Human detection method, system, control device and readable storage medium

Abstract

The present application relates to the technical field of target detection, and specifically provides a human body detection method, system, control device and readable storage medium, aiming to solve the problem of how to improve the accuracy of human body detection in a dense crowd scene. To this end, the present application can obtain the final human body detection result of the to-be-detected image according to the first human body detection result obtained by human body detection and the torso detection result obtained by torso detection. Since the intersection-over-union of the two first human body detection results is large, the intersection-over-union of the corresponding torso detection result is not necessarily large. Through the above configuration mode, the present application can effectively avoid the problem that one of the two human body frames is filtered out due to the too large intersection-over-union of the two human body frames, thereby causing the lack of human body frames, and can effectively improve the accuracy of human body detection and effectively improve the human body detection effect in a dense crowd scene.

Description

Technical Field

[0001] This invention relates to the field of target detection technology, specifically providing a human body detection method, system, control device, and readable storage medium. Background Technology

[0002] With the continuous development of artificial intelligence technology, image acquisition technology is being applied to various scenarios, such as security and finance. Human detection based on image acquisition technology, in particular, is of great significance in the security and financial fields. However, in densely populated areas, human detection is prone to problems such as inaccurate human bounding boxes or missing human bounding boxes. Experiments have revealed that the main reasons for these problems are as follows:

[0003] (1) The human detector gives higher confidence to the inaccurate human bounding boxes, resulting in only the inaccurate human bounding boxes being retained after the NMS (Non Maximum Suppression) operation.

[0004] (2) The IOU (Intersection over Union) of the two human bounding boxes is greater than the NMS threshold, which causes one of the human bounding boxes to be filtered out, resulting in the absence of human bounding boxes.

[0005] Accordingly, there is a need in this field for a new human detection solution to address the aforementioned problems. Summary of the Invention

[0006] To overcome the above-mentioned shortcomings, this invention is proposed to provide a solution, or at least a partial solution, to the problem of how to improve the accuracy of human detection in densely populated scenes.

[0007] In a first aspect, the present invention provides a human body detection method, the method comprising:

[0008] Perform human detection on the image to be detected to obtain the first human detection result of the image to be detected;

[0009] Perform torso detection on the image to be detected to obtain the torso detection result of the image to be detected;

[0010] Based on the first human body detection result and the torso detection result, the final human body detection result of the image to be detected is obtained.

[0011] In one technical solution of the above-mentioned human detection method, the step of "obtaining the final human detection result of the image to be detected based on the first human detection result and the torso detection result" includes:

[0012] The first human body detection result and the torso detection result are matched to obtain the matching result;

[0013] Based on the matching results, the final human detection result of the image to be detected is obtained.

[0014] In one technical solution of the above-mentioned human body detection method, both the first human body detection result and the torso detection result are multiple. The first human body detection result includes a first human body bounding box, and the torso detection result includes a torso bounding box and a second human body bounding box corresponding to the torso bounding box. The step of "matching the first human body detection result and the torso detection result to obtain a matching result" includes:

[0015] Obtain the intersection-union ratio (CUC) between each first human bounding box and each second human bounding box to obtain the CUC matrix;

[0016] The matching result is obtained based on the intersection-union ratio matrix.

[0017] In one technical solution of the above-mentioned human body detection method, the matching result includes a first matching result and a second matching result. The step of "obtaining the matching result according to the intersection-union matrix" includes:

[0018] Based on the intersection-union matrix, obtain the first human body frame that matches each second human body frame;

[0019] After replacing the second human body bounding box in each torso detection result with the first human body bounding box that matches the second human body bounding box, the corrected torso detection result is obtained, and the corrected torso detection result is used as the first matching result.

[0020] The first human body frame that does not match the second human body frame is taken as the second matching result.

[0021] In one technical solution of the above-mentioned human detection method, the step of "obtaining the final human detection result of the image to be detected based on the matching result" includes:

[0022] For the first matching result, non-maximum suppression calculation is performed based on the torso bounding box to obtain the retained torso bounding box; the first human body bounding box corresponding to the retained torso bounding box in the first matching result is taken as the first detection result;

[0023] For the second matching result, non-maximum suppression calculation is performed based on the first human bounding box to obtain the retained first human bounding box; the retained first human bounding box in the second matching result is used as the second detection result;

[0024] Based on the first detection result and the second detection result, obtain the final human detection result of the image to be detected; and / or,

[0025] The step of “obtaining the first human body frame matching each second human body frame according to the intersection-union matrix” includes:

[0026] The Hungarian algorithm is applied to obtain the first human body frame that matches each second human body frame based on the intersection-union matrix.

[0027] In one technical solution of the above-mentioned human body detection method, the trunk detection result includes the trunk classification confidence and trunk localization accuracy of the trunk bounding box. "Performing non-maximum suppression calculation based on the trunk bounding box" includes:

[0028] Based on the torso classification confidence and torso localization accuracy, the torso bounding box confidence is obtained;

[0029] Non-maximum suppression is calculated based on the torso bounding box and the torso bounding box confidence level; and / or,

[0030] The first human detection result includes the human classification confidence and human localization accuracy of the first human bounding box. The step of "performing nonmaximum suppression calculation based on the first human bounding box" includes:

[0031] Based on the human body classification confidence score and the human body positioning accuracy, obtain the human body bounding box confidence score of the first human body bounding box;

[0032] Non-maximum suppression is calculated based on the first human bounding box and the confidence level of the human bounding box.

[0033] In one technical solution of the above-mentioned human detection method, the step of "performing human detection on the image to be detected and obtaining a first human detection result of the image to be detected" includes:

[0034] Feature extraction is performed on the image to be detected to obtain the image features of the image to be detected;

[0035] Based on the image features, human detection is performed to obtain the first human detection result of the image to be detected;

[0036] The step of "performing torso detection on the image to be detected and obtaining the torso detection result of the image to be detected" includes:

[0037] Based on the image features, torso detection is performed to obtain the torso detection result of the image to be detected.

[0038] In a second aspect, the present invention provides a human body detection system, the system comprising:

[0039] The first human detection result acquisition module is configured to perform human detection on the image to be detected and obtain the first human detection result of the image to be detected.

[0040] A torso detection result acquisition module is configured to perform torso detection on the image to be detected and obtain the torso detection result of the image to be detected;

[0041] The final human detection result acquisition module is configured to acquire the final human detection result of the image to be detected based on the first human detection result and the torso detection result.

[0042] In a third aspect, a control device is provided, comprising a processor and a storage device, the storage device being adapted to store a plurality of program codes, the program codes being adapted to be loaded and run by the processor to perform the human detection method described in any of the above-described technical solutions of the human detection method.

[0043] In a fourth aspect, a computer-readable storage medium is provided, wherein a plurality of program codes are stored therein, the program codes being adapted to be loaded and run by a processor to perform the human detection method described in any of the above-described technical solutions of the human detection method.

[0044] The above-described technical solutions of the present invention have at least one or more of the following beneficial effects:

[0045] In implementing the technical solution of this invention, the invention can obtain the final human detection result of the image to be detected based on the first human detection result obtained from human detection and the torso detection result obtained from torso detection. Since the intersection-union ratio (IU / R) of the two first human detection results is large, the IU / R of the corresponding torso detection result may not necessarily be large. Through the above configuration, this invention can effectively avoid the problem of one human bounding box being filtered out due to an excessively large IU / R of the two human bounding boxes, thus resulting in missing human bounding boxes. It can also effectively improve the accuracy of human detection and significantly improve the human detection effect in densely populated scenes. Attached Figure Description

[0046] The disclosure of this invention will become more readily understood with reference to the accompanying drawings. It will be readily understood by those skilled in the art that these drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. Furthermore, similar numbers in the drawings are used to denote similar components, wherein:

[0047] Figure 1 This is a schematic flowchart of the main steps of a human body detection method according to an embodiment of the present invention;

[0048] Figure 2 This is an example schematic diagram of nonmaximum suppression calculation according to one embodiment of the present invention;

[0049] Figure 3This is an example schematic diagram of the torso frame and human body frame in the torso detection result according to an embodiment of the present invention.

[0050] Figure 4 This is a schematic flowchart of the main steps of a human body detection method according to one embodiment of the present invention.

[0051] Figure 5 yes Figure 4 A flowchart illustrating the main steps of step S205;

[0052] Figure 6 This is a main structural block diagram of a human body detection system according to an embodiment of the present invention.

[0053] List of reference numerals :

[0054] 1: Human body frame; 2: Torso frame. Detailed Implementation

[0055] Some embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0056] In the description of this invention, "module" and "processor" can include hardware, software, or a combination of both. A module can include hardware circuitry, various suitable sensors, communication ports, memory, and may also include software components, such as program code, or a combination of software and hardware. A processor can be a central processing unit, microprocessor, image processor, digital signal processor, or any other suitable processor. The processor has data and / or signal processing capabilities. The processor can be implemented in software, in hardware, or a combination of both. Non-transitory computer-readable storage media includes any suitable medium capable of storing program code, such as magnetic disks, hard disks, optical disks, flash memory, read-only memory, random access memory, etc. The term "A and / or B" means all possible combinations of A and B, such as only A, only B, or A and B. The terms "at least one A or B" or "at least one of A and B" have a similar meaning to "A and / or B" and can include only A, only B, or A and B. The singular terms "a" or "this" can also include plural forms.

[0057] See appendix Figure 1 , Figure 1 This is a schematic flowchart illustrating the main steps of a human body detection method according to an embodiment of the present invention. Figure 1 As shown, the human body detection method in this embodiment of the invention mainly includes the following steps S101-S103.

[0058] Step S101: Perform human detection on the image to be detected to obtain the first human detection result of the image to be detected.

[0059] In this embodiment, human detection can be performed on the image to be detected to obtain the first human detection result in the image to be detected.

[0060] In one implementation, the first human detection result may include the position (e.g., coordinates) of the first human bounding box, the confidence level of human classification, and the accuracy of human positioning.

[0061] In one implementation, common object detection algorithms such as R-CNN (Region Convolutional Neural Network) and YOLO series algorithms can be used for human detection.

[0062] Step S102: Perform torso detection on the image to be detected to obtain the torso detection result of the image to be detected.

[0063] In this embodiment, torso detection can be performed on the image to be detected to obtain the torso detection result in the image to be detected.

[0064] In one implementation, the torso detection result may include the position (e.g., coordinates) of the torso bounding box, the torso classification confidence, the torso localization accuracy, and the position (e.g., coordinates) of the second human bounding box corresponding to the torso bounding box.

[0065] In one implementation, feature extraction can be performed on the image to be detected first to obtain the image features of the image to be detected. Then, a target detection algorithm is applied to perform human body detection and torso detection based on the image features to obtain the first human body detection result and the torso detection result, respectively.

[0066] In one implementation, a convolutional neural network can be used to extract features from the image to be detected.

[0067] Step S103: Based on the first human body detection result and the torso detection result, obtain the final human body detection result of the image to be detected.

[0068] In this embodiment, the final human detection result of the image to be detected can be obtained by combining the first human detection result and the torso detection result. That is, the final human detection result takes into account both the first human detection result and the torso detection result.

[0069] Based on steps S101-S103 above, this embodiment of the invention can obtain the final human detection result of the image to be detected based on the first human detection result obtained from human detection and the torso detection result obtained from torso detection. Since the intersection-union ratio (IU) of the two first human detection results is large, the IU of the corresponding torso detection result may not necessarily be large. Through the above configuration, this embodiment of the invention can effectively avoid the problem of one human bounding box being filtered out due to an excessively large IU of the two human bounding boxes, thus resulting in missing human bounding boxes. It can also effectively improve the accuracy of human detection and effectively improve the human detection effect in dense crowd scenes.

[0070] The following is a further explanation of step S103.

[0071] In one embodiment of the present invention, step S103 may include steps S1031 and S1032:

[0072] Step S1031: Match the first human body detection result with the torso detection result to obtain the matching result.

[0073] In this embodiment, there can be multiple first human body detection results and multiple torso detection results. Step S1031 may include the following steps S10311 and S10312:

[0074] Step S10311: Obtain the cross-union ratio (CUP) between each first human body bounding box and each second human body bounding box, and obtain the CUP matrix.

[0075] In this embodiment, the Intersection over Union (IOU) ratio can be calculated between each first human body frame and each second human body frame to obtain the IOU matrix.

[0076] The method for calculating IOU is as follows: Assuming the area of ​​box A is S1, the area of ​​box B is S2, and the overlap area between box A and box B is S3, then the IOU between box A and box B can be obtained according to the following formula (1):

[0077] (1)

[0078] Assuming there are m first human body frames and n second human body frames, the resulting intersection-union matrix is ​​an m×n matrix.

[0079] Step S10312: Obtain the matching result based on the intersection-union matrix.

[0080] In this embodiment, the matching result may include a first matching result and a second matching result. Step S10312 may further include the following steps S103121 and S103123:

[0081] Step S103121: Based on the intersection-union matrix, obtain the first human body frame that matches each second human body frame.

[0082] In this embodiment, the first human body frame that matches each second human body frame can be determined based on the intersection-union matrix.

[0083] In one implementation, the Hungarian algorithm can be used to obtain the first human bounding box that matches each second human bounding box based on the intersection-union matrix. The Hungarian algorithm is an algorithm for finding the maximum matching in graph theory.

[0084] Step S103122: Replace the second human body bounding box in each torso detection result with the first human body bounding box that matches the second human body bounding box to obtain the corrected torso detection result, and use the corrected torso detection result as the first matching result.

[0085] In this embodiment, experiments show that the accuracy of the first human body detection result obtained from human body detection is higher than the accuracy of the second human body detection result obtained from torso detection. Therefore, after obtaining the first human body frame that matches the second human body frame, the second human body frames in the torso detection result can be replaced with the first human body frames that match the second human body frames, thus obtaining a corrected torso detection result. This part of the torso detection result can be used as the first matching result.

[0086] Step S103123: Take the first human body frame that does not match the second human body frame as the second matching result.

[0087] In this embodiment, the first human body frame that does not match the second human body frame can be taken as the second matching result.

[0088] Step S1032: Based on the matching results, obtain the final human detection result of the image to be detected.

[0089] In this embodiment, step S1032 may include steps S10321 and S10323:

[0090] Step S10321: For the first matching result, perform non-maximum suppression (NMS) calculation based on the torso bounding box to obtain the retained torso bounding box; take the first human body bounding box corresponding to the retained torso bounding box in the first matching result as the first detection result.

[0091] Non-maximum suppression (NMS) is a common post-processing algorithm in target detection. (See appendix for details.) Figure 2 , Figure 2 This is an example schematic diagram illustrating nonmaximum suppression calculation according to one embodiment of the present invention. Figure 2 As shown, in the target detection process, multiple detection boxes are usually output for a single target. Each detection box has a confidence score. Typically, the detection box with the highest confidence score is retained. Non-maximum suppression calculation accomplishes this function, and its specific process is as follows:

[0092] (1) Sort all detection boxes by confidence and select the detection box with the highest confidence.

[0093] (2) Iterate through all the remaining detection boxes. If there is an IOU greater than a certain threshold with the current detection box with the highest confidence, delete the remaining detection box and mark the detection box with the highest confidence as the retained detection box.

[0094] (3) Continue to select the detection box with the highest confidence from the remaining detection boxes, and repeat the above process (1) and (2) to finally obtain all the retained detection boxes.

[0095] In this embodiment, please refer to the appendix. Figure 3 , Figure 3 This is an example schematic diagram of the torso frame and human body frame in the torso detection results according to one embodiment of the present invention. Figure 3 As shown, when there is overlap between human figures in the image to be detected, the intersection-union ratio (IU) between the human bounding boxes (1) of two individuals will be relatively large, while the IU between the torso bounding boxes (2) of two individuals will be relatively small. Therefore, if human bounding boxes (1) are used directly for NMS calculation, one of the human bounding boxes may be filtered out, resulting in missing human bounding boxes. However, if torso bounding boxes (2) are used for NMS calculation, since the IU between the torso bounding boxes (2) of two individuals is small, it can effectively prevent the torso bounding box of one individual from being filtered out.

[0096] Specifically, for the first matching result, NMS calculation can be performed based on the torso bounding box to obtain the retained torso bounding box. Then, the first human bounding box corresponding to the retained torso bounding box is used as the first detection result. This can effectively avoid the situation where some human bounding boxes are missing when there is overlap in densely populated scenes.

[0097] In one implementation, step S10321 may include steps S103211 and S103212:

[0098] Step S103211: Obtain the torso bounding box confidence based on the torso classification confidence and torso localization accuracy.

[0099] Step S103212: Perform nonmaximum suppression calculation based on the torso bounding box and torso bounding box confidence.

[0100] In this embodiment, to improve the confidence of the torso detection results, a convolutional neural network for object detection can be trained to simultaneously output the torso bounding box position, torso classification confidence, and torso localization accuracy. This adds the prediction result of torso localization accuracy to the existing convolutional neural network model. The torso bounding box confidence can be obtained based on the torso classification confidence and torso localization accuracy, and NMS is calculated based on the torso bounding box and its confidence.

[0101] In one implementation, the result of multiplying the torso classification confidence score and the torso localization accuracy score can be used as the torso frame confidence score.

[0102] In one implementation, the torso classification confidence and torso localization accuracy can be weighted to obtain the torso bounding box confidence.

[0103] Step S10322: For the second matching result, perform non-maximum suppression calculation based on the first human bounding box to obtain the retained first human bounding box; use the retained first human bounding box in the second matching result as the second detection result.

[0104] In this embodiment, for the second matching result, that is, the first human body bounding box in the first human body detection result that does not match the second human body bounding box, NMS calculation can be directly performed on this part of the first human body bounding box to obtain the retained first human body bounding box. This part of the retained first human body bounding box is the second detection result.

[0105] In one implementation, step S10322 may include steps S103221 and S103222:

[0106] Step S103221: Obtain the confidence score of the first human bounding box based on the confidence score of human body classification and the accuracy of human body localization.

[0107] Step S103222: Perform nonmaximum suppression calculation based on the first human bounding box and the human bounding box confidence.

[0108] In this embodiment, to improve the confidence of the first human detection result, a convolutional neural network for object detection can be trained to simultaneously output the position of the first human bounding box, the human classification confidence score, and the human localization accuracy. This adds the prediction result of human localization accuracy to the existing convolutional neural network model. The human bounding box confidence score can be obtained based on the human classification confidence score and the human localization accuracy, and NMS calculation can be performed based on the first human bounding box and its confidence score.

[0109] In one implementation, the result of multiplying the human body classification confidence score and the human body localization accuracy score can be used as the human body bounding box confidence score.

[0110] In one implementation, the confidence score of human body classification and the accuracy of human body localization can be weighted to obtain the confidence score of human body bounding box.

[0111] Step S10323: Based on the first detection result and the second detection result, obtain the final human detection result of the image to be detected.

[0112] In this embodiment, the first detection result and the second detection result can be combined to obtain the final human detection result of the image to be detected.

[0113] In one implementation, see Appendix Figure 4 , Figure 4 This is a schematic flowchart illustrating the main steps of a human body detection method according to one embodiment of the present invention. Figure 4 As shown, the human body detection method may include the following steps S201 to S206:

[0114] Step S201: Input the image to be detected.

[0115] In this embodiment, the image to be detected can be input first.

[0116] Step S202: Feature extraction.

[0117] In this embodiment, image feature extraction can be performed on the image to be detected.

[0118] Step S203: Human body detection.

[0119] In this embodiment, the method described in step S203 is similar to the method described in step S101 above, and will not be repeated here for the sake of simplicity.

[0120] Step S204: Trunk detection.

[0121] In this embodiment, the method described in step S204 is similar to the method described in step S102 above, and will not be repeated here for the sake of simplicity.

[0122] Step S205: Post-processing.

[0123] In this embodiment, please refer to the appendix. Figure 5 , Figure 5 yes Figure 4 A flowchart illustrating the main steps of step S205 is shown below; Figure 5 As shown, step S205 may include steps S2051 to S2053:

[0124] Step S2051: Match the first human body detection result with the torso detection result to obtain the first matching result and the second matching result.

[0125] In this embodiment, the method described in step S2051 is similar to the method described in step S1031, and will not be described again here for the sake of simplicity.

[0126] Step S2052: Calculate non-maximum suppression for the first matching result and the second matching result respectively to obtain the first detection result and the second detection result respectively.

[0127] In this embodiment, the method described in step S2052 is similar to the methods described in steps S10321 and S10322, and will not be described again here for the sake of simplicity.

[0128] Step S2053: Obtain the final human body test result based on the first test result and the second test result.

[0129] In this embodiment, the method described in step S2053 is similar to the method described in step S10323, and will not be described again here for the sake of simplicity.

[0130] Step S206: Output the final human body detection results.

[0131] In this embodiment, the final human body detection results can be output.

[0132] It should be noted that although the steps in the above embodiments are described in a specific order, those skilled in the art will understand that in order to achieve the effects of the present invention, different steps do not necessarily have to be executed in such an order. They can be executed simultaneously (in parallel) or in other orders, and these variations are all within the scope of protection of the present invention.

[0133] Furthermore, the present invention also provides a human body detection system.

[0134] See appendix Figure 6 , Figure 6 This is a main structural block diagram of a human body detection system according to an embodiment of the present invention. Figure 6 As shown, the human detection system in this embodiment of the invention may include a first human detection result acquisition module, a torso detection result acquisition module, and a final human detection result acquisition module. In this embodiment, the first human detection result acquisition module can be configured to perform human detection on the image to be detected, obtaining a first human detection result for the image to be detected. The torso detection result acquisition module can be configured to perform torso detection on the image to be detected, obtaining a torso detection result for the image to be detected. The final human detection result acquisition module can be configured to obtain a final human detection result for the image to be detected based on the first human detection result and the torso detection result.

[0135] The aforementioned human detection system is used to perform Figure 1The human body detection method embodiments shown are similar in technical principle, the technical problems solved, and the technical effects produced. Those skilled in the art can clearly understand that, for the sake of convenience and brevity, the specific working process and related descriptions of the human body detection system can be referred to the content described in the embodiments of the human body detection method, which will not be repeated here.

[0136] Those skilled in the art will understand that all or part of the processes in the method of the above embodiment of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable storage medium can include any entity or device capable of carrying the computer program code, a medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory, a random access memory, an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0137] Furthermore, the present invention also provides a control device. In one embodiment of the control device according to the present invention, the control device includes a processor and a storage device. The storage device can be configured to store a program for executing the human body detection method of the above-described method embodiments, and the processor can be configured to execute the program in the storage device. The program includes, but is not limited to, the program for executing the human body detection method of the above-described method embodiments. For ease of explanation, only the parts related to the embodiments of the present invention are shown; for specific technical details not disclosed, please refer to the method section of the embodiments of the present invention. The control device can be a control device device comprising various electronic devices.

[0138] Furthermore, the present invention also provides a computer-readable storage medium. In one embodiment of the computer-readable storage medium according to the present invention, the computer-readable storage medium can be configured to store a program for performing the human detection method of the above-described method embodiments. This program can be loaded and run by a processor to implement the above-described human detection method. For ease of explanation, only the parts related to the embodiments of the present invention are shown; for specific technical details not disclosed, please refer to the method section of the embodiments of the present invention. The computer-readable storage medium can be a storage device comprising various electronic devices. Optionally, in the embodiments of the present invention, the computer-readable storage medium is a non-transitory computer-readable storage medium.

[0139] Furthermore, it should be understood that since the various modules are only provided to illustrate the functional units of the device of the present invention, the physical devices corresponding to these modules may be the processor itself, or a part of the processor's software, a part of its hardware, or a combination of software and hardware. Therefore, the number of modules shown in the figures is merely illustrative.

[0140] Those skilled in the art will understand that the various modules in the device can be adaptively split or combined. Such splitting or combining of specific modules will not cause the technical solution to deviate from the principles of the present invention; therefore, the technical solutions after splitting or combining will fall within the protection scope of the present invention.

[0141] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A method for human body detection, characterized in that, The method includes: Perform human detection on the image to be detected to obtain the first human detection result of the image to be detected; Perform torso detection on the image to be detected to obtain the torso detection result of the image to be detected; Based on the first human body detection result and the torso detection result, the final human body detection result of the image to be detected is obtained; Both the first human body detection result and the torso detection result are multiple. The first human body detection result includes a first human body bounding box, and the torso detection result includes a torso bounding box and a second human body bounding box corresponding to the torso bounding box. The step of "obtaining the final human body detection result of the image to be detected based on the first human body detection result and the torso detection result" includes: Obtain the intersection-union ratio (CUC) between each first human bounding box and each second human bounding box to obtain the CUC matrix; Based on the intersection-union matrix, obtain the first human body frame that matches each second human body frame; After replacing the second human body bounding box in each torso detection result with the first human body bounding box that matches the second human body bounding box, the corrected torso detection result is obtained, and the corrected torso detection result is used as the first matching result. The first human body bounding box that does not match the second human body bounding box is taken as the second matching result; For the first matching result, non-maximum suppression calculation is performed based on the torso bounding box to obtain the retained torso bounding box; the first human body bounding box corresponding to the retained torso bounding box in the first matching result is taken as the first detection result; For the second matching result, non-maximum suppression calculation is performed based on the first human bounding box to obtain the retained first human bounding box; the retained first human bounding box in the second matching result is used as the second detection result; Based on the first detection result and the second detection result, the final human detection result of the image to be detected is obtained.

2. The human body detection method according to claim 1, characterized in that, The step of obtaining the first human body frame matching each second human body frame based on the intersection-union matrix includes: The Hungarian algorithm is applied to obtain the first human body frame that matches each second human body frame based on the intersection-union matrix.

3. The human body detection method according to claim 1, characterized in that, The torso detection results include the torso classification confidence and torso localization accuracy of the torso bounding box. Non-maximum suppression calculation based on the torso bounding box includes: Based on the torso classification confidence and torso localization accuracy, the torso bounding box confidence is obtained; Non-maximum suppression is calculated based on the torso bounding box and the torso bounding box confidence level; and / or, The first human detection result includes the human classification confidence and human localization accuracy of the first human bounding box. The step of "performing non-maximum suppression calculation based on the first human bounding box" includes: Based on the human body classification confidence score and the human body positioning accuracy, obtain the human body bounding box confidence score of the first human body bounding box; Non-maximum suppression is calculated based on the first human bounding box and the confidence level of the human bounding box.

4. The human body detection method according to claim 1, characterized in that, The steps for performing human detection on an image to be detected and obtaining a first human detection result for the image to be detected include: Feature extraction is performed on the image to be detected to obtain the image features of the image to be detected; Based on the image features, human detection is performed to obtain the first human detection result of the image to be detected; The steps of performing torso detection on the image to be detected and obtaining the torso detection result of the image to be detected include: Based on the image features, torso detection is performed to obtain the torso detection result of the image to be detected.

5. A human body detection system, characterized in that, The system includes: The first human detection result acquisition module is configured to perform human detection on the image to be detected and obtain the first human detection result of the image to be detected. A torso detection result acquisition module is configured to perform torso detection on the image to be detected and obtain the torso detection result of the image to be detected; The final human detection result acquisition module is configured to acquire the final human detection result of the image to be detected based on the first human detection result and the torso detection result; Both the first human body detection result and the torso detection result are multiple. The first human body detection result includes a first human body bounding box, and the torso detection result includes a torso bounding box and a second human body bounding box corresponding to the torso bounding box. The final human body detection result acquisition module is further configured to: Obtain the intersection-union ratio (CUC) between each first human bounding box and each second human bounding box to obtain the CUC matrix; Based on the intersection-union matrix, obtain the first human body frame that matches each second human body frame; After replacing the second human body bounding box in each torso detection result with the first human body bounding box that matches the second human body bounding box, the corrected torso detection result is obtained, and the corrected torso detection result is used as the first matching result. The first human body bounding box that does not match the second human body bounding box is taken as the second matching result; For the first matching result, non-maximum suppression calculation is performed based on the torso bounding box to obtain the retained torso bounding box; the first human body bounding box corresponding to the retained torso bounding box in the first matching result is taken as the first detection result; For the second matching result, non-maximum suppression calculation is performed based on the first human bounding box to obtain the retained first human bounding box; the retained first human bounding box in the second matching result is used as the second detection result; Based on the first detection result and the second detection result, the final human detection result of the image to be detected is obtained.

6. A control device, comprising a processor and a storage device, said storage device being adapted to store a plurality of program codes, characterized in that, The program code is adapted to be loaded and run by the processor to perform the human body detection method according to any one of claims 1 to 4.

7. A computer-readable storage medium storing a plurality of program codes, characterized in that, The program code is adapted to be loaded and run by a processor to perform the human body detection method according to any one of claims 1 to 4.

Images