Industrial field risk early warning method, device and readable storage medium

Abstract

This application discloses an industrial site risk early warning method, apparatus, and readable storage medium, comprising: acquiring a work site image and identifying at least one human body part region and / or at least one hazardous tool region in the work site image; calculating a spatial overlap metric between each human body part region and each hazardous tool region based on the relative positional relationship between the human body part region and the hazardous tool region, wherein the spatial overlap metric is defined as the ratio of the area of ​​the intersection region of the human body part region and the hazardous tool region to the area of ​​the human body part region; determining a risk quantification value from multiple spatial overlap metric values, and matching the risk quantification value with a preset safety threshold parameter set to generate early warning signals of different levels based on the matching result. The technical solution provided by this application can quickly identify safety risks at the work site.

Description

Technical Field

[0001] This application relates to the field of industrial safety technology, and in particular to an industrial site risk early warning method, device and readable storage medium. Background Technology

[0002] In industries such as power, petrochemicals, and heavy machinery, operators frequently need to use dangerous tools such as hydraulic wrenches, cutting machines, and electric saws during production and maintenance work, posing significant safety risks. For example, hydraulic wrenches are frequently used during wind turbine construction. The high torque output of the reaction arm of a hydraulic wrench means that if the operator's hand gets too close to the reaction arm, slippage can easily cause crush injuries to the operator's hand, leading to serious accidents.

[0003] To strengthen safety management, companies generally adopt the EHS (Environment-Health-Safety) management system. Under the EHS system, whenever such high-risk operations are carried out, operators are required to take photos of the site from multiple key angles, such as tool layout and personnel positioning, and upload them to the management system for internal safety audits.

[0004] However, the traditional review method is manual review, which not only consumes a lot of manpower and takes a long time, but also makes it easy for reviewers to become fatigued and make mistakes when facing a large number of similar photos for a long time. Ultimately, this results in the inability to identify and deal with safety risks at the work site in a timely manner.

[0005] Therefore, it is necessary to provide a new method and device for early warning of industrial site risks to address the above-mentioned shortcomings. Summary of the Invention

[0006] The purpose of this application is to provide an industrial site risk early warning method, device and readable storage medium that can quickly identify safety risks at the work site.

[0007] To achieve the above objectives, this application provides an industrial site risk early warning method, the method comprising: acquiring a work site image and identifying at least one human body part region and / or at least one hazardous tool region in the work site image; calculating a spatial overlap metric between each human body part region and each hazardous tool region based on the relative positional relationship between the human body part region and the hazardous tool region, wherein the spatial overlap metric is defined as the ratio of the area of ​​the intersection region of the human body part region and the hazardous tool region to the area of ​​the human body part region; determining a risk quantification value from multiple spatial overlap metric values, and matching the risk quantification value with a preset safety threshold parameter set to generate early warning signals of different levels based on the matching results.

[0008] To achieve the above objectives, this application also provides an industrial site risk early warning device, the device comprising: a region identification module for acquiring images of the work site and identifying at least one human body part region and / or at least one hazardous tool region in the work site images; an overlap measurement module for calculating a spatial overlap metric value between each human body part region and each hazardous tool region based on the relative positional relationship between the human body part region and the hazardous tool region, wherein the spatial overlap metric value is defined as the ratio of the area of ​​the intersection region of the human body part region and the hazardous tool region to the area of ​​the human body part region; and a risk early warning module for determining a risk quantification value from multiple spatial overlap metric values ​​and matching the risk quantification value with a preset safety threshold parameter set to generate early warning signals of different levels based on the matching results.

[0009] To achieve the above objectives, this application also provides an industrial site risk warning device, comprising: a memory for storing a computer program; and a processor for executing the computer program stored in the memory, so that the device performs the industrial site risk warning method as described above.

[0010] To achieve the above objectives, this application also provides a computer-readable storage medium having instructions stored thereon, which, when executed by a processor, cause the processor to implement the industrial site risk warning method as described above.

[0011] To achieve the above objectives, this application also provides a computer program product, which includes computer program code. When the computer program code is run on a computer, the computer implements the industrial site risk warning method as described above.

[0012] To achieve the above objectives, this application also provides a chip that includes a circuit for performing the industrial site risk warning method described above.

[0013] Therefore, the technical solution provided in this application automatically identifies and locates human body parts and hazardous tool areas in work site images by analyzing the images. Then, based on the relative positional relationship between the human body parts and hazardous tool areas, a spatial overlap metric is calculated to quantify the spatial proximity of the operator and the hazardous tool, triggering a tiered early warning signal accordingly. In this application's solution, the spatial overlap metric is generated based on the area ratio of the intersection region to the human body parts region, and the smaller human body parts region is used as the normalization benchmark. This accurately reflects the risk level of a small human body part region intruding into a large hazardous tool region, ultimately significantly improving the sensitivity of identifying local intrusion behavior. Compared to traditional manual review methods, the technical solution provided in this application can significantly reduce the labor costs of manual review, effectively improve the efficiency and real-time nature of safety management, and achieve rapid and accurate identification of safety risks at the work site. Attached Figure Description

[0014] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0015] Figure 1 This is a flowchart of an industrial site risk early warning method according to one embodiment of this application; Figure 2 This is a flowchart illustrating the industrial site risk early warning method according to one embodiment of this application; Figure 3 This is a schematic diagram of the functional modules of the industrial site risk early warning device in the embodiments of this application; Figure 4 This is a schematic diagram of the industrial site risk early warning device in the embodiments of this application. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be further described in detail below with reference to the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. In the description of this application, the terms "first," "second," "third," etc., are only used to distinguish similar objects and are not necessarily used to describe a specific order or sequence, nor should they be construed as indicating or implying relative importance. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances. Furthermore, in the description of this application, unless otherwise stated, when an element is described as "connected" to another element, it can be directly connected to the other element, or there can be one or more intermediate elements between them. "Multiple" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship.

[0017] In industries such as power, petrochemicals, and heavy machinery, operators frequently need to use large and dangerous tools such as hydraulic wrenches, cutting machines, and electric saws during production and maintenance. These operations are complex and pose a high risk of mechanical injury. For example, hydraulic wrenches, widely used in wind turbine construction and heavy equipment dismantling and assembly, have a large torque output from their reaction arm and exert a strong force during operation. If an operator's hands, head, or other body parts are too close to or accidentally enter the rotation area of ​​the reaction arm, slippage can easily cause crushing, collision injuries, or even serious accidents such as fractures and limb injuries.

[0018] To strengthen on-site safety management, companies generally implement EHS (Environment, Safety, and Health) management systems. Under EHS requirements, whenever high-risk operations are performed, operators must take photos of the work from multiple key angles, including tool placement and personnel positioning, and upload them to the management system for internal safety audits. However, traditional auditing methods are manual, which is not only costly in terms of manpower, inefficient, and time-consuming, but also prone to visual fatigue due to reviewers constantly seeing similar photos, leading to missed or incorrect assessments. Furthermore, the standards for manual audits are susceptible to fluctuations due to subjective factors, ultimately resulting in the inability to promptly identify and address safety risks at the work site. Therefore, in complex industrial operation scenarios, how to achieve automatic and accurate identification of human body parts and dangerous tools, and to quantitatively assess the spatial proximity of the two in order to quickly identify safety risks at the work site, has become an urgent issue to be addressed in this field.

[0019] The present application will now be described in more detail with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are for illustrative and explanatory purposes only and do not constitute a limitation on the embodiments of the present application. The embodiments described herein are only a part of the embodiments of the present application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this application.

[0020] It should be noted that the implementation subject of this application is an industrial site risk warning device, which can be an electronic device with corresponding hardware and software environment, such as a server, mobile phone, laptop, tablet computer, vehicle-mounted device, etc. For ease of understanding, this application uses a hydraulic wrench as a typical example for description, and its technical solution can also be extended to other types of hazardous operation tools such as cutting machines and electric saws.

[0021] Please refer to the following: Figure 1 and Figure 2 , Figure 1 This is a flowchart of an industrial site risk early warning method according to one embodiment of this application. Figure 2 This is a flowchart illustrating the industrial site risk early warning method in one embodiment of this application.

[0022] S101: Acquire images of the work site and identify at least one human body part area and / or at least one hazardous tool area in the work site images.

[0023] In this embodiment, work site images can be collected in various ways according to EHS management system specifications. For example, operators can actively take photos and upload them in real time using mobile terminals (such as smartphones or digital cameras) at key work nodes, or video frames can be automatically collected by monitoring cameras deployed in the work area at time intervals or event-triggered mechanisms. Alternatively, the operation process can be continuously recorded from a first-person perspective using wearable devices (such as smart safety helmets with integrated cameras). Furthermore, the collected work site images can be transmitted to edge computing nodes or cloud servers via wired or wireless networks for storage and preprocessing. In this way, the risk warning device can retrieve work site images from the aforementioned edge computing nodes or cloud servers, and then analyze the work site images to identify at least one human body part area (such as a hand, arm, or head) and / or at least one hazardous tool area (such as a hydraulic wrench body, reaction arm, or high-pressure hose connector).

[0024] It should be noted that the above recognition process can be achieved through computer vision technology, which includes, but is not limited to, object detection models (YOLO, SSD, Faster R-CNN, etc.), instance segmentation models, key point detection models, or combinations thereof. This computer vision technology is used to locate the spatial position and geometric shape of human body parts and / or hazardous tools in work site images, thereby providing basic data input for subsequent spatial relationship analysis and risk assessment. The human body part region includes at least one of the hand region, arm region, or head region, and the hazardous tool region includes at least one of the hydraulic wrench body region, reaction arm region, or high-pressure hose connector region.

[0025] In one feasible implementation, the risk warning device can analyze work site images using a target detection model to identify at least one human body part region and / or at least one hazardous tool region contained within the image. In practical applications, the target detection model can employ a fusion architecture based on a feature pyramid network. This feature pyramid network uses a residual network as its backbone network, and its construction is as follows: First, the backbone network outputs feature maps with different spatial resolutions at different network levels. Deeper feature maps have lower spatial resolution and richer semantic information, while shallower feature maps have higher spatial resolution and more detailed features. Then, upsampling processing is performed on the feature map output from the deepest layer of the backbone network. The upsampling result is then fused with the feature map from the adjacent shallower layer through lateral connections. This process is then passed down level by level according to the network hierarchy, sequentially completing multi-scale feature fusion, ultimately generating a feature pyramid. The higher-resolution layer feature maps in the feature pyramid are used for human body part region detection, while the lower-resolution layer feature maps are used for detecting the hazardous tool region.

[0026] For example, suppose a target detection model is constructed using a 34-layer ResNet34 as the backbone, a Feature Pyramid Network (FPN) as the core for multi-scale feature fusion, and a detection head as the core for final judgment and localization. In this fusion architecture, ResNet34 is responsible for performing layer-by-layer convolution and pooling operations on the input work site image, transforming the original pixel information into multi-level visual features that can be recognized by the machine. At the same time, the residual structure solves the gradient vanishing problem in deep neural network training, ensuring that the 34-layer network can stably extract effective features. The FPN connects ResNet34 and the detection head, performing upsampling, feature fusion, and lateral connection processing on the basic features output by ResNet34 at each level to generate a multi-scale fused feature map. Based on the multi-scale fused feature map output by the FPN, the detection head classifies candidate regions in the image, outputting clear category labels such as "hand," "arm," "hydraulic wrench," and "head." It also predicts the location of the identified target and outputs the coordinates of the corresponding rectangular bounding box, ultimately defining the human body parts and dangerous tool areas in the work site image.

[0027] Specifically, the residual structure formula for the ResNet network is as follows: y=F(x,{W i})+x, Where x is the input feature F(x,{W i}) represents the residual mapping, {W i} represents the network weight parameters.

[0028] Considering the differences in target size caused by shooting angle, the Feature Pyramid Network (FPN) can be used to capture image features at different scales. The construction formula of the FPN is as follows: P n =σ(Up(P n+1 )+W l ·C n ), Among them, P n+1 Let W be the output feature map of the (n+1)th layer feature pyramid, where Up(·) is the upsampling operation (usually nearest neighbor upsampling). l ·C n For 1×1 convolution pairs of bottom-up path features C n The channel adjustment, σ is the nonlinear activation function (usually ReLU), C n This is the feature map of the nth stage of a bottom-up path (such as ResNet).

[0029] To preserve spatial information, the detector head employs a fully convolutional design. The classification loss uses multi-class cross-entropy, and the multi-class cross-entropy loss function is as follows: , in, The total number of categories, For the sample Category The true label (one-hot encoding). To predict probabilities.

[0030] Considering that positioning accuracy involves the interrelationships between targets, the positioning loss function adopted is the Distance Intersection over Union (DIou) loss function. The DIou loss function additionally considers the distance between the center points of the predicted bounding box and the ground truth bounding box. The DIou loss formula is as follows: , in, The intersection-union ratio (IU) of the predicted bounding box and the ground truth bounding box. The squared Euclidean distance between the center points of the predicted bounding box and the ground truth bounding box is given. Let be the diagonal length of the smallest closed region containing both the predicted and ground truth boxes. ,in, , , , These are the boundary coordinates of the smallest rectangle containing the two boxes.

[0031] Therefore, the overall loss function is: To improve the recognition accuracy of object detection models, they can be trained. Training an object detection model can be achieved through the following methods: First, original images of the work site are collected and labeled to construct an initial training set containing human body parts and hazardous tool areas. Then, data augmentation is performed on the initial training set to generate an augmented training set, which includes at least target replication augmentation and occlusion simulation augmentation. Finally, the augmented training set is used to train the target detection model, where the localization loss function of the target detection model adopts the distance intersection-union loss function.

[0032] Specifically, firstly, images of real-world work scenarios are collected, and then the images are normalized. normalized = I / 255, ensuring all pixel values ​​are on the same order of magnitude to avoid excessively large values ​​affecting calculations. Next, the image is standardized, I... standardized = (I normalized-μ) / σ, where I = [R, G, B] is the original image pixel value (range 0-1), μ = [0.485, 0.456, 0.406] is the ImageNet channel mean, and σ = [0.229, 0.224, 0.225] is the ImageNet channel standard deviation. The images are then labeled to construct an initial training set containing key categories such as "hands" and "hydraulic wrench reaction arms," ​​forming the basic annotation information for human body parts and dangerous tool regions. After constructing the initial training set, data augmentation processing can be performed to generate an augmented training set. Data augmentation processing includes at least two methods: target copying augmentation and occlusion simulation augmentation. The target copying augmentation process involves selecting sample images from the initial training set, copying the labeled human body parts or dangerous tool regions from the sample images, and pasting them into a background region in the sample image that does not spatially overlap with the existing labeled regions. Simultaneously, the pasted region is scaled and rotated to improve the model's generalization ability to recognize diverse target positions and postures. The occlusion simulation enhancement process involves selecting sample images from the initial training set and adjusting the relative spatial position or scale ratio between at least one human body part and at least one dangerous tool region within the sample images to generate enhanced training samples with different overlap ratios. This enhances the model's ability to identify and locate targets under partially occluded conditions. After obtaining the enhanced training set, the target detection model can be trained using it. During training, the Distance Intersection over Union (IoU) loss function is used as the localization loss function. This function, by introducing a center point distance penalty term, can effectively improve the localization accuracy of small targets (such as hands) and alleviate the gradient vanishing problem caused by traditional IoU loss when the bounding boxes do not overlap.

[0033] After training the target detection model until it converges or reaches a set number of rounds, the risk warning device can use the target detection model to analyze the work site image and identify at least one human body part and / or at least one dangerous tool area in the work site image.

[0034] S102: Based on the relative positional relationship between the human body part region and the dangerous tool region, calculate the spatial overlap metric between each human body part region and each dangerous tool region, wherein the spatial overlap metric is defined as: the ratio of the area of ​​the intersection region of the human body part region and the dangerous tool region to the area of ​​the human body part region.

[0035] In this embodiment, after inputting the work site image into a pre-trained target detection model, the risk warning device can use the aforementioned target detection model to output bounding boxes of all human body parts (e.g., hands, hand_box) and dangerous tool parts (e.g., hydraulic wrenches, force_arm_box) in the work site image. Each bounding box can be represented by pixel coordinates; for example, each region is represented by an axis-aligned rectangle defined by a quadruple (x1, y1, x2, y2), where (x1, y1) is the coordinate of the upper left corner and (x2, y2) is the coordinate of the lower right corner. Based on the output bounding box coordinates, the spatial relative positional relationship between the human body parts and dangerous tool parts can be resolved. It should be noted that the aforementioned spatial relative positional relationship specifically refers to the spatial positional relationship of a single human body part and a single dangerous tool paired together.

[0036] It is important to note that if the risk warning device fails to identify both human body parts and hazardous tool areas from the work site image, it indicates that there is no safety risk to the operator, and therefore the risk warning device can consider the current work environment safe. For example, if the risk warning device only identifies a hydraulic wrench but not a human hand area, it will not trigger a warning signal; if it only identifies a human hand area but not a hydraulic wrench or other hazardous tools, it can output a "No hazardous tool detected" warning signal. Conversely, if the risk warning device simultaneously identifies both human body parts and hazardous tool areas from the work site image, it indicates that there is a safety risk to the operator in the current work environment, and therefore the risk warning device can initiate the risk warning process.

[0037] In the risk warning process, the risk warning device can calculate the spatial overlap metric value corresponding to the combination of each human body part area and the dangerous tool area based on the aforementioned spatial relative positional relationship. Then, it can assess the spatial proximity of the human body part area and the dangerous tool area based on this spatial overlap metric value. Considering the significant area difference between the predicted bounding boxes of the human body part area and the dangerous tool area, the commonly used intersection-union ratio (IUU) formula... The value range is limited to 0 to ratio, where ratio is the area ratio of the smaller prediction box to the larger prediction box. This range is not a continuous interval of 0 to 1, and the range fluctuates dynamically with ratio. This makes it impossible to effectively distinguish the actual spatial overlap between the two using a fixed threshold. To address these shortcomings, this application uses the area ratio of the intersection of the two prediction boxes to the area of ​​the human body part. The spatial overlap metric is defined as the ratio of the area of ​​the intersection of the human body part region and the dangerous tool region to the area of ​​the human body part region.

[0038] For example, for a combination of a human body part region A and a dangerous tool region B, assuming the area of ​​human body part region A is Area(A) and the area of ​​dangerous tool region B is Area(B), then the spatial overlap metric corresponding to this combination can be expressed as: Area(A∩B) / Area(A), where Area(A∩B) represents the area of ​​the intersection of region A and region B.

[0039] In one feasible implementation, when calculating the spatial overlap metric between each body part region and each hazardous tool region, the risk warning device can first traverse all identified hazardous tool regions, and then, for each traversed hazardous tool region, traverse all identified body part regions one by one. Through this two-layer traversal mechanism, the risk warning device can construct multiple one-to-one region combinations of "single body part region and single hazardous tool region," thereby achieving a complete match of all identified regions without omission.

[0040] After constructing all region combinations, the risk warning device can perform calculations for each combination separately. Specifically, for any region combination, the risk warning device can first calculate the area of ​​the intersection between the human body part region and the dangerous tool region in that region combination, then use this intersection area as the dividend and the area of ​​the human body part region in that region combination as the divisor to perform a division operation, ultimately generating a unique spatial overlap metric for each region combination.

[0041] S103: Determine a risk quantification value from multiple spatial overlap metrics, and match the risk quantification value with a preset set of safety threshold parameters to generate warning signals of different levels based on the matching results.

[0042] In this embodiment, after calculating the spatial overlap metric value corresponding to each area combination, the risk warning device can determine the risk quantification value from multiple spatial overlap metric values ​​using methods such as weighted summation, maximum value method, and threshold-screened extreme value method. For example, if the weighted summation method is used, different risk weight coefficients can be assigned to different body parts and different dangerous tools (e.g., the risk weight of the hand is higher than that of the arm, and the risk weight of the reaction arm of a hydraulic wrench is higher than that of the hydraulic wrench body). Then, each spatial overlap metric value is multiplied by the weight coefficient of the corresponding combination to obtain the weighted metric value. Finally, the risk quantification value is determined by summing or taking the weighted maximum value. If the maximum value method is used, the maximum value among all spatial overlap metric values ​​can be directly determined as the risk quantification value for this operation. If the threshold-screened extreme value method is used, a risk-free threshold can be set first. Spatial overlap metric values ​​below this threshold are considered risk-free. Then, values ​​below this threshold are removed from all spatial overlap metric values, and the maximum value of the remaining spatial overlap metric values ​​is taken as the risk quantification value (if the remainder is empty, the risk quantification value is 0).

[0043] After determining the risk quantification value from multiple spatially overlapping metrics, the risk warning device can match the risk quantification value with a preset set of safety threshold parameters and generate warning signals of different levels based on the matching results. Specifically, the risk warning device stores a pre-configured set of safety threshold parameters, which can be pre-calibrated into a tiered threshold range based on high-risk operation safety regulations, historical accident data, and the operating conditions of different hazardous tools. For example, the safety threshold parameter set can be divided into multiple threshold ranges such as no risk, low risk, medium risk, and high risk according to the degree of risk.

[0044] When matching the quantified risk value with the safety threshold parameter set, the risk warning device can substitute the quantified risk value into the safety threshold parameter set, determine the threshold range to which the quantified risk value belongs, and generate and output different levels of warning signals based on the risk level corresponding to that threshold range. Different levels of warning signals can be configured with differentiated output formats. For example, a "no risk" level can trigger the risk warning device's background log recording; a "low risk" level can trigger the risk warning device to output text prompts; a "medium risk" level can trigger a pop-up reminder on the risk warning device's interface; and a "high risk" level can trigger an audible and visual alarm and push an alarm message to the safety management personnel's terminal. Furthermore, for high-risk levels, the risk warning device can also link with on-site safety control devices to output intervention commands, achieving graded response and closed-loop management of risk warnings.

[0045] Please see Figure 3 This application also provides an industrial site risk early warning device, the device comprising: The area recognition module is used to acquire images of the work site and identify at least one human body part area and / or at least one hazardous tool area in the images of the work site. The overlap measurement module is used to calculate the spatial overlap metric between each human body part region and each dangerous tool region based on the relative positional relationship between the human body part region and the dangerous tool region. The spatial overlap metric is defined as the ratio of the area of ​​the intersection region of the human body part region and the dangerous tool region to the area of ​​the human body part region. The risk warning module is used to determine a risk quantification value from multiple spatial overlap measurement values, and match the risk quantification value with a preset set of safety threshold parameters to generate warning signals of different levels based on the matching results.

[0046] Please see Figure 4 This application also provides an industrial site risk early warning device, which includes a memory and a processor. The memory stores a computer program, and when the computer program is executed by the processor, the industrial site risk early warning method described above can be implemented. Specifically, at the hardware level, the industrial site risk early warning device may include a processor, an internal bus, and a memory. The memory may include main memory and non-volatile memory. The processor reads the corresponding computer program from the non-volatile memory into main memory and then runs it. Those skilled in the art will understand that... Figure 4 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned industrial site risk early warning device. For example, the aforementioned industrial site risk early warning device may also include components that are more... Figure 4 The components shown may include more or fewer components, such as other processing hardware like a GPU (Graphics Processing Unit) or external communication ports. Of course, this application does not exclude other implementation methods besides software implementations, such as logic devices or a combination of hardware and software.

[0047] In this embodiment, the processor may include a central processing unit (CPU) or a graphics processing unit (GPU), and may also include other microcontrollers, logic gates, integrated circuits, or appropriate combinations thereof with logic processing capabilities. The memory described in this embodiment can be a storage device for storing information. In digital systems, a device capable of storing binary data can be a memory; in integrated circuits, a circuit without physical form but with storage function can also be a memory, such as RAM or FIFO; in a system, a storage device with physical form can also be called a memory. In implementation, this memory can also be implemented using a cloud storage method; the specific implementation method is not limited in this specification.

[0048] It should be noted that the specific implementation method of the industrial site risk early warning device in this specification can be referred to the description of the method implementation method, and will not be repeated here.

[0049] This application also provides a computer-readable medium storing instructions that, when executed by a processor, enable the processor to implement the industrial site risk warning method described in the above embodiments.

[0050] This application also provides a computer program product, which includes computer program code. When the computer program code is run on a computer, the computer can implement the industrial site risk warning method in the above embodiments.

[0051] This application also provides a chip, the chip including a circuit, the circuit being used to execute the industrial site risk warning method in the above embodiments.

[0052] Therefore, the technical solution provided in this application automatically identifies and locates human body parts and hazardous tool areas in work site images by analyzing the images. Then, based on the relative positional relationship between the human body parts and hazardous tool areas, a spatial overlap metric is calculated to quantify the spatial proximity of the operator and the hazardous tool, triggering a tiered early warning signal accordingly. In this application's solution, the spatial overlap metric is generated based on the area ratio of the intersection region to the human body parts region, and the smaller human body parts region is used as the normalization benchmark. This accurately reflects the risk level of a small human body part region intruding into a large hazardous tool region, ultimately significantly improving the sensitivity of identifying local intrusion behavior. Compared to traditional manual review methods, the technical solution provided in this application can significantly reduce the labor costs of manual review, effectively improve the efficiency and real-time nature of safety management, and achieve rapid and accurate identification of safety risks at the work site.

[0053] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0054] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for early warning of industrial site risks, characterized in that, The method includes: Acquire images of the work site and identify at least one human body part and / or at least one hazardous tool area in the images of the work site; Based on the relative positional relationship between the human body parts region and the dangerous tool region, a spatial overlap metric is calculated between each human body parts region and each dangerous tool region. The spatial overlap metric is defined as the ratio of the area of ​​the intersection region of the human body parts region and the dangerous tool region to the area of ​​the human body parts region. A risk quantification value is determined from multiple spatial overlap metrics, and the risk quantification value is matched with a preset set of safety threshold parameters to generate warning signals of different levels based on the matching results.

2. The method according to claim 1, characterized in that, The method further includes: A target detection model based on a feature pyramid network is trained to identify at least one human body part region and at least one hazardous tool region in the work site image.

3. The method according to claim 2, characterized in that, The feature pyramid network is constructed in the following way: The residual network backbone outputs feature maps with different spatial resolutions at different network layers. The feature map output from the deepest layer of the backbone network is upsampled, and the upsampled result is fused with the feature map of the adjacent shallow layer through lateral connection. This process is passed down and fused level by level to generate a feature pyramid. In the feature pyramid, the higher resolution layer feature map is used to detect the human body part region, and the lower resolution layer feature map is used to detect the dangerous tool region.

4. The method according to claim 3, characterized in that, The training method for the object detection model includes: Collect and label original work site images to construct an initial training set containing the human body parts and the hazardous tools. Data augmentation is performed on the initial training set to generate an augmented training set, wherein the data augmentation includes at least target replication augmentation and occlusion simulation augmentation; The target detection model is trained using the enhanced training set, wherein the localization loss function of the target detection model adopts the distance intersection-union loss function.

5. The method according to claim 4, characterized in that, The target replication enhancement includes: selecting sample images from the initial training set, copying the labeled human body parts or dangerous tool areas in the sample images, and pasting them into the background area of ​​the sample images that does not spatially overlap with the labeled areas, while scaling and rotating the pasted areas. The occlusion simulation enhancement includes: selecting sample images from the initial training set, and adjusting the relative spatial position or scale ratio between at least one of the human body parts and at least one of the dangerous tool areas in the sample images to generate enhanced training samples with different overlap ratios.

6. The method according to claim 1, characterized in that, The calculation of spatial overlap metrics between each human body part region and each dangerous tool region based on the relative positional relationship between the human body part regions and the dangerous tool regions includes: Traverse all identified hazardous tool areas, and for each hazardous tool area, traverse all human body part areas to form multiple area combinations; For each of the region combinations, the area of ​​the intersection region between the human body part region and the dangerous tool region in the region combination is calculated, and the area of ​​the intersection region is divided by the area of ​​the human body part region to generate the spatial overlap metric.

7. The method according to claim 6, characterized in that, Determining the risk quantification value from the plurality of spatial overlap metrics includes: The maximum value among the various spatial overlap metrics is selected as the risk quantification value.

8. The method according to claim 1, characterized in that, The human body region includes at least one of the hand region, arm region, or head region; The hazardous tool area includes at least one of the following: the hydraulic wrench body area, the reaction arm area, or the high-pressure hose connector area.

9. An industrial site risk early warning device, characterized in that, The device includes: The area recognition module is used to acquire images of the work site and identify at least one human body part area and / or at least one hazardous tool area in the images of the work site. The overlap measurement module is used to calculate the spatial overlap metric between each human body part region and each dangerous tool region based on the relative positional relationship between the human body part region and the dangerous tool region. The spatial overlap metric is defined as the ratio of the area of ​​the intersection region of the human body part region and the dangerous tool region to the area of ​​the human body part region. The risk warning module is used to determine a risk quantification value from multiple spatial overlap measurement values, and match the risk quantification value with a preset set of safety threshold parameters to generate warning signals of different levels based on the matching results.

10. An industrial site risk early warning device, characterized in that, include: Memory, used to store computer programs; A processor for executing a computer program stored in the memory to cause the apparatus to perform the method as described in any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that, It stores instructions that, when executed by a processor, cause the processor to implement the method as described in any one of claims 1 to 8.

12. A computer program product, characterized in that, The computer program product includes computer program code that, when run on a computer, causes the computer to perform the method as described in any one of claims 1 to 8.

13. A chip, characterized in that, The chip includes circuitry for performing the method as described in any one of claims 1 to 8.

Images