be recognized and its corresponding gesture label, train the initial first gesture recognition model until the initial first gesture recognition model reaches a preset first convergence condition, and obtain a target gesture recognition model for detecting the gesture label corresponding to the gesture contained in the image.

[0125] In another embodiment of the present invention, the S302 may include the following step 031:

[0126] 031: Based on a pre-trained gesture recognition model and the images to be recognized acquired by the multiple image acquisition devices, determine the gesture label corresponding to the gesture to be recognized displayed in the gesture display area, and calibrate the gesture label corresponding to the image to be recognized, wherein the pre-trained gesture recognition model is: a model trained based on multiple groups of sample images containing sample gestures and sample gesture labels corresponding to each group of sample images, each group of sample images includes: images acquired by multiple sample image acquisition devices for sample gestures from the different angles in the same acquisition cycle.

[0127] In another embodiment of the present invention, before step 031, the method may further include:

[0128] The process of training and obtaining the pre-trained gesture recognition model may include the following steps 041-043:

[0129] 041: Obtain an initial second gesture recognition model;

[0130] 042: Obtain multiple groups of sample images and sample gesture labels corresponding to each group of sample images, wherein each group of sample images includes: images captured by multiple sample image acquisition devices for the sample gestures from the different angles in the same acquisition period;

[0131] 043: Using each group of sample images and the sample gesture labels corresponding to each group of sample images, the initial second gesture recognition model is trained until the initial second gesture recognition model reaches a preset second convergence condition, thereby obtaining a pre-trained gesture recognition model for detecting gesture labels corresponding to gestures contained in the image.

[0132] Corresponding to the above system embodiment, an embodiment of the present invention provides a gesture recognition device, which is applied to a gesture recognition system, wherein the gesture recognition system further comprises a plurality of image acquisition devices for photographing the gesture display area from different angles; Figure 4 As shown, the device comprises:

[0133] The first acquisition module 410 is configured to obtain images captured by multiple image acquisition devices in the same acquisition cycle for the gesture display area as images to be recognized, wherein the images to be recognized include: the gesture to be recognized displayed in the gesture display area;

[0134] The determination and calibration module 420 is configured to determine the gesture label corresponding to the gesture to be recognized displayed in the gesture display area based on the images to be recognized acquired by the multiple image acquisition devices and the preset gesture recognition method, and calibrate the gesture label corresponding to the image to be recognized.

[0135] By applying the embodiments of the present invention, the gesture label of the gesture to be recognized displayed in the gesture display area can be determined based on a preset gesture recognition method and the images to be recognized obtained when the gesture to be recognized displayed in the gesture display area is photographed from different angles. That is, the gesture labels corresponding to the gestures to be recognized contained in multiple images to be recognized can be determined at one time, so as to calibrate the gesture labels corresponding to the gestures contained in the images, and can also calibrate the gesture labels corresponding to the gestures to be recognized in the images to be recognized where the gestures are occluded, which can reduce the workload of calibrating the gesture labels corresponding to the images to be recognized to a certain extent.

[0136] In another embodiment of the present invention, the determination and calibration module 420 includes:

[0137] a sending unit (not shown in the figure), configured to send the image to be recognized captured by each of the multiple image acquisition devices to a display device, so that the display device displays the image to be recognized captured by each of the multiple image acquisition devices, so that the user can input a gesture label corresponding to the gesture to be recognized contained in the target image to be recognized among the images to be recognized based on the gesture to be recognized contained in the target image to be recognized;

[0138] A first obtaining unit (not shown in the figure) is configured to obtain the gesture label input by the user for the target image to be recognized;

[0139] The determination and calibration unit (not shown in the figure) is configured to determine the gesture label as the gesture label corresponding to the gesture to be recognized contained in each image in the images to be recognized, and perform calibration.

[0140] In another embodiment of the present invention, the device further comprises:

[0141] A second obtaining module (not shown in the figure) is configured to obtain an initial first gesture recognition model after determining the gesture label as the gesture label corresponding to the gesture to be recognized contained in each image in the image to be recognized and calibrating it;

[0142] The third acquisition module (not shown in the figure) is configured to obtain multiple groups of sample gesture images containing sample gestures, and calibrated gesture labels corresponding to each group of sample gesture images, wherein each group of sample gesture images includes: images collected by the multiple image acquisition devices for the gesture display area within the same acquisition cycle, and the calibrated gesture labels corresponding to each group of sample gesture images are: gesture labels corresponding to the sample gestures displayed in the gesture display area determined by the processor based on the group of sample gesture images and the preset gesture recognition method;

[0143] The first training module (not shown in the figure) is configured to use the multiple groups of sample gesture images and the calibrated gesture labels corresponding to each group of sample gesture images, as well as the images to be recognized and the gesture labels corresponding to them, to train the initial first gesture recognition model until the initial first gesture recognition model reaches a preset first convergence condition, thereby obtaining a target gesture recognition model for detecting the gesture labels corresponding to the gestures contained in the images.

[0144] In another embodiment of the present invention, the determination and calibration module 420 is specifically configured as follows:

[0145] Based on a pre-trained gesture recognition model and the images to be recognized captured by the multiple image acquisition devices, the gesture label corresponding to the gesture to be recognized displayed in the gesture display area is determined, and the gesture label corresponding to the image to be recognized is calibrated, wherein the pre-trained gesture recognition model is: a model trained based on multiple groups of sample images containing sample gestures and sample gesture labels corresponding to each group of sample images, each group of sample images includes: images captured by multiple sample image acquisition devices for sample gestures from different angles in the same acquisition cycle.

[0146] In another embodiment of the present invention, the device further comprises:

[0147] The model training module (not shown in the figure) is configured to train the pre-trained gesture recognition model before determining the gesture label corresponding to the gesture to be recognized displayed in the gesture display area based on the pre-trained gesture recognition model and the images to be recognized collected by the multiple image acquisition devices, and calibrating the gesture label corresponding to the image to be recognized, wherein the model training module is specifically configured as follows:

[0148] Obtaining an initial second gesture recognition model;

[0149] Obtain multiple groups of sample images and sample gesture labels corresponding to each group of sample images, wherein each group of sample images includes: images captured by multiple sample image acquisition devices for sample gestures from the different angles in the same acquisition cycle;

[0150] The initial second gesture recognition model is trained using each group of sample images and the sample gesture labels corresponding to each group of sample images until the initial second gesture recognition model reaches a preset second convergence condition, thereby obtaining a pre-trained gesture recognition model for detecting gesture labels corresponding to gestures contained in the image.

[0151] The above-mentioned device and method embodiments correspond to the system embodiment and have the same technical effects as the system embodiment. For specific descriptions, please refer to the system embodiment. The device embodiment is obtained based on the method embodiment. For specific descriptions, please refer to the system and method embodiment part, which will not be repeated here. It can be understood by those skilled in the art that the accompanying drawings are only schematic diagrams of one embodiment, and the modules or processes in the accompanying drawings are not necessarily required to implement the present invention.

[0152] Those skilled in the art can understand that the modules in the device in the embodiment can be distributed in the device in the embodiment according to the description of the embodiment, or can be changed accordingly and located in one or more devices different from the embodiment. The modules in the above embodiment can be combined into one module, or can be further divided into multiple sub-modules.

[0153] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention 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 invention.

Claims

1. A gesture label calibration system, characterized in that: The system includes a processor and a plurality of image acquisition devices, wherein the processor is connected to the plurality of image acquisition devices; The multiple image acquisition devices are configured to photograph the gesture display area from different angles; The processor is configured to obtain images captured by the multiple image acquisition devices in the same acquisition cycle for the gesture display area as images to be recognized; Based on the images to be recognized collected by the multiple image acquisition devices and the preset gesture recognition method, determine the gesture label corresponding to the gesture to be recognized displayed in the gesture display area, and calibrate the gesture label corresponding to the image to be recognized, wherein the image to be recognized includes: the gesture to be recognized displayed in the gesture display area; The system also includes a display device; The processor is specifically configured to: determine the gesture label corresponding to the gesture to be recognized displayed in the gesture display area based on the images to be recognized acquired by the multiple image acquisition devices and the preset gesture recognition method, and mark the gesture label corresponding to the image to be recognized. Sending the image to be recognized acquired by each of the plurality of image acquisition devices to the display device; The display device is configured to display the image to be recognized captured by each of the multiple image acquisition devices, so that the user can input a gesture label corresponding to the gesture to be recognized contained in the target image to be recognized among the images to be recognized based on the gesture to be recognized contained in the target image to be recognized; The processor obtains the gesture label input by the user for the target image to be recognized; The gesture label is determined as the gesture label corresponding to the gesture to be recognized contained in each image in the images to be recognized, and calibration is performed.

2. The system according to claim 1, characterized in that The processor is further configured to: after determining the gesture label as the gesture label corresponding to the gesture contained in each image in the to-be-recognized image and marking it, obtain an initial first gesture recognition model; Obtain multiple groups of sample gesture images containing sample gestures, and calibrated gesture labels corresponding to each group of sample gesture images, wherein each group of sample gesture images includes: images captured by the multiple image acquisition devices for the gesture display area within the same acquisition cycle, and the calibrated gesture labels corresponding to each group of sample gesture images are: gesture labels corresponding to the sample gestures displayed in the gesture display area determined by the processor based on the group of sample gesture images and the preset gesture recognition method; The initial first gesture recognition model is trained using multiple groups of sample gesture images and calibrated gesture labels corresponding to each group of sample gesture images, as well as the image to be recognized and its corresponding gesture label, until the initial first gesture recognition model reaches a preset first convergence condition, thereby obtaining a target gesture recognition model for detecting the gesture label corresponding to the gesture contained in the image.

3. The system according to claim 1, characterized in that The process in which the processor determines the gesture label corresponding to the gesture to be recognized displayed in the gesture display area based on the images to be recognized acquired by the multiple image acquisition devices and the preset gesture recognition method, and marks the gesture label corresponding to the image to be recognized, is specifically configured as follows: Based on a pre-trained gesture recognition model and the images to be recognized captured by the multiple image acquisition devices, the gesture label corresponding to the gesture to be recognized displayed in the gesture display area is determined, and the gesture label corresponding to the image to be recognized is calibrated, wherein the pre-trained gesture recognition model is: a model trained based on multiple groups of sample images containing sample gestures and sample gesture labels corresponding to each group of sample images, each group of sample images includes: images captured by multiple sample image acquisition devices for sample gestures from different angles in the same acquisition cycle.

4. The system according to claim 3, characterized in that The processor is further configured to train the pre-trained gesture recognition model before determining the gesture label corresponding to the gesture to be recognized displayed in the gesture display area based on the pre-trained gesture recognition model and the images to be recognized acquired by the multiple image acquisition devices, and calibrating the gesture label corresponding to the image to be recognized, wherein the process includes: Obtaining an initial second gesture recognition model; Obtain multiple groups of sample images and sample gesture labels corresponding to each group of sample images, wherein each group of sample images includes: images captured by multiple sample image acquisition devices for sample gestures from the different angles in the same acquisition cycle; The initial second gesture recognition model is trained using each group of sample images and the sample gesture labels corresponding to each group of sample images until the initial second gesture recognition model reaches a preset second convergence condition, thereby obtaining a pre-trained gesture recognition model for detecting gesture labels corresponding to gestures contained in the image.

5. A method for calibrating a gesture label, characterized in that: A processor applied to a gesture recognition system, wherein the gesture recognition system further comprises a plurality of image acquisition devices for photographing a gesture display area from different angles; the method comprises: Obtaining images captured by multiple image acquisition devices in the same acquisition cycle for the gesture display area as images to be recognized, wherein the images to be recognized include: the gesture to be recognized displayed in the gesture display area; Based on the images to be recognized acquired by the multiple image acquisition devices and the preset gesture recognition method, determining a gesture label corresponding to the gesture to be recognized displayed in the gesture display area, and calibrating the gesture label corresponding to the image to be recognized; The step of determining the gesture label corresponding to the gesture to be recognized displayed in the gesture display area based on the images to be recognized collected by the multiple image acquisition devices and the preset gesture recognition method, and calibrating the gesture label corresponding to the image to be recognized includes: Sending the image to be recognized captured by each of the multiple image acquisition devices to a display device, so that the display device displays the image to be recognized captured by each of the multiple image acquisition devices, so that the user can input a gesture label corresponding to the gesture to be recognized contained in the target image to be recognized among the images to be recognized based on the gesture to be recognized contained in the target image to be recognized; Obtaining the gesture label input by the user for the target image to be recognized; The gesture label is determined as the gesture label corresponding to the gesture to be recognized contained in each image in the images to be recognized, and calibration is performed.

6. The method according to claim 5, characterized in that After the step of determining the gesture label as the gesture label corresponding to the gesture to be recognized contained in each image in the image to be recognized and performing calibration, the method further includes: Obtaining an initial first gesture recognition model; Obtain multiple groups of sample gesture images containing sample gestures, and calibrated gesture labels corresponding to each group of sample gesture images, wherein each group of sample gesture images includes: images captured by the multiple image acquisition devices for the gesture display area within the same acquisition cycle, and the calibrated gesture labels corresponding to each group of sample gesture images are: gesture labels corresponding to the sample gestures displayed in the gesture display area determined by the processor based on the group of sample gesture images and the preset gesture recognition method; The initial first gesture recognition model is trained using multiple groups of sample gesture images and calibrated gesture labels corresponding to each group of sample gesture images, as well as the image to be recognized and its corresponding gesture label, until the initial first gesture recognition model reaches a preset first convergence condition, thereby obtaining a target gesture recognition model for detecting the gesture label corresponding to the gesture contained in the image.

7. The method according to claim 5, characterized in that The step of determining the gesture label corresponding to the gesture to be recognized displayed in the gesture display area based on the images to be recognized collected by the multiple image acquisition devices and the preset gesture recognition method, and calibrating the gesture label corresponding to the image to be recognized includes: Based on a pre-trained gesture recognition model and the images to be recognized captured by the multiple image acquisition devices, the gesture label corresponding to the gesture to be recognized displayed in the gesture display area is determined, and the gesture label corresponding to the image to be recognized is calibrated, wherein the pre-trained gesture recognition model is: a model trained based on multiple groups of sample images containing sample gestures and sample gesture labels corresponding to each group of sample images, each group of sample images includes: images captured by multiple sample image acquisition devices for sample gestures from different angles in the same acquisition cycle.

8. A gesture label calibration device, characterized in that: Applied to a gesture recognition system, the gesture recognition system further includes a plurality of image acquisition devices for photographing a gesture display area from different angles; the device includes: A first acquisition module is configured to obtain images captured by multiple image acquisition devices in the same acquisition cycle for the gesture display area as images to be recognized, wherein the images to be recognized include: the gesture to be recognized displayed in the gesture display area; a determination and calibration module, configured to determine a gesture label corresponding to the gesture to be recognized displayed in the gesture display area based on the images to be recognized acquired by the multiple image acquisition devices and a preset gesture recognition method, and calibrate the gesture label corresponding to the image to be recognized; The determination and calibration module comprises: a sending unit configured to send the image to be recognized captured by each of the multiple image acquisition devices to a display device, so that the display device displays the image to be recognized captured by each of the multiple image acquisition devices, so that a user can input a gesture label corresponding to the gesture to be recognized contained in the target image to be recognized among the images to be recognized based on the gesture to be recognized contained in the target image to be recognized; A first obtaining unit is configured to obtain the gesture label input by the user for the target image to be recognized; The determination and calibration unit is configured to determine the gesture label as the gesture label corresponding to the gesture to be recognized contained in each image in the images to be recognized, and perform calibration.