CN-122024178-A - Method and system for accurately associating personal protection equipment of constructor with constructor

CN122024178ACN 122024178 ACN122024178 ACN 122024178ACN-122024178-A

Abstract

The invention relates to a method and a system for precisely associating constructor personal protective equipment with personnel, wherein the method comprises the steps of obtaining images of a plurality of defined personnel target frames and protective equipment target frames, calculating the occupation ratio of each protective equipment target frame to each personnel target frame, screening candidate associated personnel based on the occupation ratio, obtaining the position of a human body key point corresponding to each personnel target frame, calculating the minimum distance from the central point of the protective equipment target frame to the relevant key point of each candidate personnel, mapping the occupation ratio and the key point distance to a percentile scale, fusing the occupation ratio and the key point distances to generate an associated quality score, and finally carrying out global optimization distribution based on the associated quality score and conforming to the standard wearing quantity constraint of a single person to realize the precise association of the personal protective equipment with the personnel. The invention can realize accurate judgment of the attribution of the personal protection equipment in complex and dense scenes, and provides core technical support for intelligent and personalized safety compliance inspection.

Inventors

LI YANG
ZHAO XIN
LI ZHENG
Tong Siyuan
SHEN PENG

Assignees

中国联合工程有限公司

Dates

Publication Date: 20260512
Application Date: 20260413

Claims (10)

1. A method for precisely associating personal protection equipment of construction personnel with personnel, which is characterized by comprising the following steps: Step S1, acquiring an image containing personnel and personal protective equipment, and acquiring a plurality of defined personnel target frames and a plurality of protective equipment target frames in the image; Step S2, calculating the cross-over ratio of each protective equipment target frame to each personnel target frame, and screening candidate associated personnel target frames for the protective equipment target frames based on the cross-over ratio, wherein the protective equipment target frames and each candidate associated personnel target frame form a candidate associated pair respectively; step S3, acquiring a group of human body key point positions corresponding to each human body target frame according to a human body posture estimation algorithm; Step S4, calculating the minimum Euclidean distance from the central point of each protective equipment target frame to a group of related human body key points corresponding to the personnel target frame according to each protective equipment target frame and each candidate related personnel target frame; Step S5, mapping the cross-over ratio and the minimum Euclidean distance to the same comparison scale for each candidate association pair to obtain a cross-over ratio mapping value and a minimum Euclidean distance mapping value, and generating association quality scores of the candidate association pairs based on the cross-over ratio mapping value and the minimum Euclidean distance mapping value; And S6, based on the association quality scores of all candidate association pairs, carrying out global optimization allocation in combination with the single person standard wearing quantity constraint of each type of personal protection equipment, and uniquely associating each protection equipment target frame to one person target frame.
2. The method for precisely associating the personal protective equipment with the person according to claim 1, wherein in the step S2, candidate associated person target frames are screened out for the protective equipment target frames by: (1) Setting an occupancy threshold for each type of personal protection equipment; (2) For a protective equipment target frame, calculating the cross-occupation ratio of the protective equipment target frame and all personnel target frames in the image respectively; (3) And determining the personnel target frames with the cross-occupation ratio larger than or equal to the corresponding threshold value of the personal protection equipment as candidate associated personnel target frames of the protection equipment target frames, wherein the protection equipment target frames and each candidate associated personnel target frame respectively form a candidate associated pair.
3. The method for precisely associating the personal protective equipment with the person for construction personnel according to claim 1, wherein the step S4 is specifically: for a protective equipment target frame, calculating the coordinates of the central point of the protective equipment target frame, and for a candidate associated person target frame corresponding to the protective equipment target frame, acquiring a related human body key point subset K t corresponding to the person, calculating the Euclidean distance from the central point of the protective equipment to each key point in the human body key point subset K t , and taking the minimum value as the minimum Euclidean distance.
4. The method for precisely associating constructor personal protective equipment with personnel according to claim 3, wherein in the step S5, the percentile mapping method is adopted to map the intersection ratio and the minimum euclidean distance to the same comparison scale: For the cross-over ratio, calculating the percentile of the current cross-over ratio in the distribution based on pre-constructed cross-over ratio empirical distribution of the personal protection equipment, and taking the percentile as the cross-over ratio mapping value; And for the minimum Euclidean distance, calculating the percentile of the current minimum Euclidean distance value in the distribution based on the pre-constructed key point distance experience distribution of the personal protection equipment, and taking the percentile as the minimum Euclidean distance mapping value.
5. The method for precisely associating constructor personal protective equipment with personnel according to claim 4, wherein the pre-constructed cross-ratio empirical distribution and the key point distance empirical distribution of each type of personal protective equipment are obtained based on statistical analysis of an image data set containing ownership labels of personnel and personal protective equipment.
6. The method for precisely associating constructor personal protective equipment with personnel according to claim 4, wherein the associating quality fraction adopts the following strategies: , Wherein q is an associated quality score, p dist is the minimum euclidean distance mapping value, p IoP is the cross-over ratio mapping value, and the symbol indicates an empty set.
7. The method for precisely associating the personal protective equipment with the personnel according to claim 1, wherein the step S6 is a step of restricting the number of the personal standard wearing of each type of personal protective equipment, specifically, ensuring that the number of the same type of protective equipment target frames allocated to the same personal target frame does not exceed the standard upper wearing limit in the associating and allocating process, wherein the standard upper wearing limit is set according to the common physical wearing knowledge of the personal protective equipment.
8. The method for precisely associating constructor personal protective equipment with personnel according to claim 1, wherein in step S6, the global optimization allocation is performed by adopting a greedy algorithm, comprising the following steps: (1) Sorting all candidate association pairs according to the association quality scores thereof from high to low; (2) Traversing the sorted candidate association pair list in sequence; (3) For the candidate association pair currently traversed, if the protective equipment target frame is not allocated and the number of the currently allocated personal protective equipment of the personnel target frame in the candidate association pair does not reach the standard number constraint, allocating the personal protective equipment to the personnel and updating the state; (4) Repeating the previous step until the traversal is completed.
9. The method for precisely associating constructor personal protective equipment with personnel according to claim 5, wherein the image dataset is constructed by the steps of: (1) Acquiring an image containing a person and a plurality of personal protection devices; (2) Marking the image, distributing a unique personnel identifier for each detected personnel, and distributing an owner identifier for each detected personal protection equipment, wherein the owner identifier is the same as the personnel identifier of the personnel if the personal protection equipment is worn by the personnel; (3) Dividing the marked image into a training set, a verification set and a test set to form the image data set.
10. A personal protection apparatus and personnel accurate association system for implementing the constructor personal protection apparatus and personnel accurate association method according to any one of claims 1 to 9, comprising: An input module for acquiring an image containing personnel and personal protection equipment and a plurality of personnel target frames and protection equipment target frames defined in the image; the candidate screening module is used for calculating the occupation ratio of each protective equipment target frame and each personnel target frame and screening candidate associated personnel target frames based on the occupation ratio; The key point acquisition module is used for acquiring the positions of the key points of the human body corresponding to each human target frame; the distance calculation module is used for calculating the minimum Euclidean distance from the central point of each protective equipment target frame to the relevant human body key point of each candidate associated person target frame; The quality scoring module is used for mapping the cross-over ratio and the minimum Euclidean distance to the same comparison scale and fusing to generate the association quality score of each candidate association pair; And the global distribution module is used for executing global optimization distribution based on the mass fraction and the individual person standard wearing quantity constraint of each type of personal protection equipment and outputting a correlation result.

Description

Method and system for accurately associating personal protection equipment of constructor with constructor Technical Field The invention relates to the technical fields of intelligent construction and computer vision, in particular to a method and a system for accurately associating personal protection equipment of constructors with personnel, which are used for improving the level of intellectualization and precision of safety management in high-risk operation scenes such as building construction and the like. Background In high-risk industries such as building construction, factory operation and the like, forced and correct use of personal protection equipment is a last line of defense for guaranteeing personnel safety and preventing accidents. The traditional supervision mode relying on manual inspection of a safety officer has the problems of low efficiency, narrow coverage, incapability of realizing all-weather continuous monitoring and the like. In recent years, an automatic safety monitoring technology based on computer vision is rapidly developed, and the wearing condition of personal protection equipment is automatically identified by analyzing a monitoring video stream, so that the personal protection equipment becomes a hot spot for research and application. Current automated methods of compliance inspection of personal protective equipment typically focus on the identification of personal protective equipment, and ignore the differentiated requirements in practical applications due to job and scene differences. Wherein, how to accurately associate the detected personal protection equipment (such as a safety helmet, a safety belt, a protective glove and the like) with specific operators in a scene is a key premise for realizing individuation and fine combination rule judgment, and is a main bottleneck faced by the current technology. The existing method mainly has the defect that firstly, the association precision is limited by simple space measurement. Most existing solutions use the cross-connect ratio as a basis for spatial association of personnel with personal protection equipment. However, in a real complex work scenario, people are dense, the work gestures are diverse, and the size differences of different personal protection devices are huge. The calculation of the cross-over ratio is significantly affected by the size of the personnel detection frames, such as a small-sized glove close to a plurality of workers, and the cross-over ratio of the small-sized glove and the plurality of workers detection frames may be very low and close in value, so that the attribution of the small-sized glove cannot be reliably judged, and error association is extremely easy to generate. Second, the realistic physical constraints are ignored. In the existing method, basic physical wearing logic such as 'one worker wearing at most one safety helmet and two gloves' is not considered in normal condition during association, and the algorithm can be caused to output an association result which violates common sense, so that the system reliability is reduced. Finally, the data base for fine assessment and optimization is lacking. Publicly available personal protective equipment datasets mostly label only object categories and locations, lacking fine-grained labeling of "equipment-wearer" ownership relationships. This prevents the performance of the correlation algorithm from being quantified and compared accurately, preventing the technology from moving towards higher accuracy and robustness. Therefore, a core method that can adapt to complex, crowded and variable real working environments and realize high-precision and high-robustness association between personal protection equipment and operators is needed. Disclosure of Invention The invention provides a method and a system for accurately associating constructor personal protection equipment with personnel, which solve the problem of inaccurate and unreliable association under a complex scene by introducing new measurement of the cross-occupation ratio, fusing key point information of human bodies, combining physical constraint and constructing an image data set, and lay a core technical foundation for realizing automatic and personalized safety compliance inspection subsequently. The invention solves the problems by adopting the following technical scheme: a method for precisely associating personal protection equipment of constructors with personnel, comprising the following steps: Step S1, acquiring an image containing personnel and personal protective equipment, and acquiring a plurality of defined personnel target frames and a plurality of protective equipment target frames in the image; Step S2, calculating the cross-over ratio of each protective equipment target frame to each personnel target frame, and screening candidate associated personnel target frames for the protective equipment target frames based on the cross-over ratio, wherein the protective