A helmet detection system based on image recognition

Abstract

The present invention belongs to the field of helmet fusion detection combining thermal images and camera images, and specifically discloses a helmet detection system based on image recognition, which includes: by extracting each visually identifiable area and each visually distorted area in the worker's helmet wearing image, on the one hand, determining the wearing quality defect rate of the visually identifiable area based on the wearing image of each visually identifiable area, and on the other hand, by collecting the thermal image of each visually identifiable area, the wearing quality defect rate of the visually distorted area is determined, thereby improving the overall utilization rate of image information and making the distorted image area that was originally regarded as "useless" valuable. By detecting the environmental parameters of the worker, a heat correction map is generated to effectively eliminate heat distribution interference. By obtaining the fitting distance between the helmet and the worker's head, the fitting degree of the worker's helmet is analyzed, which can reflect the protective association between the size of the helmet and the worker's head.

Description

Technical Field

[0001] The invention belongs to the field of helmet fusion detection combining thermal images with camera images, and relates to a helmet detection system based on image recognition. Background Art

[0002] With the rapid development of industrialization and modernization, safety issues in engineering construction, manufacturing and other fields have become increasingly prominent. In these fields, hard hats are the most basic personal protective equipment, and their wearing is directly related to the life safety of workers. However, traditional hard hat wearing monitoring methods often rely on the analysis of high-definition images, which makes it difficult to achieve efficient use of images. Therefore, it is necessary to optimize the existing hard hat image detection system.

[0003] Traditional visual monitoring may only focus on clear, undistorted image areas, while ignoring distorted areas caused by various reasons (such as light, shadow, occlusion, etc.). That is, when images are collected under poor visibility conditions and the worker's helmet-wearing normative indicators cannot be successfully fully identified, it is usually necessary to wait for the worker to move before collecting again, which makes the safety analysis less timely and limits the overall utilization of image information.

[0004] At the same time, traditional helmet detection systems may only focus on whether the helmet is worn, without in-depth analysis of the fit between the helmet and the worker's head. The fit is crucial to the protective effect, and the lack of this analysis may make it impossible to fully evaluate the protective performance of the helmet. Summary of the invention

[0005] In view of this, in order to solve the problems raised in the above background technology, a helmet detection system based on image recognition is proposed.

[0006] The purpose of the present invention can be achieved through the following technical solutions: The present invention provides a safety helmet detection system based on image recognition, the system comprising: a safety helmet wearing image detection module: used to collect workers' safety helmet wearing images, extract each visually identifiable area and each visually distorted area, obtain the workers' safety helmet wearing model, and match the standard thermal feature distribution map of the safety helmet wearing model.

[0007] Safety helmet wearing quality analysis module: used to extract the apparent identification parameters of each contour in each visually identifiable area and analyze the wearing quality defect rate of the visually identifiable area of ​​the safety helmet.

[0008] Safety helmet wearing position analysis module: used to integrate various visually identifiable areas into a closed outline of the safety helmet, which includes the worker's facial outline and the safety helmet covering outline, and evaluate the standardization of the safety helmet wearing position.

[0009] Thermal imaging acquisition and analysis module: used to use thermal imaging cameras to capture the heat distribution map of workers' heads, construct a thermal correction map of the visual distortion area of ​​the helmet, and analyze the wearing quality defect rate of the visual distortion area of ​​the helmet based on this.

[0010] Safety helmet protection quality assessment module: used to analyze the fit of workers' safety helmets, and then comprehensively analyze the safety helmet protection quality indicators.

[0011] In a specific embodiment, the extraction of each visually recognizable area and each visually distorted area is specifically as follows: extracting a recognizable linear contour on the safety helmet from the safety helmet wearing image, comparing it with the pre-stored linear contours of various models of safety helmets, and screening out the safety helmet model with the highest overlap, which is recorded as the worker's safety helmet wearing model.

[0012] The preset standard contour image of the helmet wearing model is extracted and compared with the helmet wearing image. The template matching algorithm is used to search for areas similar to the standard contour image in the helmet wearing image, which are recorded as potential areas, and the clarity of each potential area is calculated.

[0013] A clarity threshold is set, potential areas whose clarity exceeds the set clarity threshold are circled and recorded as visually recognizable areas, and potential areas whose clarity does not exceed the set clarity threshold are recorded as visually distorted areas.

[0014] In a specific embodiment, the analysis method for analyzing the wearing quality defect rate of the visually identifiable area of ​​the helmet is: extracting each contour appearance identification parameter of each visually identifiable area based on image recognition technology.

[0015] The apparent identification parameters of each contour in each visually identifiable area are compared with the preset allowable values ​​of the corresponding apparent identification parameters of the contour, and the wearing quality defect rate of the visually identifiable area of ​​the helmet is analyzed.

[0016] In a specific embodiment, the method for evaluating the standardization of the helmet wearing position is: using an edge detection algorithm, by finding edge lines that meet the shape characteristics of the helmet and the contour characteristics of the face and head, extending and connecting the edge lines that meet the shape characteristics of the helmet to form a helmet covering contour, and extending and connecting the edge lines that meet the contour characteristics of the face and head to form a worker's facial contour, thereby integrating each visually identifiable area into a closed contour of the helmet.

[0017] By obtaining the positional coverage relationship between the helmet shell and the worker's head on the closed contour of the helmet, the compliance factor of the helmet shell position is evaluated.

[0018] By obtaining the position binding relationship between the tether and the worker's head on the closed contour of the helmet, the compliance factor of the tether position corresponding to the helmet is evaluated.

[0019] Based on these two indicators, the standardization of the helmet wearing position is analyzed.

[0020] In a specific embodiment, the evaluation method of the helmet shell position compliance factor is: extracting the worker's facial contour area ratio from the worker's facial contour, identifying the worker's head deflection direction, and the worker's head deflection direction includes the front face direction and the side face direction.

[0021] When the worker's head is deflected in the frontal direction, the brim of the helmet's coverage contour is obtained, and the visually recognizable symmetrical side endpoints on the brim are used as reference points to obtain the horizontal distance L between the two reference points. The height distance L between the vertices of the helmet's coverage contour and the two reference points on the brim is extracted, and the vertical axis inclination angle of the helmet's coverage contour is obtained by comparing them.

[0022] Extract the median line of the worker's head from the closed contour of the helmet, and extract the median line of the helmet shell corresponding to the helmet covering contour. Compare and obtain the horizontal axis inclination angle θ′ of the helmet covering contour, and evaluate the compliance factor of the helmet corresponding to the helmet shell position. Wherein θ″ is the preset reference value of the horizontal axis inclination angle, and θ0 is the preset reference value of the vertical axis inclination angle.

[0023] When the worker's head is deflected in the side face direction, the connection line between the worker's ear and eye is extracted from the worker's facial contour, and it is compared with the connection line of the two reference points corresponding to the brim line on the helmet coverage contour to obtain the helmet's line of sight coverage angle.

[0024] The safety helmet sight coverage angle is compared with the preset suitable coverage angle range. If the safety helmet sight coverage angle is within the preset suitable coverage angle range, the first preset value is recorded as the safety helmet corresponding to the helmet shell position compliance factor; if the safety helmet sight coverage angle is outside the preset suitable coverage angle range, the second preset value is recorded as the safety helmet corresponding to the helmet shell position compliance factor.

[0025] In a specific embodiment, the evaluation method of the safety helmet corresponding to the tether position compliance factor is: extracting the tether binding contour trajectory and the worker's facial jaw contour trajectory from the safety helmet coverage contour, and using the contour comparison algorithm to obtain the offset amplitude of the tether binding contour trajectory.

[0026] Obtain the tether buckle binding method from the helmet coverage contour, compare it with the preset standard binding method, and analyze the compliance factor of the tether position corresponding to the helmet B2, where B1 indicates that the tether buckle binding method is consistent with the preset standard binding method, B2 indicates that the tether buckle binding method is inconsistent with the preset standard binding method, β indicates the preset influence factor when the tether buckle binding method is inconsistent with the preset standard binding method, and e is a natural constant.

[0027] In a specific embodiment, the method for constructing a thermal correction map of the visual distortion area of ​​the helmet is as follows: using environmental measuring instruments to detect the air temperature and thermal radiation intensity of the worker's environment, evaluating the amplitude of human body temperature change and the amplitude of heat absorption of the helmet, and determining the interference indicators of the heat distribution in the visually identifiable area based on a specific application algorithm.

[0028] According to the above interference indicators, the image filtering technology is used to reduce the interference of the heat distribution map of the worker's head, and the heat correction map of the visual distortion area of ​​the helmet is obtained.

[0029] In a specific embodiment, the analysis method for analyzing the wearing quality defect rate of the visual distortion area of ​​the helmet is: comparing the standard thermal characteristic distribution map of the helmet wearing model with the thermal correction map of the visual distortion area of ​​the helmet, and extracting each heat flow area in the thermal correction map that is different from the standard thermal characteristic distribution map.

[0030] The corresponding texture features of each heat flow area are extracted, and the damage type of each heat flow area is identified based on image recognition technology, and then the texture area of ​​each heat flow area is extracted.

[0031] The damage type of each heat flow area is compared with the preset quality reduction weight corresponding to each damage type to obtain the corresponding quality reduction weight of each heat flow area, and the wearing quality defect rate of the visual distortion area of ​​the helmet is analyzed based on this.

[0032] In a specific embodiment, the analysis of the worker's helmet wearing fit is specifically as follows: matching the worker's helmet wearing image with the pre-stored identity verification images of each worker to identify the worker's identity, and then extracting the worker's head length distance d 长 , and extract the preset top thickness distance d corresponding to the worker's helmet wearing model 厚 .

[0033] Get the distance d between the vertex of the helmet's contour and the lowest point of the worker's head jaw 总 , compare it with the length of the worker's head, and obtain the fitting distance d=d between the helmet and the worker's head 总 -(d 长 +d 厚 ).

[0034] When the fitting distance is greater than 0, it means that the size of the safety helmet is too small. At this time, the fitting degree of the worker's safety helmet is set to the specified value.

[0035] When the fitting distance is equal to 0, locate the binding points of the tether and the helmet shell at both ends from the helmet coverage contour, obtain the close distance between the tether and the worker's facial contour, record them as d1 and d2 respectively, obtain the fitting length L between the tether binding trajectory and the worker's facial contour, and analyze the tightness of the tether Among them, d0 is the set reference close distance, L0 is the set reference fitting length, which is recorded as the worker's helmet wearing fit.

[0036] In a specific embodiment, the specific analysis method of the helmet protection quality index is: summing the wearing quality defect rates of the visually identifiable area and the visually distorted area of ​​the helmet to obtain the apparent defect degree of the helmet.

[0037] The helmet protection quality index is obtained through standardized calculation based on the apparent degree of defect of the helmet, the standardization of the helmet wearing position, and the fit of the worker's helmet.

[0038] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) The present invention extracts each visually identifiable area and each visually distorted area in the worker's safety helmet wearing image. On the one hand, the wearing quality defect rate of the visually identifiable area is determined based on the wearing image of each visually identifiable area. On the other hand, the wearing quality defect rate of the visually distorted area is determined by collecting the thermal image of each visually identifiable area. This not only improves the overall utilization rate of the image information, but also makes those distorted image areas that were originally regarded as "useless" valuable.

[0039] (2) The present invention determines the position conformity factor of the helmet shell corresponding to the helmet by identifying the deflection direction of the worker's head, which can improve the diversity of scene applicability of image acquisition.

[0040] (3) The present invention can generate a heat correction map by detecting the environmental parameters of the workers, effectively eliminating heat distribution interference, improving image quality, and making the monitoring of the helmet wearing status more accurate.

[0041] (4) The present invention analyzes the fit of the worker's safety helmet by obtaining the fit distance between the safety helmet and the worker's head, which can reflect the protective relationship between the size of the safety helmet and the worker's head. The close fit not only helps to improve the protective effect, but also reduces the risk of accidental falling off of the safety helmet during work. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0043] Figure 1 This is a schematic diagram of the connection of the system modules of the present invention.

[0044] Figure 2 It is a schematic diagram of the covering profile of the safety helmet of the present invention.

[0045] Reference numerals: 1 is the vertex of the helmet's covering contour, 21 and 22 are visually identifiable symmetrical side endpoints on the brim line. DETAILED DESCRIPTION

[0046] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0047] See also Figure 1 As shown, the present invention provides a helmet detection system based on image recognition, which includes: a helmet wearing image detection module, a helmet wearing quality analysis module, a helmet wearing position analysis module, a thermal imaging acquisition and analysis module, and a helmet protection quality assessment module.

[0048] The safety helmet wearing image detection module is respectively connected to the safety helmet wearing quality analysis module and the safety helmet wearing position analysis module; the safety helmet wearing quality analysis module is respectively connected to the thermal imaging acquisition and analysis module; and the safety helmet protection quality assessment module is respectively connected to the safety helmet wearing position analysis module and the thermal imaging acquisition and analysis module.

[0049] The helmet wearing image detection module is used to collect the worker's helmet wearing image, extract each visually identifiable area and each visually distorted area, obtain the worker's helmet wearing model, and match the standard thermal feature distribution map of the helmet wearing model.

[0050] In a preferred embodiment, the extraction of each visually recognizable area and each visually distorted area is specifically: extracting a recognizable linear contour on the safety helmet from the safety helmet wearing image, comparing it with the pre-stored linear contours of various models of safety helmets, screening out the safety helmet model with the highest overlap, and recording it as the worker's safety helmet wearing model.

[0051] The preset standard contour image of the helmet wearing model is extracted and compared with the helmet wearing image. The template matching algorithm is used to search for areas similar to the standard contour image in the helmet wearing image, which are recorded as potential areas. The detailed texture area and texture chromaticity depth of each potential area in the helmet wearing image are extracted based on image processing technology, and the clarity of each potential area is calculated accordingly.

[0052] Specifically, the clarity of each potential area is determined by the detail texture area and texture chromaticity depth of the corresponding area. The larger the detail texture area and the greater the texture chromaticity depth of the area, the higher the clarity value.

[0053] A clarity threshold is set, potential areas whose clarity exceeds the set clarity threshold are circled and recorded as visually recognizable areas, and potential areas whose clarity does not exceed the set clarity threshold are recorded as visually distorted areas.

[0054] Specifically, the distortion causes of each visual distortion area include light, shadow, occlusion, etc.

[0055] The helmet wearing quality analysis module is used to extract the apparent identification parameters of each contour of each visually identifiable area and analyze the wearing quality defect rate of the visually identifiable area of ​​the helmet.

[0056] In a preferred embodiment, the analysis method for analyzing the wearing quality defect rate of the visually identifiable area of ​​the helmet is: extracting the contour appearance identification parameters S of each visually identifiable area based on image recognition technology jz , each contour appearance identification parameter includes the damaged area, the dirty area, etc., j is the number of the visually identifiable area, j = 1, 2, ..., n, z is the number of the contour appearance identification parameter, z = 1, 2, ..., m.

[0057] The contour appearance identification parameters of each visually identifiable area are compared with the preset allowable value S′ of the corresponding contour appearance identification parameter. z Compare and analyze the wearing quality defect rate of the visually identifiable area of ​​the helmet

[0058] The helmet wearing position analysis module is used to integrate each visually identifiable area into a helmet closed contour, which includes the worker's facial contour and the helmet covering contour, and evaluate the standardization of the helmet wearing position.

[0059] In a preferred embodiment, the method for evaluating the standardization of the wearing position of a safety helmet is as follows: using an edge detection algorithm, by finding edge lines that conform to the shape features of the safety helmet and the contour features of a human face and head, the edge lines that conform to the shape features of the safety helmet are extended and connected to form a safety helmet covering contour, and the edge lines that conform to the contour features of a human face and head are extended and connected to form a worker's facial contour, thereby integrating each visually identifiable area into a safety helmet closed contour.

[0060] By obtaining the positional coverage relationship between the helmet shell and the worker's head on the closed contour of the helmet, the position of the helmet corresponding to the helmet shell is evaluated to meet the factor λ.

[0061] Specifically, the position coverage relationship between the helmet shell and the worker's head on the closed contour of the safety helmet refers to the inclination angle of the vertical axis of the safety helmet covering contour, the inclination angle of the horizontal axis of the safety helmet covering contour, and the safety helmet sight coverage angle.

[0062] By obtaining the position binding relationship between the tether and the worker's head on the closed contour of the helmet, the corresponding tether position of the helmet is evaluated to meet the factor

[0063] Specifically, the position binding relationship between the tether and the worker's head on the closed contour of the safety helmet refers to the offset amplitude of the tether binding contour trajectory and the tether buckle binding method.

[0064] Based on these two indicators, the standardization of helmet wearing position is analyzed.

[0065] See also Figure 2 As shown, in a further preferred embodiment, the evaluation method of the safety helmet corresponding to the helmet shell position compliance factor is: extracting the worker's facial contour area ratio from the worker's facial contour, when it exceeds a preset first threshold of the facial contour area ratio, recording the worker's head deflection direction as the front face direction, when it exceeds a preset second threshold of the facial contour area ratio, recording the worker's head deflection direction as the side face direction, when it does not exceed the preset second threshold of the facial contour area ratio, re-collecting the worker's safety helmet wearing image, and identifying the worker's head deflection direction based on this, the worker's head deflection direction includes the front face direction and the side face direction.

[0066] When the worker's head is deflected in the frontal direction, the brim of the helmet's coverage contour is obtained, and the visually recognizable symmetrical side endpoints on the brim are used as reference points to obtain the horizontal distance L between the two reference points. The height distance L between the vertices of the helmet's coverage contour and the two reference points on the brim is extracted, and the vertical axis inclination angle of the helmet's coverage contour is obtained by comparing them.

[0067] Extract the median line of the worker's head from the closed contour of the helmet, and extract the median line of the helmet shell corresponding to the helmet covering contour. Compare and obtain the horizontal axis inclination angle θ′ of the helmet covering contour, and evaluate the compliance factor of the helmet corresponding to the helmet shell position. Wherein θ″ is the preset reference value of the horizontal axis inclination angle, and θ0 is the preset reference value of the vertical axis inclination angle.

[0068] When the worker's head is deflected in the side face direction, the connection line between the worker's ear and eye is extracted from the worker's facial contour, and it is compared with the connection line of the two reference points corresponding to the brim line on the helmet coverage contour to obtain the helmet's line of sight coverage angle.

[0069] The safety helmet sight coverage angle is compared with the preset suitable coverage angle range. If the safety helmet sight coverage angle is within the preset suitable coverage angle range, the first preset value is recorded as the safety helmet corresponding to the helmet shell position compliance factor; if the safety helmet sight coverage angle is outside the preset suitable coverage angle range, the second preset value is recorded as the safety helmet corresponding to the helmet shell position compliance factor.

[0070] The present invention determines the position conformity factor of the helmet shell corresponding to the helmet by identifying the deflection direction of the worker's head, which can improve the scene applicability diversity of image acquisition.

[0071] In a further preferred embodiment, the evaluation method of the safety helmet corresponding to the tether position compliance factor is: extracting the tether binding contour trajectory and the worker's facial jaw contour trajectory from the safety helmet coverage contour, and using the contour comparison algorithm to obtain the offset amplitude of the tether binding contour trajectory

[0072] Obtain the tether buckle binding method from the helmet coverage contour, compare it with the preset standard binding method, and analyze the compliance factor of the tether position corresponding to the helmet B2, where B1 indicates that the tether buckle binding method is consistent with the preset standard binding method, B2 indicates that the tether buckle binding method is inconsistent with the preset standard binding method, β indicates the preset influence factor when the tether buckle binding method is inconsistent with the preset standard binding method, and e is a natural constant.

[0073] The thermal imaging acquisition and analysis module is used to capture the heat distribution map of the worker's head using a thermal imaging camera, construct a heat correction map of the visual distortion area of ​​the helmet, and analyze the wearing quality defect rate of the visual distortion area of ​​the helmet based on it.

[0074] In a preferred embodiment, the method for constructing the heat correction map of the visual distortion area of ​​the helmet is: using an environmental measuring instrument to detect the air temperature T and the thermal radiation intensity E of the worker's environment, evaluating the amplitude of human body temperature change and the amplitude of heat absorption of the helmet, and determining the interference indicator of the heat distribution in the visually recognizable area according to a specific application algorithm.

[0075] Specifically, the human body temperature variation amplitude is obtained by comparing the air temperature of the worker's environment with the preset conventional ambient temperature T0. The heat absorption amplitude of the helmet is obtained by comparing the thermal radiation intensity with the preset reference thermal radiation intensity E0.

[0076] According to the above interference indicators, image filtering technology (such as Gaussian filtering, median filtering, etc.) is used to reduce interference on the heat distribution map of the worker's head to obtain a heat correction map of the visual distortion area of ​​the helmet.

[0077] The present invention can generate a heat correction map by detecting the environmental parameters of the workers, effectively eliminate the interference of heat distribution, improve the image quality, and make the monitoring of the helmet wearing status more accurate.

[0078] In a further preferred embodiment, the analysis method for analyzing the wearing quality defect rate of the visual distortion area of ​​the helmet is: comparing the standard thermal characteristic distribution map of the helmet wearing model with the thermal correction map of the visual distortion area of ​​the helmet, and extracting each heat flow area in the thermal correction map that is different from the standard thermal characteristic distribution map.

[0079] Extract the texture features corresponding to each heat flow area, identify the damage type of each heat flow area based on image recognition technology, the damage type includes crack type and dirt type, and then extract the texture area S of each heat flow area r , r is the number of the heat flow area, r = 1, 2,…, a.

[0080] Compare the damage type of each heat flow area with the preset quality reduction weight corresponding to each damage type, and obtain the quality reduction weight H corresponding to each heat flow area r Based on this, the wearing quality defect rate of the helmet visual distortion area is analyzed Where S0 represents the set reference texture area, and a is the number of heat flow areas.

[0081] The present invention extracts visually identifiable areas and visually distorted areas in the worker's safety helmet wearing image. On the one hand, the wearing quality defect rate of the visually identifiable area is determined based on the wearing image of each visually identifiable area. On the other hand, the wearing quality defect rate of the visually distorted area is determined by collecting thermal images of each visually identifiable area. This not only improves the overall utilization rate of image information, but also makes those distorted image areas that were originally regarded as "useless" valuable.

[0082] The helmet protection quality assessment module is used to analyze the fit of workers' helmets, and then comprehensively analyze the helmet protection quality indicators.

[0083] In a preferred embodiment, the analysis of the fit of the worker's helmet is specifically as follows: matching the worker's helmet wearing image with pre-stored identity verification images of each worker to identify the worker's identity, and then extracting the worker's head length distance d long, and extracting the preset top thickness distance d thick corresponding to the worker's helmet wearing model.

[0084] Obtain the distance dtotal between the vertex of the helmet's coverage contour and the lowest point of the worker's chin, compare it with the worker's head length distance, and obtain the fitting distance d=dtotal-(dlength+dthickness) between the helmet and the worker's head.

[0085] When the fitting distance is greater than 0, it means that the size of the safety helmet is too small. At this time, the fitting degree of the worker's safety helmet is set to the specified value.

[0086] When the fitting distance is equal to 0, locate the binding points of the tether and the helmet shell at both ends from the helmet coverage contour, obtain the close distance between the tether and the worker's facial contour, record them as d1 and d2 respectively, obtain the fitting length L between the tether binding trajectory and the worker's facial contour, and analyze the tightness of the tether Among them, d0 is the set reference close distance, L0 is the set reference fitting length, which is recorded as the worker's helmet wearing fit.

[0087] The present invention analyzes the fit of the worker's safety helmet by obtaining the fit distance between the safety helmet and the worker's head, which can reflect the protective relationship between the safety helmet size and the worker's head. The close fit not only helps to improve the protective effect, but also reduces the risk of accidental falling off of the safety helmet during work.

[0088] In a further preferred embodiment, the specific analysis method of the helmet protection quality index is: summing the wearing quality defect rates of the visually identifiable area and the visually distorted area of ​​the helmet to obtain the apparent defect degree σ of the helmet.

[0089] The helmet protection quality index is obtained by standardized calculation based on the apparent defect degree of the helmet, the standardization degree of the helmet wearing position φ, and the worker's helmet wearing fit τ.

[0090] The present invention can comprehensively evaluate the protective performance of the safety helmet by comprehensively considering the apparent degree of defect, wearing position standardization and wearing fit of the safety helmet. This comprehensive evaluation method can more accurately reflect the actual protective effect of the safety helmet than a single indicator, and provide more reliable safety protection for workers.

[0091] The above contents are merely examples and explanations of the concept of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, they shall all fall within the protection scope of the present invention.

Claims

1. A helmet detection system based on image recognition, characterized in that: The system includes: a helmet wearing image detection module: used to collect images of workers wearing helmets, extract each visually recognizable area and each visually distorted area, obtain the worker's helmet wearing model, and match the standard thermal feature distribution map of the helmet wearing model; Safety helmet wearing quality analysis module: used to extract the apparent identification parameters of each contour in each visually identifiable area and analyze the wearing quality defect rate of the visually identifiable area of ​​the safety helmet; Helmet wearing position analysis module: used to integrate each visually identifiable area into a closed contour of the helmet, which includes the worker's facial contour and the helmet coverage contour, and evaluate the standardization of the helmet wearing position; Thermal imaging acquisition and analysis module: used to capture the heat distribution map of the worker's head using a thermal imaging camera, construct a thermal correction map of the visual distortion area of ​​the helmet, and analyze the wearing quality defect rate of the visual distortion area of ​​the helmet based on this; Safety helmet protection quality assessment module: used to analyze the fit of workers' safety helmets, and then comprehensively analyze the safety helmet protection quality indicators; The analysis method for analyzing the wearing quality defect rate of the visually identifiable area of ​​the helmet is: extracting the contour appearance identification parameters S of each visually identifiable area based on image recognition technology jz , each contour appearance identification parameter includes the damaged area, the dirty area, etc., j is the number of the visually identifiable area, j = 1, 2, ..., n, z is the number of the contour appearance identification parameter, z = 1, 2, ..., m; The contour appearance identification parameters of each visually identifiable area are compared with the preset allowable value S′ of the corresponding contour appearance identification parameter. z Compare and analyze the wearing quality defect rate of the visually identifiable area of ​​the helmet The method for evaluating the standardization of the helmet wearing position is: Using edge detection algorithms, by finding edge lines that match the shape features of the helmet and the contour features of the human face, the edge lines that match the shape features of the helmet are extended and connected to form the helmet covering contour, and the edge lines that match the contour features of the human face are extended and connected to form the worker's facial contour, thereby integrating each visually recognizable area into a closed contour of the helmet; By obtaining the positional coverage relationship between the helmet shell and the worker's head on the closed contour of the helmet, the corresponding helmet shell position of the helmet is evaluated to meet the factor λ; By obtaining the position binding relationship between the tether and the worker's head on the closed contour of the helmet, the corresponding tether position of the helmet is evaluated to meet the factor Based on these two indicators, the standardization of helmet wearing position is analyzed. The analysis method for analyzing the wearing quality defect rate of the visual distortion area of ​​the helmet is: The standard heat characteristic distribution map of the helmet wearing model is compared with the heat correction map of the helmet visual distortion area, and the heat flow areas in the heat correction map that are different from the standard heat characteristic distribution map are extracted; Extract the texture features corresponding to each heat flow area, identify the damage type of each heat flow area based on image recognition technology, and then extract the texture area S of each heat flow area r , r is the number of the heat flow area, r = 1, 2, ..., a; Compare the damage type of each heat flow area with the preset quality reduction weight corresponding to each damage type, and obtain the quality reduction weight H corresponding to each heat flow area. r Based on this, the wearing quality defect rate of the helmet visual distortion area is analyzed. Where S0 represents the set reference texture area, and a is the number of heat flow areas; The analysis of the worker's helmet wearing fit is specifically as follows: The worker's helmet wearing image is matched with the pre-stored identity verification images of each worker to identify the worker's identity and then extract the worker's head length distance d. 长 , and extract the preset top thickness distance d corresponding to the worker's helmet wearing model 厚 ; Get the distance d between the vertex of the helmet's contour and the lowest point of the worker's head jaw 总 , compare it with the length of the worker's head, and obtain the fitting distance d=d between the helmet and the worker's head 总 -(d 长 +d 厚 ); When the fitting distance is greater than 0, it means that the size of the helmet is too small, and the fitting degree of the worker's helmet is set to the specified value; When the fitting distance is equal to 0, locate the binding points of the tether and the helmet shell at both ends from the helmet coverage contour, obtain the close distance between the tether and the worker's facial contour, record them as d1 and d2 respectively, obtain the fitting length L between the tether binding trajectory and the worker's facial contour, and analyze the tightness of the tether Where d0 is the set reference close distance, L0 is the set reference fit length, which is recorded as the worker's helmet fit; The specific analysis method of the helmet protection quality index is as follows: The wearing quality defect rate of the helmet's visually recognizable area and visually distorted area is summed to obtain the apparent defect degree of the helmet; The helmet protection quality index is obtained through standardized calculation based on the apparent degree of defect of the helmet, the standardization of the helmet wearing position, and the fit of the worker's helmet.

2. The helmet detection system based on image recognition according to claim 1, characterized in that: The extracting of each visually identifiable area and each visually distorted area specifically comprises: extracting a recognizable linear contour on the helmet from the helmet wearing image, comparing it with the pre-stored linear contours of various models of helmets, screening out the helmet model with the highest overlap, and recording it as the helmet wearing model of the worker; Extract the preset standard contour image of the helmet wearing model, compare it with the helmet wearing image, use the template matching algorithm to search for areas similar to the standard contour image in the helmet wearing image, record them as potential areas, and calculate the clarity of each potential area; A clarity threshold is set, potential areas whose clarity exceeds the set clarity threshold are circled and recorded as visually recognizable areas, and potential areas whose clarity does not exceed the set clarity threshold are recorded as visually distorted areas.

3. The helmet detection system based on image recognition according to claim 1, characterized in that: The evaluation method of the helmet shell position compliance factor is as follows: Extract the worker's facial contour area ratio from the worker's facial contour and identify the worker's head deflection direction, which includes the frontal face direction and the side face direction; When the worker's head is deflected in the direction of the front face, the brim line of the helmet's coverage contour is obtained, and the visually recognizable symmetrical side endpoints on the brim line are used as reference points to obtain the horizontal distance L between the two reference points. 平 , extract the height distance L between the vertices of the helmet coverage contour and the two reference points on the brim line 高 , and compare them to get the vertical axis inclination angle of the helmet coverage profile Extract the median line of the worker's head from the closed contour of the helmet, and extract the median line of the helmet shell corresponding to the helmet covering contour. Compare and obtain the horizontal axis inclination angle θ′ of the helmet covering contour, and evaluate the compliance factor of the helmet corresponding to the helmet shell position. Wherein θ″ is the preset reference value of the horizontal axis inclination angle, and θ0 is the preset reference value of the vertical axis inclination angle; When the worker's head is deflected in the side face direction, the connection line between the worker's ear and eye is extracted from the worker's facial contour, and compared with the connection line of the two reference points corresponding to the brim line on the helmet coverage contour to obtain the helmet's sight coverage angle; The safety helmet sight coverage angle is compared with the preset suitable coverage angle range. If the safety helmet sight coverage angle is within the preset suitable coverage angle range, the first preset value is recorded as the safety helmet corresponding to the helmet shell position compliance factor; if the safety helmet sight coverage angle is outside the preset suitable coverage angle range, the second preset value is recorded as the safety helmet corresponding to the helmet shell position compliance factor.

4. The helmet detection system based on image recognition according to claim 1, characterized in that: The evaluation method of the safety helmet corresponding to the tether position compliance factor is: Extract the tether binding contour trajectory and the worker's facial jaw contour trajectory from the helmet coverage contour, and use the contour comparison algorithm to obtain the offset amplitude of the tether binding contour trajectory Obtain the tether buckle binding method from the helmet coverage contour, compare it with the preset standard binding method, and analyze the compliance factor of the tether position corresponding to the helmet B2, where B1 indicates that the tether buckle binding method is consistent with the preset standard binding method, B2 indicates that the tether buckle binding method is inconsistent with the preset standard binding method, β indicates the preset influence factor when the tether buckle binding method is inconsistent with the preset standard binding method, and e is a natural constant.

5. The helmet detection system based on image recognition according to claim 1, characterized in that: The method for constructing a heat correction map of the visual distortion area of ​​the helmet is as follows: using an environmental measuring instrument to detect the air temperature T and the thermal radiation intensity E of the worker's environment, evaluating the amplitude of human body temperature change and the amplitude of heat absorption of the helmet, and determining the interference indicator of the heat distribution in the visually recognizable area according to a specific application algorithm The human body temperature variation amplitude is obtained by comparing the air temperature of the worker's environment with the preset normal ambient temperature T0. The heat absorption amplitude of the helmet is obtained by comparing the thermal radiation intensity with the preset reference thermal radiation intensity E0. According to the above interference indicators, the image filtering technology is used to reduce the interference of the heat distribution map of the worker's head, and the heat correction map of the visual distortion area of ​​the helmet is obtained.

Images