A construction safety risk early warning method and system
The construction safety risk early warning method combining frame difference method and multi-pose verification template realizes real-time detection of dynamic objects in the construction environment and accurate identification of helmet wearing status. It solves the problem of insufficient accuracy of dynamic recognition and helmet wearing recognition in the existing technology, and improves the early warning efficiency and reliability of construction safety risks.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BEIJING JINGHONG YUNTAI TECH CO LTD
- Filing Date
- 2026-04-21
- Publication Date
- 2026-07-24
AI Technical Summary
Existing construction safety monitoring technologies lack accuracy in identifying dynamic objects in the construction environment, are easily affected by changes in light, equipment obstruction, and dust, and have weak targeting for helmet-wearing identification. They also fail to form a complete closed-loop logic of dynamic identification, human body locking, and helmet verification, resulting in insufficient correlation between early warning signals and actual safety hazards.
The frame difference method is used to identify dynamic pixels. Combined with multi-pose verification templates and helmet model comparison, dynamic contour center point tracking and contour feature matching are used to realize real-time detection of dynamic objects and accurate identification of helmet wearing status, forming a progressive risk screening.
It improves the accuracy and real-time nature of dynamic target positioning in the construction area, reduces the misjudgment rate of helmet wearing recognition, ensures the accuracy and timeliness of early warning signals, and enhances the efficiency of construction safety management.
Smart Images
Figure CN122454480A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of safe construction technology, specifically to a method and system for early warning of safety risks in engineering construction. Background Technology
[0002] In the field of engineering construction, safety risk early warning is a core link in ensuring the safety of construction workers and the smooth progress of projects. Among them, real-time monitoring of the wearing of safety helmets by construction workers is of paramount importance in risk prevention and control.
[0003] Currently, traditional construction safety monitoring relies heavily on manual inspections, which suffers from limited coverage, delayed response, and significant subjective judgment bias. This makes it particularly difficult to capture dynamic risks in a timely manner, especially in construction scenarios with dense personnel and dispersed work areas.
[0004] While some image recognition-based safety monitoring solutions have been attempted in the current technology, they generally suffer from the following shortcomings: On the one hand, the accuracy of dynamic object recognition is insufficient, and it is easily affected by changes in light, equipment obstruction, dust interference, etc. in the construction environment, leading to missed detection or misjudgment of dynamic targets; on the other hand, the specificity of human body and helmet recognition is weak. Most solutions use a single posture template or single-view verification, which is difficult to adapt to the diverse working postures of construction workers such as standing, walking, and bending over. Moreover, the helmet contour judgment lacks quantitative standards, which can easily lead to misjudgment of wearing status due to inaccurate contour feature extraction.
[0005] In addition, some solutions do not form a complete logical closed loop of "dynamic recognition - human body locking - helmet verification", and the correlation between early warning signals and actual safety hazards is insufficient, making it difficult to effectively support the rapid response and decision-making of management personnel. The efficiency and reliability of construction safety risk prevention and control need to be further improved. Summary of the Invention
[0006] To address the shortcomings of existing technologies, this invention provides a method and system for early warning of safety risks in engineering construction, which solves the problem of not forming a complete logical closed loop of "dynamic identification - human body locking - helmet verification".
[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for early warning of safety risks in engineering construction, comprising the following steps: Step 1: Conduct real-time monitoring of the construction area and confirm the frame difference between adjacent monitored frames to identify dynamic pixels. Then, based on the dynamic pixels, confirm whether there are dynamic objects. The specific method is as follows: Based on the monitoring videos generated in the construction area, identify the monitoring images associated with adjacent frames, and perform grayscale processing on each group of monitoring images to identify the grayscale images associated with the corresponding monitoring images. Based on the grayscale image associated with each set of monitoring images, the grayscale value associated with pixels at the same location is identified, and the grayscale value associated with the corresponding pixel is recorded as HD. i Then, mark the grayscale values associated with the corresponding pixels in the previous set of frames as HH. i If two sets of gray values belonging to the same pixel position satisfy: |HD i -HH i If |≥Y1, then the corresponding pixel in the grayscale image is recorded as a dynamic pixel, where Y1 is a preset threshold; otherwise, no marking is performed. Several groups of dynamic pixels in the grayscale image are marked, and adjacent dynamic pixels are connected to confirm a dynamic contour. Then the center point of this dynamic contour is identified. Record the location of the center point of the dynamic contour corresponding to the current grayscale image, then identify the dynamic contour that exists in the next frame of the current grayscale image, and simultaneously confirm the center point of the dynamic contour in the next frame. From the monitoring area, confirm whether the two sets of center points are the same location point. If so, continue monitoring. If not, simultaneously mark the internal area associated with the dynamic contour as a dynamic object, and track the specific position of the dynamic object in subsequent frames. Step 2: Identify several sets of dynamic contours generated by the dynamic object in the construction area, and verify these contours using a pre-set verification template. Determine if the dynamic object is the target human body. If so, lock the head region; otherwise, no processing is required. The specific method is as follows: The generated sets of dynamic contours are compared sequentially with multiple preset verification templates to confirm the center point of the dynamic contours. The center point of the dynamic contours is made to coincide with the center point of the verification templates. The dynamic contours are then scaled and rotated, with the center of rotation as the center point. The maximum overlap between the dynamic contours and the verification templates during the processing is confirmed. The maximum overlap is determined as follows: the dynamic contour is identified as being located in the contour area of the verification template, and the area feature M1 of the corresponding contour area is confirmed. Then, the area feature M2 of the entire verification template is confirmed. The maximum overlap is calculated as M1 ÷ M2, where the verification template is a preset template. The maximum overlap is then confirmed to meet the following condition: maximum overlap ≥ 95%. If it meets the condition, the corresponding verification template is recorded as the selected template, and the dynamic object is marked as the target human body. Otherwise, other verification templates are continuously confirmed until the selected template is confirmed, and the dynamic object is marked as the target human body. Based on the selected template and the associated target human body, the partial outline associated with the preset head in the selected template is marked as the head region. Step 3: Compare and verify the head region of the target human body with the set helmet model. From several sets of associated dynamic contours, identify two sets of head regions to be verified, and extract the helmet contours associated with the two sets of head regions to be verified. Based on the extraction process, confirm whether the target human body is correctly wearing a safety helmet. The specific method is as follows: The selected template associated with the target human body is determined, and then the verification template that is perpendicular to the selected template is directly locked. This verification template is recorded as the secondary template. The secondary template is compared and verified with several sets of dynamic contours so that the center point of the dynamic contour coincides with the center point of the verification template. The dynamic contour is scaled and rotated, and the maximum overlap between different dynamic contours and the secondary template is determined. The maximum value is selected from several confirmed maximum overlaps, and the dynamic contour associated with the maximum value is recorded as the selected contour. Based on the preset head of the secondary template, the head region is confirmed from the selected contour and this head region is recorded as the secondary head region. The head region associated with the selected template is recorded as the primary head region. The vertical relationship between the verification templates is a pre-preset state. The helmet outline associated with the main head region and the secondary head region is confirmed: both the main head region and the secondary head region are recorded as head regions. The outer outline of the head region is identified, and three sets of continuous outer outline points are grouped together. The outer tangent of the middle outline point of each set of outline points is confirmed. Then, the angle between the outline points on both sides of each set of outline points and the outer tangent is confirmed. If both sets of angles satisfy: angle ≤ 15°, then the outline points of this set are recorded as helmet outline points. Otherwise, no marking is made. Several sets of continuous helmet outline points are connected to confirm the helmet outline associated with the main head region and the secondary head region. The helmet outline of the main head region is recorded as the main helmet outline, and the helmet outline of the secondary head region is recorded as the secondary helmet outline. If no helmet outline is determined in either the main head region or the secondary head region, a warning signal is generated directly. Within the main helmet contour and the secondary helmet contour, identify the single set of contour points with the smallest angle to the external tangent. Record the middle contour point of the single set of contour points as the feature high point. Based on the marked feature high point, compare and verify the main helmet contour with the set helmet model: make the feature high point coincide with the vertex of the helmet model, and control the main helmet contour to rotate according to the feature high point. During the rotation process, record the overlap ratio between the main helmet contour and the outer contour of the helmet model: determine the overlapping segment between the main helmet contour and the outer contour of the helmet model, and then determine the line length ratio of the overlapping segment on the outer contour of the helmet model. The line length ratio is the overlap ratio. Record the process with the maximum overlap ratio from several sets of rotation processes, and record the overlap ratio recorded in the corresponding process as the determined ratio. Using the same determination method for the main helmet outline, the determination ratio associated with the secondary helmet outline is reconfirmed. The two sets of determination ratios are averaged to determine the average ratio. If the average ratio is ≥90%, no processing is required, indicating that the target person is correctly wearing a safety helmet; otherwise, a warning signal is generated directly.
[0008] Preferably, a construction safety risk early warning system includes: The dynamic object marking end monitors the construction area in real time, confirms the frame difference between adjacent monitored frames, identifies dynamic pixels, and then confirms the existence of dynamic objects based on the dynamic pixels. The target human body recognition end confirms several sets of dynamic contours generated by dynamic objects in the construction area, and performs verification processing on several sets of dynamic contours in combination with the preset verification template to identify whether the dynamic object is the target human body. If it is, the head area is locked; if not, no processing is required. The safety warning and identification terminal compares and verifies the head area of the target human body with the set helmet model. From several sets of associated dynamic contours, it identifies two sets of head areas to be verified and extracts the helmet contours associated with the two sets of head areas to be verified. Based on the extraction process, it confirms whether the target human body is wearing a safety helmet.
[0009] This invention provides a method and system for early warning of safety risks in engineering construction. Compared with the prior art, it has the following advantages: This invention combines frame difference method with dynamic pixel connection and center point tracking to achieve real-time detection and continuous tracking of dynamic objects in the construction area, avoids static background interference, ensures the accuracy and real-time performance of dynamic target positioning, and lays a reliable foundation for subsequent risk assessment. The system employs multi-pose verification templates for comparison, covering various human body states in construction scenarios such as standing and walking, thereby improving the comprehensiveness of target human body recognition. By verifying the head area from both primary and secondary fixed templates, and combining the tangent angle of the contour points with the proportion of overlap with the helmet model for quantitative judgment, the system significantly reduces the misjudgment rate of helmet wearing recognition and ensures recognition accuracy. Focusing on the core risks of safety helmet wearing by construction workers, the helmet verification process is triggered only on the target human body, avoiding interference from non-human dynamic objects. This achieves a progressive risk screening of "dynamic object - target human body - helmet wearing," with early warning signals accurately pointing to the safety hazards of not wearing a helmet. This helps managers respond quickly and improves the efficiency and pertinence of construction safety management. Attached Figure Description
[0010] Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2 This is a schematic diagram of the principle framework of the present invention. Detailed Implementation
[0011] 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, and 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.
[0012] First Embodiment Please see Figure 1 This application provides a method for early warning of safety risks in engineering construction, including the following steps: Step 1: Conduct real-time monitoring of the construction area and confirm the frame difference between adjacent frames to identify dynamic pixels. Then, based on the dynamic pixels, confirm whether there are dynamic objects. If they exist, mark them; otherwise, do not mark them. Specifically, within the corresponding monitoring area, the monitoring video consists of several sets of frames. Between each adjacent frame, there are pixels at the same position. When the image does not change, the pixel features associated with the pixels will not fluctuate significantly. When the corresponding pixels fluctuate significantly, it means that the corresponding pixels have changed, that is, there is an intervening object. In this case, it is necessary to mark the object, identify whether it is a dynamic object, and confirm it accordingly. Specifically, the method for confirming the presence of dynamic objects within the construction area is as follows: Based on the monitoring videos generated in the construction area, the monitoring images associated with adjacent frames are identified, and each set of monitoring images is processed into grayscale to identify the grayscale image associated with the corresponding monitoring image. Specifically, each pixel in the monitoring image is associated with a set of RGB values. By assigning different weights to the three sets of RGB values, a set of grayscale values can be effectively identified, which will result in = 0.299×R + 0.587×G + 0.114×B. Based on the grayscale image associated with each set of monitoring images, the grayscale value associated with pixels at the same location is identified, and the grayscale value associated with the corresponding pixel is recorded as HD. i Where i represents different pixels, and the grayscale values associated with the corresponding pixels in the previous set of frames of the current grayscale image are marked as HH. i If two sets of gray values belonging to the same pixel position satisfy: |HD i -HH iIf |≥Y1, then the corresponding pixel in the grayscale image is recorded as a dynamic pixel, where Y1 is a preset threshold, generally between 50 and 100, and its specific value is preset in advance by relevant personnel. Otherwise, no marking is made. Several groups of dynamic pixels in the grayscale image are marked, and adjacent dynamic pixels are connected to confirm a dynamic contour. Then, the center point of this dynamic contour is identified. Specifically, by combining the dynamic contour with a two-dimensional coordinate system, the two-dimensional coordinates associated with each group of contour points can be confirmed. The average of the confirmed groups of two-dimensional coordinates is then processed to confirm the average coordinates, which are the center points of the corresponding dynamic contours. The location of the center point of the dynamic contour corresponding to the current grayscale image is recorded (that is, recorded within the location points associated with the monitoring area, which is a region including different points, so the center point of the contour can be simultaneously marked within the monitoring area). Then, the dynamic contour existing in the next frame of the current grayscale image is identified, and the center point of the dynamic contour in the next frame is simultaneously confirmed. From the monitoring area, it is confirmed whether the two sets of center points are the same location point. If so, monitoring continues. If not, the internal region associated with the dynamic contour is simultaneously marked as a dynamic object, and the specific position of the dynamic object in subsequent frames is tracked. Specifically, during tracking, within each frame, the dynamic contour of the object intersects with the dynamic contour of the previous frame. Thus, the dynamic contour of the next frame is the specific location of the dynamic object. Similarly, the dynamic contour of the next set of frames also belongs to the specific location of the corresponding dynamic object. This process can be repeated to enable real-time tracking of dynamic objects. Step 2: Confirm several sets of dynamic contours generated by the dynamic object in the construction area, and verify the several sets of dynamic contours in combination with the preset verification template to identify whether the dynamic object is the target human body. If so, lock the head area; if not, no processing is required. Specifically, this mainly involves identifying whether the corresponding construction personnel are wearing the corresponding safety helmet in the construction area. If so, the helmet is used. The specific method for identifying whether a dynamic object is the target human body is as follows: Several sets of dynamic contours are sequentially compared with multiple pre-set verification templates to confirm the center point of the dynamic contour. The center point of the dynamic contour is then aligned with the center point of the verification template. The dynamic contour is then scaled and rotated, with the center of rotation as the center point. The maximum overlap between the dynamic contour and the verification template is confirmed during the processing. The maximum overlap is determined as follows: the contour region of the dynamic contour within the verification template is identified, and the area feature M1 of the corresponding contour region is confirmed. Then, the area feature M2 of the entire verification template is confirmed. The maximum overlap is calculated as M1 ÷ M2. The verification template is a pre-set template, determined in advance by relevant personnel based on experience. The maximum overlap is then confirmed to meet the following requirement: maximum overlap ≥ 95%. If it meets this requirement, the corresponding verification template is marked as the selected template, and the dynamic object is labeled... If a target human body is identified, the dynamic object is marked as such. Otherwise, other verification templates are continuously checked until a selected template is identified. If no template is selected, the dynamic object does not belong to the target human body and no processing is required. Specifically, the preset verification templates are human body templates in motion, including standing, walking, and other bending or different postures. Each different state is associated with a different human body template. These human body templates are preset by relevant personnel in advance. Feature verification is performed on various dynamic contours generated by the same group of dynamic objects in the entire monitoring area to identify whether the dynamic object associated with the corresponding dynamic contour is the target human body. This identification method is more comprehensive and can effectively identify the target human body, facilitating the subsequent head area confirmation process. Based on the selected template and the associated target human body, the partial outline associated with the preset head in the selected template is marked as the head region. Step 3: Compare and verify the head area of the target human body with the set helmet model. From the associated dynamic contours, identify two sets of head areas to be verified, and extract the helmet contours associated with the two sets of head areas to be verified. Based on the extraction process, confirm whether the target human body is wearing a safety helmet correctly. The specific method for confirmation is as follows: The process involves identifying the template associated with the target human body, then directly locking onto the verification template that is perpendicular to the selected template. This verification template is referred to as the secondary template. The perpendicular relationship between the verification templates is a pre-set state, such as the front view template and the side view template. The secondary template is compared and verified with several sets of dynamic contours to ensure that the center point of the dynamic contour coincides with the center point of the verification template. The dynamic contours are then scaled and rotated to determine the maximum overlap between different dynamic contours and the secondary template. The maximum value is selected from the confirmed sets of maximum overlap, and the dynamic contour associated with the maximum value is referred to as the selected contour. Based on the preset head of the secondary template, the head region is identified from the selected contour and referred to as the secondary head region. The head region associated with the selected template is referred to as the primary head region. The helmet outline associated with the main head region and the secondary head region is confirmed as follows: both the main head region and the secondary head region are recorded as head regions. The outer outline of the head region is identified, and three sets of continuous outer outline points are grouped together. The outer tangent of the middle outline point in each group is confirmed, and the angle between the outline points on both sides of each group and the outer tangent is confirmed. If both groups of angles satisfy: angle ≤ 15°, then the outline points in this group are recorded as helmet outline points. Otherwise, no marking is made. Several sets of continuous helmet outline points are connected to confirm the helmet outline associated with the main head region and the secondary head region. The helmet outline of the main head region is recorded as the main helmet outline, and the helmet outline of the secondary head region is recorded as the secondary helmet outline. If no helmet outline is determined in either the main head region or the secondary head region, a warning signal is directly generated, indicating that the corresponding target person is not wearing a safety helmet, and relevant management personnel are alerted to implement corresponding management measures. Within the main helmet contour and the secondary helmet contour, identify the single set of contour points with the smallest angle to the external tangent. Record the middle contour point of the single set of contour points as the feature high point. Based on the marked feature high point, compare and verify the main helmet contour with the set helmet model: make the feature high point coincide with the vertex of the helmet model, and control the main helmet contour to rotate according to the feature high point. During the rotation process, record the overlap ratio between the main helmet contour and the outer contour of the helmet model: determine the overlapping segment between the main helmet contour and the outer contour of the helmet model, and then determine the line length ratio of the overlapping segment on the outer contour of the helmet model. The line length ratio is the overlap ratio. Record the process with the maximum overlap ratio from several sets of rotation processes, and record the overlap ratio recorded in the corresponding process as the determined ratio. Using the same determination method for the main helmet outline, the determination of the determination of the secondary helmet outline is reconfirmed. The two sets of determination of the determination of the proportion are averaged to determine the average proportion. If the average proportion is ≥90%, no processing is required, which means that the target person is wearing a safety helmet. Otherwise, a warning signal is directly generated, which means that the corresponding target person is not wearing a safety helmet, and relevant management personnel are alerted to implement the corresponding management measures. Specifically, during the verification and checking process, the head area is first located within the target human body. Then, the helmet outline within the head area is selected and confirmed, and the selected helmet outline is verified to determine whether the helmet is properly fitted. This comprehensive verification and identification process provides a safety warning, ensuring the actual accuracy of the identification and confirmation process, improving the identification effect, and effectively controlling the comprehensiveness of the safety warning.
[0013] Second Embodiment Combination Figure 2 A construction safety risk early warning system, comprising: The dynamic object marking end monitors the construction area in real time, confirms the frame difference between adjacent monitored frames, identifies dynamic pixels, and then confirms the existence of dynamic objects based on the dynamic pixels. The target human body recognition terminal confirms several sets of dynamic contours generated by dynamic objects in the construction area, and performs verification processing on several sets of dynamic contours in combination with a preset verification template to identify whether the dynamic object is the target human body. If it is, the head area is locked; otherwise, no processing is required.
[0014] The safety warning and identification terminal compares and verifies the head area of the target human body with the set helmet model. From several sets of associated dynamic contours, it identifies two sets of head areas to be verified and extracts the helmet contours associated with the two sets of head areas to be verified. Based on the extraction process, it confirms whether the target human body is wearing a safety helmet.
[0015] Some of the data in the above formulas are numerical calculations with dimensions removed, and the contents not described in detail in this specification are all prior art known to those skilled in the art.
[0016] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.
Claims
1. A method for early warning of safety risks in engineering construction, characterized in that, Includes the following steps: Step 1: Monitor the construction area in real time, confirm the frame difference between adjacent frames, identify dynamic pixels, and then confirm the presence of dynamic objects based on the dynamic pixels. Step 2: Confirm the several sets of dynamic outlines generated by the dynamic object in the construction area, and verify the several sets of dynamic outlines in combination with the preset verification template to identify whether the dynamic object is the target human body. If it is, lock the head area; if not, no processing is required. Step 3: Compare and verify the head area of the target human body with the set helmet model. From the associated dynamic contours, identify two sets of head areas to be verified, and extract the helmet contours associated with the two sets of head areas to be verified. Based on the extraction process, confirm whether the target human body is wearing a safety helmet correctly.
2. The method for early warning of engineering construction safety risks according to claim 1, characterized in that, In step one, the specific method for identifying dynamic pixels is as follows: Based on the monitoring videos generated in the construction area, identify the monitoring images associated with adjacent frames, and perform grayscale processing on each group of monitoring images to identify the grayscale images associated with the corresponding monitoring images. Based on the grayscale image associated with each set of monitoring images, the grayscale value associated with pixels at the same location is identified, and the grayscale value associated with the corresponding pixel is recorded as HD. i Then, mark the grayscale values associated with the corresponding pixels in the previous set of frames as HH. i If two sets of gray values belonging to the same pixel position satisfy: |HD i -HH i If |≥Y1, then the corresponding pixel in the grayscale image is recorded as a dynamic pixel, where Y1 is a preset threshold; otherwise, no marking is performed.
3. The method for early warning of engineering construction safety risks according to claim 2, characterized in that, In step one, the specific method for confirming whether there are dynamic objects in the construction area is as follows: Several groups of dynamic pixels in the grayscale image are marked, and adjacent dynamic pixels are connected to confirm a dynamic contour. Then the center point of this dynamic contour is identified. Record the location of the center point of the dynamic contour corresponding to the current grayscale image, then identify the dynamic contour existing in the next frame of the current grayscale image, and simultaneously confirm the center point of the dynamic contour in the next frame. From the monitoring area, confirm whether the two sets of center points are the same location point. If so, continue monitoring. If not, simultaneously mark the internal area associated with the dynamic contour as a dynamic object, and track the specific position of the dynamic object in subsequent frames.
4. The method for early warning of engineering construction safety risks according to claim 1, characterized in that, In step two, the specific method for identifying whether a dynamic object is the target human body is as follows: The generated sets of dynamic contours are compared sequentially with multiple preset verification templates to confirm the center point of the dynamic contour. The center point of the dynamic contour is made to coincide with the center point of the verification template. The dynamic contour is then scaled and rotated, with the center of rotation as the center point. The maximum overlap between the dynamic contour and the verification template is confirmed during the processing. The maximum overlap is determined as follows: the dynamic contour is identified as being located in the contour area of the verification template, and the area feature M1 of the corresponding contour area is confirmed. Then, the area feature M2 of the entire verification template is confirmed. The maximum overlap is calculated as M1 ÷ M2, where the verification template is a preset template. The maximum overlap is then confirmed to meet the following condition: maximum overlap ≥ 95%. If it meets the condition, the corresponding verification template is marked as the selected template, and the dynamic object is marked as the target human body. Otherwise, other verification templates are continuously confirmed until the selected template is confirmed, and the dynamic object is marked as the target human body.
5. The method for early warning of engineering construction safety risks according to claim 4, characterized in that, In step two, the specific method for locking the head region is as follows: Based on the selected template and the associated target human body, the partial outline associated with the preset head in the selected template is marked as the head region.
6. The method for early warning of engineering construction safety risks according to claim 1, characterized in that, In step three, the specific method for identifying the two sets of head regions to be verified is as follows: The process involves identifying the selected template associated with the target human body, then directly locking onto the verification template that is perpendicular to the selected template. This verification template is designated as the secondary template. The secondary template is then compared and verified with several sets of dynamic contours to ensure that the center point of the dynamic contour coincides with the center point of the verification template. The dynamic contours are then scaled and rotated to determine the maximum overlap between different dynamic contours and the secondary template. The maximum value is selected from the confirmed maximum overlap, and the dynamic contour associated with the maximum value is designated as the selected contour. Based on the preset head of the secondary template, the head region is identified from the selected contour and designated as the secondary head region. The head region associated with the selected template is designated as the primary head region. The vertical relationship between the verification templates is a pre-preset state.
7. The method for early warning of engineering construction safety risks according to claim 6, characterized in that, In step three, the specific method for confirming whether the target person is correctly wearing a safety helmet is as follows: The helmet outline associated with the main head region and the secondary head region is confirmed: both the main head region and the secondary head region are recorded as head regions. The outer outline of the head region is identified, and three sets of continuous outer outline points are grouped together. The outer tangent of the middle outline point of each set of outline points is confirmed. Then, the angle between the outline points on both sides of each set of outline points and the outer tangent is confirmed. If both sets of angles satisfy: angle ≤ 15°, then the outline points of this set are recorded as helmet outline points. Otherwise, no marking is made. Several sets of continuous helmet outline points are connected to confirm the helmet outline associated with the main head region and the secondary head region. The helmet outline of the main head region is recorded as the main helmet outline, and the helmet outline of the secondary head region is recorded as the secondary helmet outline. If no helmet outline is determined in either the main head region or the secondary head region, a warning signal is generated directly. Within the main helmet contour and the secondary helmet contour, identify the single set of contour points with the smallest angle to the external tangent. Record the middle contour point of the single set of contour points as the feature high point. Based on the marked feature high point, compare and verify the main helmet contour with the set helmet model: make the feature high point coincide with the vertex of the helmet model, and control the main helmet contour to rotate according to the feature high point. During the rotation process, record the overlap ratio between the main helmet contour and the outer contour of the helmet model: determine the overlapping segment between the main helmet contour and the outer contour of the helmet model, and then determine the line length ratio of the overlapping segment on the outer contour of the helmet model. The line length ratio is the overlap ratio. Record the process with the maximum overlap ratio from several sets of rotation processes, and record the overlap ratio recorded in the corresponding process as the determined ratio. Using the same determination method for the main helmet outline, the determination ratio associated with the secondary helmet outline is reconfirmed. The two sets of determination ratios are averaged to determine the average ratio. If the average ratio is ≥90%, no processing is required, indicating that the target person is correctly wearing a safety helmet; otherwise, a warning signal is generated directly.
8. A construction safety risk early warning system, wherein the system operates according to any one of claims 1-7, characterized in that, include: The dynamic object marking end monitors the construction area in real time, confirms the frame difference between adjacent monitored frames, identifies dynamic pixels, and then confirms the existence of dynamic objects based on the dynamic pixels. The target human body recognition end confirms several sets of dynamic contours generated by dynamic objects in the construction area, and performs verification processing on several sets of dynamic contours in combination with the preset verification template to identify whether the dynamic object is the target human body. If it is, the head area is locked; if not, no processing is required. The safety warning and identification terminal compares and verifies the head area of the target human body with the set helmet model. From several sets of associated dynamic contours, it identifies two sets of head areas to be verified and extracts the helmet contours associated with the two sets of head areas to be verified. Based on the extraction process, it confirms whether the target human body is wearing a safety helmet.