Construction site safety wearing detection method and system based on improved YOLOv5
By improving the YOLOv5 target detection network, introducing a cross-level path aggregation network and an improved nonmaximum suppression algorithm, the problem of low recall rate in construction site safety wear detection is solved, achieving efficient detection of small targets and improved prediction accuracy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SOUTH CHINA UNIV OF TECH
- Filing Date
- 2024-05-31
- Publication Date
- 2026-07-10
AI Technical Summary
Existing target detection networks have low recall rates in detecting safety wearables on construction sites, making it difficult to effectively detect small targets. Furthermore, multi-class nonmaximum suppression algorithms cannot effectively handle bounding boxes with inclusion relationships and small intersection-merge relationships.
The YOLOv5 object detection network is improved by introducing a cross-level path aggregation network and an improved non-maximum suppression algorithm. The cross-level connection module accelerates feature fusion and filters and removes prediction results that have inclusion relationships and mutually exclusive categories for the same target.
It improves the recall and prediction accuracy of the target detection network, reduces the false negative rate, and enhances the detection capability for small targets.
Smart Images
Figure CN118629069B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of image processing technology, and in particular to a method, system, terminal device, and computer-readable storage medium for detecting safety wearables on construction sites based on an improved YOLOv5. Background Technology
[0002] The voluntary wearing of safety equipment by construction workers is the primary prerequisite for ensuring production safety on construction sites. While manual supervision through video surveillance systems is necessary for the timely detection and elimination of safety hazards, the required manpower is prohibitively high. In recent years, significant advancements in computer vision technology, particularly in object detection, have made its practical application possible. Object detection has evolved through two main stages: one based on manual feature extraction, and the other based on neural networks. Neural network-based object detection technology is becoming increasingly mature.
[0003] Safety wear detection on construction sites requires detecting targets such as the head, torso, hands, and shoes of construction workers, which often include numerous small targets in real-world scenarios. Current mainstream research on target detection networks focuses on improving precision, with less attention paid to recall, which is related to the false negative rate. Improving the recall of target detection networks can reduce the false negative rate, especially by detecting more small targets.
[0004] For the same target, if multiple overlapping prediction boxes are detected or multiple mutually exclusive prediction results are obtained, the multi-class nonmaximum suppression algorithm can only solve the two cases of high crossover-union ratio (CUI) where the boxes overlap but do not contain each other or almost completely overlap. However, for two prediction boxes that have an inclusion relationship and a small CUI, it is difficult to filter out or correct the prediction results using multi-class nonmaximum suppression. Summary of the Invention
[0005] To address at least one of the problems in the prior art, the present invention provides a construction site safety wearable detection method, system, terminal device, and computer-readable storage medium based on an improved YOLOv5.
[0006] The first objective of this invention is to provide a construction site safety wearable detection method based on an improved YOLOv5.
[0007] The second objective of this invention is to provide a construction site safety wearable detection system based on an improved YOLOv5.
[0008] The third objective of this invention is to provide a terminal device.
[0009] A fourth objective of this invention is to provide a computer-readable storage medium.
[0010] The first objective of this invention can be achieved by adopting the following technical solution:
[0011] A construction site safety wearable detection method based on improved YOLOv5, the method comprising:
[0012] Obtain a training set; the samples in the training set are images of construction site clothing and their labeled categories;
[0013] The improved YOLOv5 object detection network is trained using the training set to obtain a construction site safety wearable detection model. The path aggregation network used in the neck network of the YOLOv5 object detection network is a three-layer double-pyramid structure network, consisting of a top-down pyramid structure and a bottom-up pyramid structure. Cross-level connection modules are added between the bottom-most input and the middle-layer output of the path aggregation network, and between the middle-layer input and the top-most output, to obtain a cross-level path aggregation network. Replacing the path aggregation network used in the neck network of the YOLOv5 object detection network with the cross-level path aggregation network yields the improved YOLOv5 object detection network. The cross-level connection module includes a C3 module and a convolutional module connected sequentially.
[0014] The construction site safety wear detection model is deployed in the on-site safety wear detection system to achieve real-time detection and early warning of safety wear.
[0015] Furthermore, in the improved YOLOv5 object detection network, CLPAN is repeatedly stacked according to the network depth requirements.
[0016] Furthermore, the improved YOLOv5 object detection network also includes improvements to the Non-Maximum Suppression (NMS) algorithm, including:
[0017] Calculate the inclusion degree and crossover ratio among all predicted boxes;
[0018] Treat any two prediction boxes as a prediction box pair;
[0019] Perform the following operation on all predicted bounding box pairs, and retain the predicted bounding boxes as the final prediction output:
[0020] If the coverage between two predicted boxes is greater than the first set threshold and they belong to the same category, then: if the intersection-union ratio between the two predicted boxes is less than or equal to the second set threshold, then remove the predicted box with the smaller area and retain the predicted box with the larger area; otherwise, remove the predicted box with the lower confidence and retain the predicted box with the higher confidence.
[0021] If the coverage between two predicted boxes is greater than the first set threshold and they do not belong to the same category, or if the coverage between two predicted boxes is less than or equal to the first set threshold, then: if the crossover ratio between two predicted boxes is greater than the second set threshold, then remove the predicted box with the lower confidence and retain the predicted box with the higher confidence.
[0022] Furthermore, the containment C between the two prediction boxes i and j ij for:
[0023]
[0024] In the formula, S i and S j S represents the area of the predicted boxes i and j, respectively. i∩j To predict the area of the intersection of boxes i and j.
[0025] Furthermore, the Intersection over Union (IoU) between the two predicted boxes i and j ij for:
[0026]
[0027] In the formula, S i∩j and S i∪j Let i and j be the areas of the intersection and union of the predicted boxes i and j, respectively.
[0028] Furthermore, the categories are divided into five main categories based on head, torso, hands, feet, and crossing movements. Each main category is further divided into two subcategories: those that comply with safety regulations and those that do not.
[0029] Furthermore, subclasses within the same major category have a mutually exclusive relationship with the same target.
[0030] The second objective of this invention can be achieved by adopting the following technical solution:
[0031] A construction site safety wearable detection system based on an improved YOLOv5, the system comprising:
[0032] The acquisition module is used to acquire the training set; the samples in the training set are images of construction site clothing and their labeled categories.
[0033] The training module is used to train the improved YOLOv5 object detection network using the training set to obtain a construction site safety wearable detection model. The path aggregation network used in the neck network of the YOLOv5 object detection network is a three-layer double-pyramid structure network, consisting of a top-down pyramid structure and a bottom-up pyramid structure. Cross-level connection modules are added between the bottom-most input and the middle-layer output of the path aggregation network, and between the middle-layer input and the top-most output, to obtain a cross-level path aggregation network. Replacing the path aggregation network used in the neck network of the YOLOv5 object detection network with the cross-level path aggregation network yields the improved YOLOv5 object detection network. The cross-level connection module includes a C3 module and a convolutional module connected sequentially.
[0034] The detection module is used to deploy the construction site safety wear detection model in the on-site safety wear detection system to achieve real-time detection and early warning of safety wear.
[0035] The third objective of this invention can be achieved by adopting the following technical solution:
[0036] A terminal device includes a processor and a memory for storing processor-executable programs. When the processor executes the program stored in the memory, it implements the above-described construction site safety wearable detection method based on the improved YOLOv5.
[0037] The fourth objective of this invention can be achieved by adopting the following technical solution:
[0038] A computer-readable storage medium storing a program that, when executed by a processor, implements the above-described construction site safety wearable detection method based on improved YOLOv5.
[0039] The present invention has the following advantages over the prior art:
[0040] 1. This invention improves the neck path aggregation network of the YOLOv5 target detection network by introducing a cross-layer connection module to accelerate feature fusion between layers, thereby improving the recall rate of the target detection network.
[0041] 2. This invention improves the nonmaximum suppression algorithm by filtering and removing prediction results that have an inclusion relationship with the same target and are mutually exclusive in class, thereby expanding the applicability of the algorithm and improving the accuracy of network prediction. Attached Figure Description
[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.
[0043] Figure 1 This is a flowchart of the construction site safety wearable detection method based on the improved YOLOv5 according to Embodiment 1 of the present invention;
[0044] Figure 2 This is a schematic diagram of the structure of the YOLOv5 target detection network in Embodiment 1 of the present invention;
[0045] Figure 3 This is a simplified diagram comparing the neck network structure before and after the improvement in Embodiment 1 of the present invention;
[0046] Figure 4This is a structural diagram of the improved YOLOv5 network in Embodiment 1 of the present invention;
[0047] Figure 5 This is a schematic diagram of the overlap and inclusion relationship of the target boxes in Embodiment 1 of the present invention;
[0048] Figure 6 The flowchart of the improved nonmaximum suppression algorithm of Embodiment 1 of the present invention is shown below;
[0049] Figure 7 This is a structural block diagram of the construction site safety wearable detection system based on the improved YOLOv5 according to Embodiment 2 of the present invention;
[0050] Figure 8 This is a structural block diagram of the terminal device according to Embodiment 3 of the present invention. Detailed Implementation
[0051] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. It should be understood that the specific embodiments described are merely used to explain this application and are not intended to limit this application.
[0052] Example 1:
[0053] like Figure 1 As shown, the construction site safety wearable detection method based on the improved YOLOv5 provided in this embodiment includes the following steps:
[0054] S101. Construct the dataset.
[0055] By acquiring publicly available images online and simulating and photographing images of safety attire on construction sites, a dataset was constructed to train a network for detecting safety attire on construction sites.
[0056] S102, labeled dataset.
[0057] The images in the dataset are labeled and categorized into five main categories based on head, torso, hands, feet, and crossing motions. Each main category is further divided into two subcategories: those that comply with safety standards and those that do not. Subcategories within the same main category are mutually exclusive for the same target.
[0058] S103. Improve the YOLOv5 network.
[0059] This embodiment improves upon the YOLOv5 object detection network, where the YOLOv5 network structure is as follows: Figure 2 As shown, it is divided into three main parts: backbone, neck and head. The neck adopts the path aggregation network PAN, and its main components include the standard convolution module Conv, the three-convolution cross-stage partial connection bottleneck module C3, the upsampling module and the channel splicing module.
[0060] Specifically, the path aggregation network PAN in the YOLOv5 object detection network is a three-layer (P3, P4, P5) double pyramid structure network, including a top-down and a bottom-up pyramid structure for feature fusion.
[0061] To facilitate rapid ascent of lower-level information, this embodiment introduces a cross-level connection module between P3 input and P4 output, and between P4 input and P5 output. This cross-level connection module includes a C3 module and a Conv module. Simplified structural diagrams of the neck network before and after the improvement are shown below. Figure 3 .
[0062] In this embodiment, the modified path aggregation network is referred to as the cross-level path aggregation network CLPAN.
[0063] Understandably, CLPANs can be repeatedly stacked according to network depth requirements, and the CLPAN stacked n times is denoted as CLPAN Rn. The improved YOLOv5 network structure, using CLPAN Rn to replace the neck PAN, can be found in [link to YOLOv5 network structure]. Figure 4 .
[0064] In this embodiment, CLPAN R2 is used as the neck of YOLOv5.
[0065] S104. Improvements to the Non-Maximum Suppression (NMS) algorithm.
[0066] The overlap relationship between the prediction boxes is as follows: Figure 5 Multi-class NMS can only handle two types of overlap relationships: overlap without inclusion and almost complete overlap. Therefore, the NMS algorithm is improved. The improved NMS algorithm flow is as follows: Figure 6 As shown, the specific steps are as follows:
[0067] (1) Calculate the coverage and crossover ratio among all prediction boxes.
[0068] Iterate through the containment degree c and intersection-union ratio (IoU) between all predicted boxes.
[0069] The inclusion degree between prediction boxes i and j is and intersection are compared as Where S i and S j S represents the area of the bounding boxes i and j, respectively. i∩j and S ivjLet i and j be the areas of the intersection and union of the target boxes i and j, respectively.
[0070] (2) Select two prediction boxes i and j. If the coverage of prediction boxes i and j is c ij Greater than the threshold c th If the condition is met, proceed to (3); otherwise, proceed to (6).
[0071] This embodiment includes a degree threshold c. th =0.8.
[0072] (3) If the predicted categories of prediction boxes i and j belong to the same major category, continue to (4); otherwise, jump to (6).
[0073] (4) If the intersection-union ratio (IoU) of predicted boxes i and j is... ij IoU less than or equal to the threshold th If the condition is met, proceed to (5); otherwise, proceed to (7).
[0074] In this embodiment, the cross-union ratio threshold IoU th =0.5.
[0075] (5) Remove the smaller prediction box from prediction boxes i and j, keep the larger prediction box, and jump to (9).
[0076] (6) If the intersection-union ratio (IoU) of predicted boxes i and j is... ij IoU greater than the threshold th If yes, continue with (7); otherwise, skip to (8).
[0077] (7) Remove the prediction boxes with lower confidence in prediction boxes i and j, keep the prediction boxes with higher confidence, and jump to (9).
[0078] (8) Do not remove the prediction boxes i and j; both prediction boxes are retained.
[0079] (9) Repeat (2) to (8) until all prediction box pairs (i,j) have been traversed. The remaining prediction boxes are the final prediction output.
[0080] S105. The improved YOLOv5 network is trained using the labeled dataset to obtain the construction site safety wear detection model.
[0081] In this embodiment, the training parameters are set as follows: 100 iterations, batch size 32, and input size 640×640 pixels.
[0082] S106. Deploy the construction site safety wear detection model into the on-site safety wear detection system to achieve real-time detection and early warning of safety wear.
[0083] The model obtained from S105 was used to deploy a safety wear detection system on the construction site to detect and provide early warnings for safety wear in real time.
[0084] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware, and the corresponding program can be stored in a computer-readable storage medium.
[0085] It should be noted that although the method operations of the above embodiments are described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. On the contrary, the order of execution of the described steps may be changed. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.
[0086] Example 2:
[0087] like Figure 7 As shown, this embodiment provides a construction site safety wearable detection system based on an improved YOLOv5. The system includes an acquisition module 701, a training module 702, and a detection module 703, wherein:
[0088] The acquisition module 701 is used to acquire a training set; the samples in the training set are images of construction site clothing and their labeled categories.
[0089] Training module 702 is used to train the improved YOLOv5 object detection network using the training set to obtain a construction site safety wearable detection model. The path aggregation network used in the neck network of the YOLOv5 object detection network is a three-layer double-pyramid structure network, consisting of a top-down pyramid structure and a bottom-up pyramid structure. Cross-level connection modules are added between the bottom-most input and the middle-layer output of the path aggregation network, and between the middle-layer input and the top-most output, to obtain a cross-level path aggregation network. Replacing the path aggregation network used in the neck network of the YOLOv5 object detection network with the cross-level path aggregation network yields the improved YOLOv5 object detection network. The cross-level connection module includes a C3 module and a convolutional module connected sequentially.
[0090] The detection module 703 is used to deploy the construction site safety wear detection model in the construction site safety wear detection system to realize real-time detection and early warning of safety wear.
[0091] The specific implementation of each module in this embodiment can be found in Embodiment 1 above, and will not be repeated here. It should be noted that the system provided in this embodiment is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure can be divided into different functional modules to complete all or part of the functions described above.
[0092] Example 3:
[0093] This embodiment provides a terminal device, which can be a computer, such as... Figure 8 As shown, the system is connected via a system bus 801 to a processor 802, a memory, an input device 803, a display 804, and a network interface 805. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium 806 and internal memory 807. The non-volatile storage medium 806 stores the operating system, computer programs, and a database. The internal memory 807 provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. When the processor 802 executes the computer programs stored in the memory, it implements the construction site safety wearable detection method based on the improved YOLOv5 described in Embodiment 1, as follows:
[0094] Obtain a training set; the samples in the training set are images of construction site clothing and their labeled categories;
[0095] The improved YOLOv5 object detection network is trained using the training set to obtain a construction site safety wearable detection model. The path aggregation network used in the neck network of the YOLOv5 object detection network is a three-layer double-pyramid structure network, consisting of a top-down pyramid structure and a bottom-up pyramid structure. Cross-level connection modules are added between the bottom-most input and the middle-layer output of the path aggregation network, and between the middle-layer input and the top-most output, to obtain a cross-level path aggregation network. Replacing the path aggregation network used in the neck network of the YOLOv5 object detection network with the cross-level path aggregation network yields the improved YOLOv5 object detection network. The cross-level connection module includes a C3 module and a convolutional module connected sequentially.
[0096] The construction site safety wear detection model is deployed in the on-site safety wear detection system to achieve real-time detection and early warning of safety wear.
[0097] Example 4:
[0098] This embodiment provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the construction site safety wearable detection method based on the improved YOLOv5 described in Embodiment 1 above, as follows:
[0099] Obtain a training set; the samples in the training set are images of construction site clothing and their labeled categories;
[0100] The improved YOLOv5 object detection network is trained using the training set to obtain a construction site safety wearable detection model. The path aggregation network used in the neck network of the YOLOv5 object detection network is a three-layer double-pyramid structure network, consisting of a top-down pyramid structure and a bottom-up pyramid structure. Cross-level connection modules are added between the bottom-most input and the middle-layer output of the path aggregation network, and between the middle-layer input and the top-most output, to obtain a cross-level path aggregation network. Replacing the path aggregation network used in the neck network of the YOLOv5 object detection network with the cross-level path aggregation network yields the improved YOLOv5 object detection network. The cross-level connection module includes a C3 module and a convolutional module connected sequentially.
[0101] The construction site safety wear detection model is deployed in the on-site safety wear detection system to achieve real-time detection and early warning of safety wear.
[0102] It should be noted that the computer-readable storage medium in this embodiment can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof.
[0103] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope disclosed in the present invention, based on the technical solution and inventive concept of the present invention, shall fall within the scope of protection of the present invention.
Claims
1. A method for detecting safety wearables on construction sites based on an improved YOLOv5, characterized in that, The method includes: Obtain a training set; the samples in the training set are images of construction site clothing and their labeled categories; The improved YOLOv5 object detection network is trained using the training set to obtain a construction site safety wearable detection model. The path aggregation network used in the neck network of the YOLOv5 object detection network is a three-layer double-pyramid structure network, consisting of a top-down pyramid structure and a bottom-up pyramid structure. Cross-level connection modules are added between the bottom-most input and the middle-layer output of the path aggregation network, and between the middle-layer input and the top-most output, to obtain a cross-level path aggregation network. Replacing the path aggregation network used in the neck network of the YOLOv5 object detection network with the cross-level path aggregation network yields the improved YOLOv5 object detection network. The cross-level connection module includes a C3 module and a convolutional module connected sequentially. The construction site safety wear detection model is deployed in the construction site safety wear detection system to achieve real-time detection and early warning of safety wear; The improved YOLOv5 object detection network also includes improvements to the Non-Maximum Suppression (NMS) algorithm, including: Calculate the inclusion degree and crossover ratio among all predicted boxes; Treat any two prediction boxes as a prediction box pair; Perform the following operation on all predicted bounding box pairs, and retain the predicted bounding boxes as the final prediction output: If the coverage between two predicted boxes is greater than the first set threshold and they belong to the same category, then: if the intersection-union ratio between the two predicted boxes is less than or equal to the second set threshold, then remove the predicted box with the smaller area and retain the predicted box with the larger area; otherwise, remove the predicted box with the lower confidence and retain the predicted box with the higher confidence. If the coverage between two predicted boxes is greater than a first set threshold and they do not belong to the same category, or if the coverage between two predicted boxes is less than or equal to the first set threshold, then: if the intersection-union ratio (IU) between two predicted boxes is greater than a second set threshold, then remove the predicted box with the lower confidence and retain the predicted box with the higher confidence; where the coverage C between two predicted boxes i and j is... ij for: ； In the formula, and Let i and j be the areas of the predicted bounding boxes, respectively. To predict the area of the intersection of boxes i and j.
2. The construction site safety wear detection method according to claim 1, characterized in that, In the improved YOLOv5 object detection network, CLPAN is repeatedly stacked according to the network depth requirements.
3. The construction site safety wear detection method according to claim 1, characterized in that, Intersection over Union (IoU) between two prediction boxes i and j ij for: ； In the formula, and Let i and j be the areas of the intersection and union of the predicted boxes i and j, respectively.
4. The construction site safety wear detection method according to any one of claims 1 to 3, characterized in that, The categories are divided into five main categories based on head, torso, hands, feet, and crossing movements. Each main category is further divided into two subcategories: those that comply with safety regulations and those that do not.
5. The construction site safety wear detection method according to claim 4, characterized in that, Subclasses within the same major category have a mutually exclusive relationship with the same target.
6. A construction site safety wearable detection system based on an improved YOLOv5, characterized in that, The system includes: The acquisition module is used to acquire the training set; the samples in the training set are images of construction site clothing and their labeled categories. The training module is used to train the improved YOLOv5 object detection network using the training set to obtain a construction site safety wearable detection model. The path aggregation network used in the neck network of the YOLOv5 object detection network is a three-layer double-pyramid structure network, consisting of a top-down pyramid structure and a bottom-up pyramid structure. Cross-level connection modules are added between the bottom-most input and the middle-layer output of the path aggregation network, and between the middle-layer input and the top-most output, to obtain a cross-level path aggregation network. Replacing the path aggregation network used in the neck network of the YOLOv5 object detection network with the cross-level path aggregation network yields the improved YOLOv5 object detection network. The cross-level connection module includes a C3 module and a convolutional module connected sequentially. The detection module is used to deploy the construction site safety wear detection model in the construction site safety wear detection system to realize real-time detection and early warning of safety wear; The improved YOLOv5 object detection network also includes improvements to the Non-Maximum Suppression (NMS) algorithm, including: Calculate the inclusion degree and crossover ratio among all predicted boxes; Treat any two prediction boxes as a prediction box pair; Perform the following operation on all predicted bounding box pairs, and retain the predicted bounding boxes as the final prediction output: If the coverage between two predicted boxes is greater than the first set threshold and they belong to the same category, then: if the intersection-union ratio between the two predicted boxes is less than or equal to the second set threshold, then remove the predicted box with the smaller area and retain the predicted box with the larger area; otherwise, remove the predicted box with the lower confidence and retain the predicted box with the higher confidence. If the coverage between two predicted boxes is greater than a first set threshold and they do not belong to the same category, or if the coverage between two predicted boxes is less than or equal to the first set threshold, then: if the intersection-union ratio (IU) between two predicted boxes is greater than a second set threshold, then remove the predicted box with the lower confidence and retain the predicted box with the higher confidence; where the coverage C between two predicted boxes i and j is... ij for: ； In the formula, and Let i and j be the areas of the predicted bounding boxes, respectively. To predict the area of the intersection of boxes i and j.
7. A terminal device, comprising a processor and a memory for storing a processor-executable program, characterized in that, When the processor executes the program stored in the memory, it implements the construction site safety wear detection method according to any one of claims 1 to 5.
8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the construction site safety wear detection method according to any one of claims 1 to 5.