Adjustment method and device based on human body characteristic parameters, computer equipment, readable storage medium and program product

Abstract

The invention relates to an adjustment method and device based on human body characteristic parameters, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: acquiring a to-be-processed image; extracting features from the to-be-processed image to obtain a target human body feature parameter map; inputting the target human body characteristic parameter diagram into a trained position recognition model, and obtaining a position recognition result output by the position recognition model; the position identification model is generated after training a sample set based on a historical human body characteristic parameter graph; and adjusting the position of the corresponding target object in the target area according to the position identification result. By adopting the method, the regulation preparation can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of machine learning models, and in particular to an adjustment method, apparatus, computer equipment, computer-readable storage medium and computer program product based on human body characteristic parameters. Background Art

[0002] With the development of computer technology, machine learning model technology has emerged and machine learning models have been widely used.

[0003] In traditional technology, when a user enters a bank's self-service area, he or she needs to manually adjust the supporting facilities in the area, such as the height of the table, the height of the chair, etc.

[0004] However, the current adjustment method has the problem of low adjustment accuracy and requires multiple manual adjustments. Summary of the invention

[0005] Based on this, it is necessary to provide a method, device, computer equipment, computer-readable storage medium and computer program product based on human characteristic parameters that can improve accuracy in order to solve the above technical problems.

[0006] In a first aspect, the present application provides a method for adjusting based on human body characteristic parameters, the method comprising:

[0007] Get the image to be processed;

[0008] Extract features from the image to be processed to obtain a target human body feature parameter map;

[0009] Input the target human body feature parameter map into a trained position recognition model, and obtain the position recognition result output by the position recognition model; the position recognition model is generated after training based on the historical human body feature parameter map as a sample set;

[0010] The position of the corresponding target object in the target area is adjusted according to the position recognition result.

[0011] In one embodiment, extracting features from the image to be processed to obtain a target human feature parameter map includes:

[0012] Extracting initial human body features from the image to be processed to obtain initial feature maps;

[0013] Performing a feature fusion operation on each of the initial feature maps to obtain a target feature map corresponding to each of the initial feature maps;

[0014] Each of the target feature maps is detected to obtain a target human body and a corresponding target human body feature parameter map.

[0015] In one embodiment, the initial feature map includes initial feature maps of different sizes; the step of performing a feature fusion operation on each of the initial feature maps to obtain a target feature map corresponding to each of the initial feature maps includes:

[0016] Acquire a first order; wherein the first order is based on the order in which the initial feature maps are arranged from small to large in size;

[0017] Performing feature extraction on the first initial feature map to obtain a first intermediate feature map corresponding to the first initial feature map; wherein the first initial feature map is a feature map with the smallest or largest size;

[0018] The first intermediate feature map and the next initial feature map of the first initial feature map are subjected to feature fusion according to the first order to obtain a first target feature map corresponding to the next initial feature map.

[0019] In one embodiment, performing a feature fusion operation on each of the initial feature maps to obtain a target feature map corresponding to each of the initial feature maps includes:

[0020] Acquire a second order, wherein the second order is an order in which the first target feature graphs are arranged from large to small in terms of size;

[0021] Taking the first target feature map with the largest size as a new first initial feature map, performing feature extraction on the new first initial feature map, and obtaining a new first intermediate feature map corresponding to the new initial feature map;

[0022] The new first intermediate feature map and the previous initial feature map of the new first initial feature map are subjected to feature fusion in the second order to obtain a second target feature map corresponding to the previous initial feature map.

[0023] In one embodiment, before inputting the human body feature parameters into the trained position recognition model, the method includes:

[0024] Acquire a historical human body characteristic parameter map and a corresponding position tag; the historical human body characteristic parameter map includes a human body local characteristic parameter map;

[0025] Inputting the human body local feature parameter map into an initial position recognition model, and obtaining an initial predicted position label output by the initial position recognition model;

[0026] Calculating a total loss value for the initial predicted position label and the position label by using a preset total loss function; wherein the total loss function value represents the difference between the initial predicted position label and the position label, and the confidence of the local feature parameter map of the human body;

[0027] The adjustable parameters in the initial position recognition model are adjusted according to the total loss value until the total loss value meets the convergence condition, and the initial position recognition model when the total loss value meets the convergence condition is used as the target initial position recognition model.

[0028] In one embodiment, calculating the total loss value of the initial predicted position label and the position label by using a preset total loss function includes:

[0029] Calculating a first loss value for the initial predicted position label and the position label by using a preset first loss function; the first loss value represents the difference between the initial predicted position label and the position label;

[0030] Calculating a first loss value for the initial predicted position label and the position label by using a preset second loss function; the second loss value represents the confidence of the human body local feature parameter map;

[0031] The first loss value and the second loss value are summed to obtain the total loss value.

[0032] In a second aspect, the present application further provides an adjustment device based on human body characteristic parameters, the device comprising:

[0033] An acquisition module, used for acquiring an image to be processed;

[0034] An extraction module is used to extract human body feature parameters from the image to be processed to obtain a target human body feature parameter map;

[0035] An input module, used to input the target human body feature parameter map into a trained position recognition model, and obtain the position recognition result output by the position recognition model; the position recognition model is generated after training based on the historical human body feature parameter map as a sample set;

[0036] The adjustment module is used to adjust the position of the corresponding target object in the target area according to the position recognition result.

[0037] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0038] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the above method when executed by a processor.

[0039] In a fifth aspect, the present application also provides a computer program product, including a computer program, which implements the steps of the above method when executed by a processor.

[0040] The above-mentioned adjustment method, device, computer equipment, computer-readable storage medium and computer program product based on human feature parameters first extract features from the acquired image to be processed to obtain a target human feature parameter map; secondly, the target human feature parameter map is input into a trained position recognition model, and the position recognition result output by the position recognition model is obtained; the position recognition model is generated after training based on the historical human feature parameter map as a sample set, which can improve the accuracy of position recognition; finally, the positions of corresponding target objects such as tables, chairs, lights, etc. in the target area are adjusted according to the position recognition results output by the position recognition model, the accurate position of the human body is obtained according to the human feature parameters, and the positions of corresponding objects in the target area are adjusted according to the accurate position of the human body, thereby achieving the accuracy of position adjustment of target objects in the target area. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0042] Figure 1 is a flow chart of an adjustment method based on human body characteristic parameters in one embodiment;

[0043] Figure 2 A schematic diagram of a process for obtaining a target human body characteristic parameter map in one embodiment;

[0044] Figure 3 is a schematic diagram of a flow chart of a feature fusion operation in one embodiment;

[0045] Figure 4 is a flowchart of a feature fusion operation in another embodiment;

[0046] Figure 5 A schematic diagram of a process of model training and adjustment in one embodiment;

[0047] Figure 6 A schematic diagram of a process for calculating a total loss value in one embodiment;

[0048] Figure 7 is a structural block diagram of an adjustment device based on human body characteristic parameters in one embodiment;

[0049] Figure 8 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0051] In an exemplary embodiment, Figure 1 As shown, a method for adjusting based on human body characteristic parameters is provided, which can be applied to a server or a terminal device. The method is described by taking a server as an example, and includes the following steps S102 to S108. Among them:

[0052] Step S102, obtaining an image to be processed.

[0053] The image to be processed is an image captured and processed by an image acquisition device, such as a video camera, when the target human body enters the target area.

[0054] Optionally, before obtaining the image to be processed, the server obtains initial image information, which is collected by a camera or a still camera; the initial image information may be an image, a group of images or a video. If the initial image information is a video, the server captures an image from the video as the image to be processed.

[0055] Optionally, the server may also perform standardization processing on the initial image information to obtain an image to be processed.

[0056] Step S104, extracting features from the image to be processed to obtain a target human body feature parameter map.

[0057] Optionally, the server can extract features from the image to be processed through a backbone network, and perform feature fusion on the extracted features to obtain a target human feature parameter map. The backbone network is a neural network that extracts features from an input image through multiple network layers based on deep learning. The backbone network can be a network such as VGG16, Resnet50, Resnet101, and MobileNet.

[0058] Step S106, input the target human body feature parameter map into the trained position recognition model, and obtain the position recognition result output by the position recognition model; the position recognition model is generated after training based on the historical human body feature parameter map as a sample set.

[0059] Optionally, before inputting the target human body feature parameter map into the trained position recognition model, the server generates a position recognition model after training the sample set based on the historical human body feature parameter map.

[0060] Optionally, after the server obtains the target human body feature parameter map, the target human body feature parameter map is input into a trained position recognition model, and the trained position recognition model outputs a position recognition result corresponding to the target human body feature parameter map. For example, the position recognition model outputs the specific position of the target human body a in the target area, and the position can be marked with two-dimensional coordinates, three-dimensional coordinates or other marking methods.

[0061] Step S108: adjusting the position of the corresponding target object in the target area according to the position recognition result.

[0062] The target area may be an independent or non-independent business processing area. The target object is an item in the business processing area, such as a smart desk, a smart office chair, a lamp, and a smart terminal on the smart desk.

[0063] Optionally, the server adjusts the positions of the smart desk, smart office chair, lamp, and smart terminal on the smart desk according to the position recognition result to more accurately adapt to the target human body.

[0064] In the above-mentioned adjustment method based on human body feature parameters, first, features are extracted from the acquired image to be processed to obtain a target human body feature parameter map; secondly, the target human body feature parameter map is input into a trained position recognition model, and the position recognition result output by the position recognition model is obtained; the position recognition model is generated after training based on the historical human body feature parameter map as a sample set, which can improve the accuracy of position recognition; finally, the positions of corresponding target objects such as tables, chairs, lights, etc. in the target area are adjusted according to the position recognition results output by the position recognition model, the accurate position of the human body is obtained according to the human body feature parameters, and the positions of corresponding objects in the target area are adjusted according to the accurate position of the human body, thereby achieving the accuracy of position adjustment of target objects in the target area.

[0065] In an exemplary embodiment, Figure 2 As shown, extracting features from the image to be processed to obtain a target human feature parameter map includes steps S202 to S206. Among them:

[0066] Step S202, extracting initial human body features from the image to be processed to obtain initial feature maps.

[0067] Optionally, the server extracts initial human features, such as overall human features, or local human features, such as height, leg length, arm length, face height, etc., from the image to be processed to obtain initial feature maps.

[0068] Optionally, the server extracts overall human features from the image to be processed to obtain initial feature maps of different sizes.

[0069] Step S204, performing a feature fusion operation on each initial feature map to obtain a target feature map corresponding to each initial feature map.

[0070] Optionally, the server performs a feature fusion operation on each initial feature map to obtain a target feature map corresponding to each initial feature map. If each initial feature map has five sizes, a feature fusion operation is performed on initial feature maps of different sizes to obtain a target feature map corresponding to each initial feature map.

[0071] Step S206, detecting each target feature map to obtain a target human body and a corresponding target human body feature parameter map.

[0072] Optionally, the server detects each target feature map to obtain a target human body in the image to be processed; if there are multiple human bodies in the image to be processed, one is determined as the target human body, and a target human body feature parameter map corresponding to the target human body is obtained.

[0073] In this embodiment, the feature information emphasized in the initial feature map is different. By performing a feature fusion operation on the initial feature map, a target feature map emphasizing different feature information is obtained, thereby improving the accuracy of the target human body feature parameter map.

[0074] In an exemplary embodiment, Figure 3 As shown, the initial feature maps include initial feature maps of different sizes; a feature fusion operation is performed on each initial feature map to obtain a target feature map corresponding to each initial feature map, including steps S302 to S306. Wherein:

[0075] Step S302, obtaining the first sequence.

[0076] The first order is based on the order in which the initial feature maps are arranged from small to large in size.

[0077] Optionally, if there are five initial feature maps corresponding to five different sizes, the order in which the initial feature maps are arranged from small to large in size is used as the first order; or the order in which the initial feature maps are arranged from large to small in size is used as the first order.

[0078] Step S304: extract features from the first initial feature map to obtain a first intermediate feature map corresponding to the first initial feature map.

[0079] The first initial feature map is the initial feature map with the smallest or largest size.

[0080] Optionally, the server selects a feature map with the smallest size from the five initial feature maps as the first initial feature map; the initial feature maps are numbered 1, 2, 3, 4, and 5; the initial feature map numbered 1 is used as the first initial feature map, and the initial feature map Figure 1 Perform feature extraction to obtain initial features Figure 1 The corresponding first intermediate feature map is denoted as 21.

[0081] Step S306: perform feature fusion on the first intermediate feature map and the next initial feature map of the first initial feature map according to the first order to obtain a first target feature map corresponding to the next initial feature map.

[0082] Optionally, the server converts the first intermediate feature map 21 and the first initial feature map (initial feature map) into Figure 1 ) of the next initial feature map (initial feature Figure 2 ) to perform feature fusion and obtain the next initial feature Figure 2 The corresponding first target feature map (22). Similarly, the server Figure 2 The first target feature map (22) is subjected to feature fusion to obtain the next initial feature map (initial feature Figure 3 ) corresponding to the first target feature map (23); until the feature fusion of the last initial feature map is completed.

[0083] In this embodiment, feature extraction is performed on the initial feature map to obtain a first intermediate feature map, and feature fusion is performed on the first intermediate feature map and the next initial feature map of the first initial feature map according to a first order to obtain a first target feature map corresponding to the next initial feature map, thereby realizing fusion between feature maps of different sizes. The feature extraction of different network layers has different emphases, and correspondingly obtains feature maps of different sizes, and fusion between feature maps of different sizes.

[0084] In the above embodiment, Figure 4 As shown, a feature fusion operation is performed on each initial feature map to obtain a target feature map corresponding to each initial feature map, including steps S402 to S406. Among them:

[0085] Step S402, obtaining a second order, wherein the second order is an order based on arranging the first target feature graphs in descending order of size.

[0086] The first order is the opposite order of the second order. If the first order is from small to large, then the second order is from large to small. If the first order is from large to small, then the second order is from small to large.

[0087] Optionally, the initial feature map includes 1, 2, 3, 4, and 5; the features extracted from the initial feature map are the first intermediate feature map (21), and the first target feature map also includes (22)(23)(24)(25).

[0088] Step S404: taking the first target feature map with the largest size as a new first initial feature map, performing feature extraction on the new first initial feature map to obtain a new first intermediate feature map corresponding to the new initial feature map.

[0089] Among them, the largest first target feature map is used as the first target feature map.

[0090] The server uses the largest first target feature map (25) as a new first initial feature map, performs feature extraction on the new first initial feature map (1), and obtains a new first intermediate feature map (35) corresponding to the new initial feature map (1).

[0091] Step S406, performing feature fusion on the new first intermediate feature map and the previous initial feature map of the new first initial feature map in the second order to obtain a second target feature map corresponding to the previous initial feature map.

[0092] Optionally, the new first intermediate feature map (35) and the previous initial feature map (24) of the new first initial feature map (25) are subjected to feature fusion in the second order to obtain a second target feature map (34) corresponding to the previous initial feature map (24). Similarly, the server performs feature fusion on the second target feature map (34) and the previous initial feature map (23) of the new first initial feature map (24) to obtain a second target feature map (33) corresponding to the previous initial feature map (23), until the feature fusion of the last initial feature map (21) is completed.

[0093] In this embodiment, by taking the first target feature map with the largest size as the new first initial feature map, feature extraction is performed on the new first initial feature map to obtain a new first intermediate feature map corresponding to the new initial feature map, and feature fusion is performed on the new first intermediate feature map and the previous initial feature map of the new first initial feature map in the second order to obtain the second target feature map corresponding to the previous initial feature map, so as to achieve fusion between feature maps of different sizes. The focus of feature extraction in different network layers is different, and feature maps of different sizes are obtained accordingly, and fusion between feature maps of different sizes is achieved.

[0094] In an exemplary embodiment, Figure 5 As shown, before inputting the human body feature parameters into the trained position recognition model, the process includes steps S502 to S508. Among them:

[0095] Step S502, obtaining a historical human body characteristic parameter map and a corresponding position tag; the historical human body characteristic parameter map includes a human body local characteristic parameter map.

[0096] Optionally, the server obtains historical human feature parameter maps and corresponding location tags; the historical human feature parameter maps include human local feature parameter maps, such as height feature parameter maps, leg length feature parameter maps, arm length feature parameter maps, facial feature parameter maps, and the like.

[0097] Step S504: input the human body local feature parameter map into the initial position recognition model, and obtain the initial predicted position label output by the initial position recognition model.

[0098] Optionally, the server inputs each human body local feature parameter map into an initial position recognition model, and obtains an initial predicted position label output by the initial position recognition model.

[0099] Step S506, calculating the total loss value of the initial predicted position label and the position label using a preset total loss function.

[0100] Among them, the total loss function value represents the difference between the initial predicted position label and the position label, as well as the confidence of the local feature parameter map of the human body.

[0101] Optionally, the server calculates the total loss value for the initial predicted position label and the position label through a preset total loss function. The total loss function is recorded as L = αL1 + βL2. The first loss function L1 represents the difference between the initial predicted position label and the position label, and the second loss function L2 represents the confidence of the local feature parameter map of the human body. α is the weight of L1 corresponding to the first loss function; β is the weight of L2 corresponding to the second loss function.

[0102] Step S508, adjusting the adjustable parameters in the initial position recognition model according to the total loss value until the total loss value meets the convergence condition, and taking the initial position recognition model when the total loss value meets the convergence condition as the target initial position recognition model.

[0103] Optionally, after obtaining the first total loss value, if the total loss value does not meet the convergence conditions, the server adjusts the adjustable parameters in the initial position recognition model according to the total loss value. Afterwards, the server inputs another sample image into the adjusted initial position recognition model, obtains the second initial predicted position label output by the adjusted initial position recognition model, and calculates the second total loss value based on the second initial predicted position label. If the second total loss value does not meet the convergence conditions, the adjustable parameters in the adjusted initial position recognition model are adjusted again according to the second total loss value. And so on, until the calculated total loss value meets the convergence conditions, and the initial position recognition model whose total loss value meets the convergence conditions is used as the target initial position recognition model. The embodiments of the present disclosure do not limit the adjustable parameters and convergence conditions, and can be set according to actual conditions.

[0104] In this embodiment, by considering not only the difference between the initial predicted position label and the position label, but also the confidence of the human body local feature parameter map, during the training process, the trained position recognition model is made more accurate.

[0105] In an exemplary embodiment, Figure 6 As shown, the loss value is calculated for the initial predicted position label and the position label by using a preset total loss function, including steps S602 to S606. Wherein:

[0106] Step S502, calculating a first loss value for the initial predicted position label and the position label using a preset first loss function.

[0107] The first loss value represents the difference between the initial predicted position label and the position label.

[0108] Optionally, the server calculates the mean square error between the initial predicted position label and the position label by formula (1) to obtain the first loss value:

[0109]

[0110] In the formula, N represents the number of samples, K is the local feature parameter map of the human body in each sample,

[0111] is the predicted coordinates of the jth human body local feature parameter map in the i-th sample, p ij is the corresponding real location label, that is, the real coordinate.

[0112] Step S504: Calculate a first loss value for the initial predicted position label and the position label using a preset second loss function.

[0113] The second loss value represents the confidence of the local feature parameter map of the human body.

[0114] Optionally, the server determines whether the human body local feature parameter map exists by formula (2), and if the human body local feature parameter map exists, it is recorded as 1, and if it does not exist, it is recorded as 0.

[0115]

[0116] In the formula, y ij is the true confidence of the jth local feature parameter map of the human body in the i-th sample, is the confidence level of the prediction.

[0117] Step S506: sum the first loss value and the second loss value to obtain a total loss value.

[0118] Optionally, the server obtains the total loss value by weights corresponding to the first loss value and the second loss value. α is the weight of the first loss value; β is the weight of the second loss value L2. The total loss function is recorded as L = αL1 + βL2

[0119] In this embodiment, by considering not only the difference between the initial predicted position label and the position label but also the confidence of the local feature parameter map of the human body, the accuracy of the position recognition model can be improved.

[0120] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0121] Based on the same inventive concept, the embodiment of the present application also provides a human body characteristic parameter-based adjustment method device for implementing the human body characteristic parameter-based adjustment method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more human body characteristic parameter-based adjustment method device embodiments provided below can refer to the limitations of the human body characteristic parameter-based adjustment method in the above text, and will not be repeated here.

[0122] In an exemplary embodiment, Figure 7 As shown, a method and device for adjusting based on human body characteristic parameters is provided, comprising: an acquisition module 801, an extraction module 802, an input module 803 and an adjustment module 804, wherein:

[0123] An acquisition module 801 is used to acquire an image to be processed;

[0124] An extraction module 802 is used to extract human body feature parameters from the image to be processed to obtain a target human body feature parameter map;

[0125] Input module 803, used to input the target human body feature parameter map into the trained position recognition model, and obtain the position recognition result output by the position recognition model; the position recognition model is generated after training based on the historical human body feature parameter map as a sample set;

[0126] The adjustment module 804 is used to adjust the position of the corresponding target object in the target area according to the position recognition result.

[0127] In an exemplary embodiment, the extraction module 802 includes:

[0128] The extraction unit is used to extract initial human body features from the image to be processed to obtain initial feature maps.

[0129] A feature fusion unit is used to perform a feature fusion operation on each initial feature map to obtain a target feature map corresponding to each initial feature map;

[0130] The detection unit is used to detect each target feature map to obtain a target human body and a corresponding target human body feature parameter map.

[0131] In an exemplary embodiment, the initial feature map includes initial feature maps of different sizes; the feature fusion unit includes:

[0132] The first acquisition subunit is used to acquire a first order; wherein the first order is an order in which the initial feature maps are arranged in ascending order of size.

[0133] The first extraction subunit is used to perform feature extraction on the first initial feature map to obtain a first intermediate feature map corresponding to the first initial feature map; wherein the first initial feature map is a feature map with the smallest or largest size.

[0134] The first fusion subunit is used to perform feature fusion on the first intermediate feature map and the next initial feature map of the first initial feature map according to the first order to obtain a first target feature map corresponding to the next initial feature map.

[0135] In an exemplary embodiment, the feature fusion unit includes:

[0136] The second acquisition subunit is used to acquire a second order, wherein the second order is an order in which the first target feature graphs are arranged in descending order of size.

[0137] The second extraction subunit is used to use the first target feature map with the largest size as a new first initial feature map, perform feature extraction on the new first initial feature map, and obtain a new first intermediate feature map corresponding to the new initial feature map.

[0138] The second fusion subunit is used to perform feature fusion on the new first intermediate feature map and the previous initial feature map of the new first initial feature map in a second order to obtain a second target feature map corresponding to the previous initial feature map.

[0139] In an exemplary embodiment, a method and apparatus for adjusting based on human body characteristic parameters further includes:

[0140] A training adjustment module is used to obtain a historical human feature parameter map and a corresponding position label; the historical human feature parameter map includes a human local feature parameter map; the human local feature parameter map is input into an initial position recognition model, and an initial predicted position label output by the initial position recognition model is obtained; a total loss value is calculated for the initial predicted position label and the position label through a pre-set loss function; wherein the total loss function value represents the difference between the initial predicted position label and the position label, and the confidence of the human local feature parameter map; the adjustable parameters in the initial position recognition model are adjusted according to the total loss value until the total loss value meets the convergence condition, and the initial position recognition model when the total loss value meets the convergence condition is used as the target initial position recognition model.

[0141] In an exemplary embodiment, the training adjustment module is also used to calculate a first loss value for the initial predicted position label and the position label through a preset first loss function; the first loss value represents the difference between the initial predicted position label and the position label; the first loss value is calculated for the initial predicted position label and the position label through a preset second loss function; the second loss value represents the confidence of the local feature parameter map of the human body; the first loss value and the second loss value are summed to obtain a total loss value.

[0142] Each module in the above-mentioned adjustment device based on human body characteristic parameters can be implemented in whole or in part by software, hardware and a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0143] In an exemplary embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. Figure 8As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store human body characteristic parameter data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a regulation method based on human body characteristic parameters is implemented.

[0144] Those skilled in the art will understand that Figure 8 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0145] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0146] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0147] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0148] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0149] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0150] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be construed as limiting the scope of the present application. It should be noted that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for adjusting based on human body characteristic parameters, characterized in that: The method comprises: Get the image to be processed; Extract features from the image to be processed to obtain a target human body feature parameter map; Input the target human body feature parameter map into a trained position recognition model, and obtain the position recognition result output by the position recognition model; the position recognition model is generated after training based on the historical human body feature parameter map as a sample set; The position of the corresponding target object in the target area is adjusted according to the position recognition result.

2. The method according to claim 1, characterized in that The step of extracting features from the image to be processed to obtain a target human body feature parameter map includes: Extracting initial human body features from the image to be processed to obtain initial feature maps; Performing a feature fusion operation on each of the initial feature maps to obtain a target feature map corresponding to each of the initial feature maps; Each of the target feature maps is detected to obtain a target human body and a corresponding target human body feature parameter map.

3. The method according to claim 2, characterized in that The initial feature map includes initial feature maps of different sizes; The performing a feature fusion operation on each of the initial feature maps to obtain a target feature map corresponding to each of the initial feature maps includes: Acquire a first order; wherein the first order is based on the order in which the initial feature maps are arranged from small to large in size; Performing feature extraction on the first initial feature map to obtain a first intermediate feature map corresponding to the first initial feature map; wherein the first initial feature map is a feature map with the smallest or largest size; The first intermediate feature map and the next initial feature map of the first initial feature map are subjected to feature fusion according to the first order to obtain a first target feature map corresponding to the next initial feature map.

4. The method according to claim 3, characterized in that The performing a feature fusion operation on each of the initial feature maps to obtain a target feature map corresponding to each of the initial feature maps includes: Acquire a second order, wherein the second order is an order in which the first target feature graphs are arranged from large to small in terms of size; Taking the first target feature map with the largest size as a new first initial feature map, performing feature extraction on the new first initial feature map, and obtaining a new first intermediate feature map corresponding to the new initial feature map; The new first intermediate feature map and the previous initial feature map of the new first initial feature map are subjected to feature fusion in the second order to obtain a second target feature map corresponding to the previous initial feature map.

5. The method according to claim 1, characterized in that Before inputting the human body feature parameters into the trained position recognition model, the method includes: Acquire a historical human body characteristic parameter map and a corresponding position tag; the historical human body characteristic parameter map includes a human body local characteristic parameter map; Inputting the human body local feature parameter map into an initial position recognition model, and obtaining an initial predicted position label output by the initial position recognition model; Calculating a total loss value for the initial predicted position label and the position label by using a preset total loss function; wherein the total loss function value represents the difference between the initial predicted position label and the position label, and the confidence of the local feature parameter map of the human body; The adjustable parameters in the initial position recognition model are adjusted according to the total loss value until the total loss value meets the convergence condition, and the initial position recognition model when the total loss value meets the convergence condition is used as the position recognition model.

6. The method according to claim 5, characterized in that Calculating a total loss value for the initial predicted position label and the position label using a preset total loss function includes: Calculating a first loss value for the initial predicted position label and the position label by using a preset first loss function; the first loss value represents the difference between the initial predicted position label and the position label; Calculating a first loss value for the initial predicted position label and the position label by using a preset second loss function; the second loss value represents the confidence of the human body local feature parameter map; The first loss value and the second loss value are summed to obtain the total loss value.

7. A regulating device based on human body characteristic parameters, characterized in that: The device comprises: An acquisition module, used for acquiring an image to be processed; An extraction module is used to extract human body feature parameters from the image to be processed to obtain a target human body feature parameter map; An input module, used to input the target human body feature parameter map into a trained position recognition model, and obtain the position recognition result output by the position recognition model; the position recognition model is generated after training based on the historical human body feature parameter map as a sample set; The adjustment module is used to adjust the position of the corresponding target object in the target area according to the position recognition result.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images