Search

CN-121982905-A - Highway maintenance site AI video perception prevents vehicle intrusion and hidden danger self-checking early warning system

CN121982905ACN 121982905 ACN121982905 ACN 121982905ACN-121982905-A

Abstract

The invention discloses an AI video perception vehicle intrusion and hidden danger self-checking early warning system for a highway maintenance site, and belongs to the technical field of highway maintenance safety prevention and control. The system adopts a layered architecture of a perception layer, a transmission layer, a platform layer and an application layer, the double AI cameras divide work to realize accurate identification of vehicle intrusion, hidden danger and three-violation behaviors, and an acousto-optic early warning device and a double transmission mode are matched to construct an identification-early warning-supervision-rectification-tracing management closed loop. The system can be rapidly deployed in various working vehicles or fixed structures, is suitable for complex outdoor environments, has the advantages of high identification precision, quick response and strong compatibility, is suitable for maintenance scenes such as high-speed, national province trunk lines, urban roads and the like, and can greatly improve the safety management level of maintenance operations.

Inventors

  • MA ZHENBAO

Assignees

  • 甘肃省张掖公路事业发展中心

Dates

Publication Date
20260505
Application Date
20260203

Claims (10)

  1. 1. A highway maintenance site AI video perception vehicle intrusion and hidden danger self-checking early warning system is characterized by adopting a layered architecture design and comprising a perception layer (1), a transmission layer (2), a platform layer (3) and an application layer (4), wherein the perception layer (1) is used for data acquisition and site early warning, the transmission layer (2) is used for data transmission of the perception layer (1) and the platform layer (3), the platform layer (3) is a core control and data processing center, and the application layer (4) provides a visual operation interface to cooperatively realize vehicle intrusion recognition early warning, hidden danger and three-violation behavior self-checking early warning and remote monitoring management functions.
  2. 2. The highway maintenance site AI video perception vehicle intrusion and hidden danger self-checking early warning system of claim 1 is characterized in that the perception layer (1) comprises a first AI camera (11), a second AI camera (12), a controllable stroboscopic alarm lamp (13), a loudspeaker (14), a lifting rod (15) and a fixable clamping seat (16), wherein the first AI camera (11) is arranged on the upstream transition zone side of an operation zone, the resolution is not lower than 1080P, the frame rate is not lower than 30fps, the protection level is not lower than IP67, the second AI camera (12) is arranged on the operation zone side, and the horizontal view angle is not lower than 120 degrees and is internally provided with an infrared night vision module.
  3. 3. The AI video perception vehicle intrusion and hidden danger self-checking early warning system for the highway maintenance site is characterized in that the controllable stroboscopic warning lamp (13) is of a strip-shaped flat shell structure, red LED luminous areas are arranged at two ends of the lamp, a blue LED luminous area is arranged in the middle of the lamp, normal-brightness, stroboscopic and explosion mode switching is supported, the stroboscopic frequency can be adjusted within the range of 5-20Hz, the bottom of the controllable stroboscopic warning lamp (13) is fixed with the mounting seat (17) through a connecting bracket, and the mounting seat (17) is coaxially connected with the top end of the liftable rod (15).
  4. 4. The highway maintenance site AI video perception vehicle intrusion and hidden danger self-checking early warning system is characterized in that two sound amplifying horns (14) are arranged in a truncated cone-shaped structure and respectively correspond to a first AI camera (11) and a second AI camera (12) in linkage, the sound intensity can be adjusted within the range of 60-120dB, a plurality of groups of warning voices are prestored and can be automatically switched and played according to early warning scenes, and the sound amplifying horns (14) and the rear end of a camera shell are integrally injection molded, so that horn mouths face the horizontal outer side.
  5. 5. The AI video perception vehicle intrusion and hidden danger self-checking early warning system for the highway maintenance site is characterized in that the lifting rod (15) is of a hydraulic lifting upright post structure, is made of high-strength aluminum alloy, has a maximum lifting height of 4m and a lifting speed of 0.1-0.2m/s, can be adjusted by remote control or a local button, and is provided with two annular metal reinforcing hoops at the outer side of the lower section of the lifting rod (15), and the bottom end of the lifting rod is connected with a cylindrical hydraulic driving bin.
  6. 6. The AI video perception vehicle intrusion and hidden danger self-checking early warning system for the highway maintenance site is characterized in that the fixable clamping seat (16) comprises two groups of symmetrically distributed metal clamping structures, each group of clamping structures comprises an arc-shaped clamping piece and a locking bolt, anti-skid rubber pads are arranged on the inner sides of the arc-shaped clamping pieces, the two groups of clamping structures are axially distributed along the liftable rod (15) at intervals, the distance between the two groups of clamping structures is 30-50cm, and the clamping structures are matched with columnar carriers with the diameters of 5-12 cm.
  7. 7. The highway maintenance site AI video perception vehicle intrusion and hidden danger self-checking early warning system of claim 1 is characterized in that a transmission layer (2) adopts a double transmission mode of combining a 4G/5G wireless network and a wired network, a 4G/5G wireless router (21) and a gigabit wired network switch (22) are configured, the 4G/5G wireless router (21) supports multi-band communication, the maximum transmission rate is not lower than 1Gbps, the gigabit wired network switch (22) is provided with 4-8 gigabit Ethernet ports, and the transmission process adopts an SSL/TLS encryption protocol and is provided with a data retransmission mechanism.
  8. 8. The highway maintenance site AI video perception vehicle intrusion and hidden danger self-checking early warning system of claim 1 is characterized in that a distributed cloud platform is built by a platform layer (3) through a cloud computing technology and comprises a data storage module (31), a data processing module (32), an algorithm model management module (33) and a system management module (34), wherein the algorithm model management module (33) comprises a vehicle running track and speed analysis algorithm, an operation area hidden danger recognition algorithm and a personnel three-violation behavior recognition algorithm.
  9. 9. The highway maintenance site AI video perception vehicle intrusion and hidden danger self-checking early warning system is characterized in that a YOLO target detection algorithm and a Kalman filtering target tracking algorithm are adopted by the vehicle running track and speed analysis algorithm, a CNN convolutional neural network algorithm is adopted by the operation area hidden danger recognition algorithm, a OpenPose human body posture estimation algorithm is adopted by the personnel 'three violating' behavior recognition algorithm, and an online upgrading function is supported by the algorithm model management module (33).
  10. 10. The system for AI video perception vehicle intrusion and hidden danger self-checking and early warning on highway maintenance sites according to claim 1, wherein the application layer (4) comprises computer software (41) and mobile phone APP (42), supports real-time monitoring, early warning and reminding, video playback, data statistics and analysis, remote control and modification closed loop management functions, and is adaptive to Android and iOS operating systems.

Description

Highway maintenance site AI video perception prevents vehicle intrusion and hidden danger self-checking early warning system Technical Field The invention relates to the technical field of highway maintenance safety prevention and control and intelligent management, in particular to a highway maintenance site vehicle intrusion and hidden danger self-checking early warning system integrating AI video real-time perception, multidimensional early warning and remote collaborative management, which is suitable for various maintenance operation scenes such as expressways, national province trunk roads, urban roads and the like, and can realize the full-flow intelligent supervision of the external vehicle intrusion risk prevention and control and internal hidden danger and three-violation behaviors of an operation area. Background The road is used as a core infrastructure of a traffic transportation system, and the normalized development of maintenance operation is a key for guaranteeing the traffic capacity of the road, prolonging the service life of the road and reducing the occurrence rate of traffic accidents. However, maintenance work sites are typically in an open traffic environment, with the work area adjacent to the traffic lane and with extremely high safety risks. Currently, highway maintenance sites face two major core safety issues: Firstly, the running-in risk of social vehicles is frequently generated when the past social vehicles run into a maintenance operation area due to the fact that the speed of the vehicles is too high, the operation is improper or warning is not timely, and the life safety of operators is directly threatened, so that maintenance equipment is damaged and the operation progress is delayed; Secondly, hidden danger and 'three violations' behavior (violation command, violation operation, violation of labor discipline) in the operation area are not enough in prevention and control, equipment placement is not standard, safety warning signs are missing, operators wear protective articles and violation operation equipment according to regulations, and secondary safety accidents are easy to cause. The traditional maintenance site safety management mainly relies on manual supervision, and has the inherent defects of limited supervision range, poor real-time performance, high labor cost, easy occurrence of human negligence and the like. Most of the existing early warning devices are designed with a single function, such as an independent warning lamp, a common camera and the like, lack of intelligent AI recognition capability, and cannot actively judge the risk of vehicle intrusion and hidden danger of an operation area, and the problems of poor mobility, low recognition precision, delayed early warning response, insufficient compatibility of software and hardware and the like of part of similar systems exist, and closed-loop management of recognition-early warning-supervision-rectification is not formed, so that the dynamic safety management requirement of modern highway maintenance operation is difficult to meet. Along with the rapid development of technologies such as artificial intelligence, the internet of things and cloud computing, intelligent supervision has become a development trend of highway maintenance safety management. In the existing part of related technical schemes, although there are attempts to monitor by adopting a camera and combining a simple algorithm, the following defects generally exist: Firstly, the mobility is poor, most equipment is fixedly installed, and the equipment is difficult to adapt to maintenance operation scenes in different positions and in different scales; secondly, the recognition precision is insufficient, the algorithm model lacks special training aiming at complex environments of maintenance sites, and the misjudgment rate and the missed judgment rate of the vehicle intrusion risk, hidden danger and three-violation behavior are high; thirdly, the early warning mode is single, only single warning is needed by light or sound, the warning effect is limited, and the cooperation of on-site early warning and remote warning is not realized; fourth, the system compatibility and expansibility are insufficient, the software and hardware cooperativity is poor, and the function upgrading and scene adaptation are difficult to carry out according to actual requirements; Fifthly, a complete management closed loop is lacked, only an identification-early warning link can be realized, and subsequent management flows such as hidden danger correction, behavior correction, data statistics analysis and the like are not involved. Therefore, the AI video perception early warning system with high mobility, high precision identification, multidimensional early warning and full flow management capability is developed, the defects of the prior art are overcome, the safety prevention and control level and management efficiency of a highway maintenance site are improved