CN-121998422-A - Engineering supervision potential safety hazard early warning method and system based on edge calculation

CN121998422ACN 121998422 ACN121998422 ACN 121998422ACN-121998422-A

Abstract

The application provides an engineering supervision potential safety hazard early warning method and system based on edge calculation, which are used for acquiring multi-source heterogeneous sensing data acquired by a plurality of monitoring points in a target construction area, performing feature fuzzy extraction on the multi-source heterogeneous sensing data to identify current construction working conditions and dominant risk factors, determining a potential safety hazard detection task to be executed according to the identified construction working conditions and risk factors, adaptively selecting a target model corresponding to the potential safety hazard detection task from a plurality of lightweight identification models pre-deployed in edge computing equipment to activate, and performing forward reasoning on the real-time multi-source heterogeneous sensing data of the corresponding monitoring points by using the activated target model in the edge computing equipment to generate a potential safety hazard identification result. By adopting the scheme of the application, the dispatching reasoning of the edge-side self-adaptive model based on the working condition perception can be realized, so that the potential safety hazard early warning efficiency of engineering supervision in a resource-limited environment is improved.

Inventors

JIN XIAOMING
WEI ZHONGHAO
HUANG QINZHAO
YUE JINLONG
HUANG JIEMIN
CHEN HUOLIN
YUE YINGCHUN
HE ZHIYUAN
LIU DANNA
HE SHANXIAN
DONG YI
WANG YONGQUAN
ZHAO SHIWU
LIU JINGTAO

Assignees

广州建筑工程监理有限公司

Dates

Publication Date: 20260508
Application Date: 20260123

Claims (10)

1. The utility model provides an engineering supervision potential safety hazard early warning method based on edge calculation, which is characterized by comprising the following steps: acquiring multi-source heterogeneous sensing data acquired by a plurality of monitoring points in a target construction area; performing feature fuzzy extraction on the multi-source heterogeneous sensing data to identify current construction working conditions and dominant risk factors, and determining potential safety hazard detection tasks to be executed according to the identified construction working conditions and risk factors; Adaptively selecting a target model corresponding to the potential safety hazard detection task from a plurality of light-weight identification models pre-deployed in edge computing equipment to activate, wherein the light-weight identification models are subjected to differential optimization design aiming at different construction stages or different risk types respectively; In the edge computing equipment, forward reasoning is carried out on the real-time multi-source heterogeneous sensing data of the corresponding monitoring points by utilizing the activated target model, and a potential safety hazard identification result is generated; And judging whether the potential safety hazard exists in the corresponding monitoring point according to the potential safety hazard identification result, generating early warning information when the potential safety hazard exists, and pushing the early warning information to an engineering supervision terminal.
2. The method of claim 1, wherein the multi-source heterogeneous sensing data specifically comprises video stream data acquired through visual sensors deployed at the monitoring points, and environmental sound and vibration spectrum data acquired through acoustic sensors deployed at the monitoring points.
3. The method of claim 1, wherein the feature fuzzy extraction of the multi-source heterogeneous awareness data to identify the current construction condition and dominant risk elements specifically comprises: Carrying out light feature matching on the collected video stream data and a preset typical construction scene template to obtain scene matching degree; Identifying an ongoing key job type based on the collected environmental acoustic and vibration spectrum data; fusing the scene matching degree and the identified key operation type, and calculating the membership degree of different preset construction working conditions by using a preset lightweight fuzzy inference rule base; And determining a preset construction working condition with the highest membership degree as the current construction working condition, and taking the high-risk factors of the history statistics under the working condition as the dominant risk factors.
4. The method of claim 1, wherein determining the potential safety hazard detection task to be performed based on the identified construction conditions and risk factors comprises: Acquiring at least one basic detection task to be periodically executed under the construction working condition; Acquiring at least one special detection task to be monitored in an important way from a predefined task mapping table according to the risk factors; combining and de-duplicating the at least one basic detection task and the at least one special detection task to obtain a potential safety hazard detection task to be executed, wherein the potential safety hazard detection task at least comprises one detection task.
5. The method of claim 1, wherein adaptively selecting a target model corresponding to the potential safety hazard detection task from a plurality of lightweight identification models pre-deployed in an edge computing device to activate comprises: extracting a task identifier of each detection task in the potential safety hazard detection tasks; inquiring in a model index table locally stored in the edge computing equipment according to the task identifiers, wherein the model index table records the mapping relation between different task identifiers and corresponding lightweight recognition model storage paths; And loading corresponding model parameters from a local storage space of the edge computing equipment to the memory according to the queried storage path, and taking the model which is loaded to the memory and corresponds to the detection task as the target model of the current activation.
6. The method of claim 1, wherein in the edge computing device, performing forward reasoning on real-time multi-source heterogeneous awareness data of the corresponding monitoring points by using the activated target model, and generating the potential safety hazard identification result specifically comprises: Intercepting input data segments of corresponding types and time sequences from the real-time multi-source heterogeneous perception data according to detection tasks corresponding to each activated target model; Carrying out standardized preprocessing on each input data segment to meet the input requirement of a corresponding target model; Feeding each input data segment subjected to standardized pretreatment into a corresponding target model respectively, and executing forward reasoning calculation on a calculation unit of the edge calculation equipment in parallel; Summarizing the identification results which are output by each target model and represent the risk state under the corresponding detection task, and generating the potential safety hazard identification results.
7. The method of claim 1, wherein determining whether a potential safety hazard exists at the corresponding monitoring point based on the potential safety hazard identification result comprises: Comparing each identification result in the potential safety hazard identification results with a preset judgment threshold value of the corresponding risk type; If any one of the identification results exceeds the corresponding judgment threshold value, judging that the hidden danger of the specific type exists; and generating a comprehensive hidden danger judgment conclusion according to the risk level and the risk position corresponding to all the overrun recognition results.
8. Engineering supervision potential safety hazard early warning system based on edge calculation, characterized by comprising: The acquisition module is used for acquiring multi-source heterogeneous sensing data acquired by a plurality of monitoring points in the target construction area; The processing module is used for carrying out feature fuzzy extraction on the multi-source heterogeneous sensing data so as to identify the current construction working condition and dominant risk elements, and determining a potential safety hazard detection task to be executed according to the identified construction working condition and risk elements; The processing module is further used for adaptively selecting a target model corresponding to the potential safety hazard detection task from a plurality of lightweight identification models pre-deployed in the edge computing equipment to activate; the processing module is further used for performing forward reasoning on the real-time multi-source heterogeneous sensing data of the corresponding monitoring points by utilizing the activated target model in the edge computing equipment to generate potential safety hazard identification results; And the pushing module is used for judging whether the potential safety hazard exists in the corresponding monitoring point according to the potential safety hazard identification result, generating early warning information when the potential safety hazard exists, and pushing the early warning information to the engineering supervision terminal.
9. A computer device comprising a memory storing code and a processor, wherein the processor is configured to obtain the code and perform the edge calculation based engineering supervision safety hazard warning method of any one of claims 1 to 7.
10. A computer-readable storage medium storing a computer program, wherein the computer program when executed by a processor implements the method for warning of an engineering supervision safety hazard based on edge calculation according to any one of claims 1 to 7.

Description

Engineering supervision potential safety hazard early warning method and system based on edge calculation Technical Field The application relates to the technical field of safety precaution, in particular to an engineering supervision potential safety hazard precaution method and system based on edge calculation. Background The main content of engineering supervision work comprises continuous supervision and risk control on the operation behaviors of construction site personnel, the operation states of mechanical equipment, the execution conditions of construction procedures and the environmental safety conditions, so as to ensure that the construction activities meet the safety standards and technical standards, engineering supervision safety early warning is a core link for ensuring the safety of the construction site personnel and equipment, the traditional early warning method mainly relies on manual inspection and fixed sensors, has the problems of large coverage blind areas, delayed response and the like, and in recent years, the automatic monitoring technology based on the Internet of things and the artificial intelligence is applied, but the intelligent early warning on the change of the real-time, accurate and complex construction environment is realized, so that the key to breakthrough is still urgent in the industry. In the prior art, engineering supervision safety real-time monitoring is realized by arranging various recognition algorithms on the edge, the design initially aims at assisting engineering supervision personnel to automatically recognize and prompt the risk of key safety elements of a construction site, so that the manual supervision intensity is reduced, the supervision coverage rate is improved, but a static model calling strategy is generally adopted, and the strategy has the remarkable defects that no matter how the current actual construction working condition and the risk focus are, the system always needs to continuously or alternately run all or a large number of preset models, so that limited calculation, memory and energy consumption resources are consumed evenly, as a result, the reasoning resources aiming at the instant high-risk task are diluted, the response delay of the whole system is increased, a large number of unnecessary low-priority model operations still occupy resources, the technical requirements of engineering supervision on rapid recognition and differentiation of key risks are difficult to meet, and the optimal balance of calculation efficiency and early warning precision supervision cannot be realized. Therefore, how to realize the scheduling reasoning of the edge-side self-adaptive model based on the working condition perception, so that the potential safety hazard early warning efficiency of engineering supervision in a resource-limited environment is a difficult problem faced by the industry. Disclosure of Invention The application provides an engineering supervision potential safety hazard early warning method and system based on edge calculation, which can realize edge side self-adaptive model scheduling reasoning based on working condition perception, thereby improving the potential safety hazard early warning efficiency of engineering supervision in a resource-limited environment. In a first aspect, the application provides an engineering supervision potential safety hazard early warning method based on edge calculation, which comprises the following steps: acquiring multi-source heterogeneous sensing data acquired by a plurality of monitoring points in a target construction area; performing feature fuzzy extraction on the multi-source heterogeneous sensing data to identify current construction working conditions and dominant risk factors, and determining potential safety hazard detection tasks to be executed according to the identified construction working conditions and risk factors; Adaptively selecting a target model corresponding to the potential safety hazard detection task from a plurality of light-weight identification models pre-deployed in edge computing equipment to activate, wherein the light-weight identification models are subjected to differential optimization design aiming at different construction stages or different risk types respectively; In the edge computing equipment, forward reasoning is carried out on the real-time multi-source heterogeneous sensing data of the corresponding monitoring points by utilizing the activated target model, and a potential safety hazard identification result is generated; And judging whether the potential safety hazard exists in the corresponding monitoring point according to the potential safety hazard identification result, generating early warning information when the potential safety hazard exists, and pushing the early warning information to an engineering supervision terminal. Preferably, the multi-source heterogeneous sensing data specifically comprises video stream data acquired through vis