Video recognition based construction road visual blind area intelligent zebra crossing early warning system

Abstract

A smart zebra crossing early warning system for blind spots in construction access roads, based on video recognition, includes a video data acquisition module, a data analysis module, and an on-site execution module. The video data acquisition module includes one or more high-definition cameras; the data analysis module includes an AI edge computing box; and the on-site execution module includes an integrated controller and an IP network speaker. The integrated controller and AI edge computing box are installed in the office area, and the IP network speaker is fixed to the top of the office entrance. The AI ​​edge computing box is connected to the high-definition cameras via signal matching, and the AI ​​edge computing box is connected to the integrated controller via a wired connection. The integrated controller is equipped with a wireless transmission module that connects to the IP network speaker. This system possesses intelligent perception and early warning capabilities, and can use video recognition technology to monitor the status of people and vehicles in the blind spots of construction access roads in real time, proactively preventing pedestrian-vehicle conflicts and avoiding collisions caused by blind spots or delayed driver reaction.

Description

Technical Field

[0001] This utility model relates to the technical field of intelligent transportation, specifically a video recognition-based intelligent zebra crossing early warning system for blind spots in construction access roads. Background Technology

[0002] During construction, access roads serve as vital passageways for personnel and materials, making their safety paramount. However, due to the complex terrain of these access roads, numerous blind spots often exist at office area entrances, such as sharp turns, steep slopes, and areas obscured by obstacles. Within these blind spots, pedestrians and construction vehicles (such as forklifts and dump trucks) cannot detect each other in time, greatly increasing the risk of collisions.

[0003] Traditionally, static zebra crossings are installed at the entrances and exits of office areas to guide pedestrian traffic. However, this method has many drawbacks. On the one hand, it fails to provide warnings in blind spots and lacks dynamic perception and proactive intervention capabilities. On the other hand, construction environments are complex and changeable, and traditional zebra crossings suffer from poor adaptability, inability to prevent dangerous situations in real time, and low traffic efficiency (such as pedestrians "testing the waters" causing vehicles to repeatedly start and stop). Therefore, a smart zebra crossing early warning system for blind spots in construction access roads needs to be designed. Summary of the Invention

[0004] The purpose of this invention is to address the problems existing in the prior art by providing a video recognition-based intelligent zebra crossing warning system for blind spots in construction access roads. This system possesses intelligent perception and warning capabilities, enabling it to monitor the status of people and vehicles in blind spots of construction access roads in real time using video recognition technology. It proactively prevents pedestrian-vehicle conflicts, promptly blocks dangerous scenarios, and avoids collisions caused by blind spots or delayed driver reactions. The system is highly adaptable and can meet the needs of different construction scenarios. It features low-cost upgrades, easy turnover, and no pollution. By optimizing traffic efficiency and preventing pedestrians from "trial crossings" that cause vehicles to repeatedly start and stop, it greatly improves traffic safety and flow in construction access road areas.

[0005] To achieve the above objectives, the technical solution adopted by this utility model is as follows:

[0006] A smart zebra crossing early warning system for blind spots in construction access roads based on video recognition includes a video data acquisition module, a data analysis module, and an on-site execution module. The video data acquisition module includes one or more high-definition cameras for monitoring the entrance, sharp turns, and steep slopes of the construction access road's blind spots; these cameras can be bolted to the side wall of the office area. The data analysis module includes an AI edge computing box, which deploys AI recognition algorithms and intelligent early warning algorithms. The AI ​​recognition algorithms include algorithms for recognizing the front and rear of construction vehicles and algorithms for recognizing left and right directions after vehicles intrude into the electronic fence. The on-site execution module includes... The system includes an integrated controller and an IP network speaker. The integrated controller has a built-in backup power supply. The integrated controller and the AI ​​edge computing box are installed in the office area. The integrated controller can be fixed to the side wall of the office area with bolts, and the IP network speaker is fixed to the top of the office area entrance. The AI ​​edge computing box and the high-definition camera are connected via signal matching to complete video material acquisition and AI algorithm training. The AI ​​edge computing box and the integrated controller are connected by a wired connection to generate trigger signals and transmit them to the integrated controller. The integrated controller is equipped with a wireless transmission module that connects to the IP network speaker. The integrated controller controls the IP network speaker to start the voice program and provide voice warnings.

[0007] A further preferred embodiment: the on-site execution module also includes a projector. The integrated controller is connected to the projector via a wired connection. The projector is installed in a suitable location in the office area, and projects dynamic zebra crossings onto the pedestrian walkway area. The integrated controller controls the projector after receiving a signal from the AI ​​edge computing box.

[0008] A further preferred embodiment: the field execution module also includes a multi-frequency flashlight, which is bolted to the top of the office area entrance. The integrated controller controls the multi-frequency flashlight after receiving a signal from the AI ​​edge computing box.

[0009] This intelligent zebra crossing early warning system for blind spots in construction access roads, based on video recognition, possesses intelligent perception and early warning capabilities, monitoring the dynamics of people and vehicles in blind spots of construction access roads in real time. Its highly adaptable design caters to the needs of various complex construction scenarios. After processing system data, it provides intelligent alerts. When a vehicle enters the section of road ahead of the office area and touches the designated electronic fence, the IP network speaker activates a voice program, repeatedly playing a voice prompt that construction vehicles are approaching and passage is prohibited. The projector executes a program to switch the displayed content, changing the ground projection pattern from a zebra crossing style to a "No Entry" sign. The system also controls the activation of multi-frequency flashing lights in a high-frequency flashing mode, using alternating red and blue strong light to warn pedestrians. Compared to traditional zebra crossings, this system can be upgraded at low cost, is reusable, and is environmentally friendly and pollution-free. It also reduces labor costs, saves time and effort, and is safer. Attached Figure Description

[0010] Figure 1 This is a schematic diagram of the three-dimensional layout structure of this system;

[0011] Figure 2 This is a top view of the system's layout structure;

[0012] The names corresponding to the serial numbers in the figure are:

[0013] 1. Construction access road; 2. Office area; 3. High-definition camera; 4. Projector; 5. IP network speaker; 6. Multi-frequency flashlight; 7. Integrated control module; 8. AI edge computing box. Detailed Implementation

[0014] To provide a more detailed description of this utility model, the following description, in conjunction with the embodiments and accompanying drawings, will further illustrate this utility model. Example

[0015] A smart zebra crossing early warning system for blind spots in construction access roads based on video recognition includes a video data acquisition module, a data analysis module, and an on-site execution module. The video data acquisition module includes one or more high-definition cameras 3 for monitoring the entrance, sharp turns, and steep slopes of the blind spots in the construction access road 1. The data analysis module includes an AI edge computing box 8, which is equipped with AI recognition algorithms and intelligent early warning algorithms. The AI ​​recognition algorithms include algorithms for recognizing the front and rear of construction vehicles and algorithms for recognizing left and right directions after vehicles intrude into the electronic fence. The on-site execution module includes an integrated controller 7 and an IP network speaker 5. The integrated controller 7 and the AI ​​edge computing box 8 are installed in the office area 2. The IP network speaker 5 is fixed at the top of the office area entrance. The AI ​​edge computing box 8 and the high-definition cameras 3 are connected via signal matching to complete video material acquisition and AI algorithm training. The AI ​​edge computing box 8 and the integrated controller 7 are connected via a wired connection to generate trigger signals and transmit them to the integrated controller 7. The integrated controller 7 is equipped with a wireless transmission module connected to the IP network speaker 5 to control the IP network speaker 5 to start a voice program and provide voice warnings.

[0016] The on-site execution module also includes a projector 4. The integrated controller 7 is connected to the projector 4 via a wired connection. The projector 4 is installed at a suitable position on the top of the office area 2, and projects a dynamic zebra crossing onto the pedestrian walkway area. When the integrated controller 7 receives a signal transmitted from the AI ​​edge computing box 8, it controls the projector 4.

[0017] The field execution module also includes a multi-frequency flashlight 6, which is bolted to the top of the entrance to the office area 2. The integrated controller 7 controls the multi-frequency flashlight 6 after receiving a signal from the AI ​​edge computing box 8.

[0018] The application principle of this intelligent zebra crossing early warning system for blind spots in construction access roads based on video recognition is as follows: The AI ​​edge computing box 8 collects video monitoring materials of construction vehicles passing through sharp turns and steep slopes in front of and behind office area 2 through high-definition cameras 3. After converting the materials into image materials, the front and rear of the construction vehicles are labeled. The labeled image materials are used to train and optimize the AI ​​algorithm. Then, the trained AI algorithm is deployed to the AI ​​edge computing box 8, and electronic fences are set up at a safe distance from the sharp turns and steep slopes in front of office area 2 to the section of road directly in front of office area 2, based on the field of view of each camera. When the AI ​​edge computing box 8 detects, through its trained AI algorithm, that the front of a construction vehicle touches a pre-set electronic fence on one end of a camera, it immediately generates a trigger signal and transmits it to the integrated controller 7. Upon receiving the signal, the integrated controller 7 controls the IP network speaker 5 to start a voice program, continuously playing a voice prompt that a construction vehicle is approaching the construction access road and passage is prohibited; it controls the projector 4 to execute a display content switching program, changing the ground projection pattern from a zebra crossing style to a no-entry sign; and it controls the multi-frequency flashlight 6 to start a high-frequency flashing mode, using alternating red and blue strong light to warn passing personnel. When the AI ​​edge computing box 8 detects, through its trained AI algorithm, that the front of a construction vehicle touches a pre-set electronic fence on the other end of a camera, it immediately generates a trigger signal and transmits it to the integrated controller 7. Upon receiving the signal, the integrated controller 7 controls the IP network speaker 5 to stop the voice program; it controls the projector 4 to execute a display content switching program, changing the ground projection pattern from a no-entry sign to a zebra crossing style; and it controls the multi-frequency flashlight 6 to stop the high-frequency flashing mode.

[0019] Implementation Case:

[0020] The construction access road at a certain construction site is 6 meters wide and features a two-lane, two-way design. The office area is located on one side of the access road. Above the office area is a sharp bend in the access road, and below is a steep slope, creating blind spots when entering or exiting the office area. A smart zebra crossing warning system for blind spots in the construction access road has been installed. Its operation steps are as follows:

[0021] 1. System Deployment: Based on the terrain of the construction site and the visibility of the office area, high-definition cameras are installed on the upper and lower sides of the office area, facing the sharp turns and steep slopes of the construction access road above and below the office area; a projector is installed at the top center of the office area to ensure that the projected dynamic zebra crossings and no-entry signs are clearly displayed on the construction access road in front of the office area; IP network speakers and multi-frequency flashlights are evenly distributed at the entrance of the office area and around the construction access road to ensure that the warning signals can effectively cover the entire blind spot; an integrated controller is installed at the entrance of the office area for easy management and maintenance, and is connected to the AI ​​edge computing box, projector, and multi-frequency flashlights via wired connection, and to the IP network speakers via wireless signal to achieve collaborative work between the modules; the AI ​​edge computing box is installed inside the office area and establishes a stable data transmission connection with the video data acquisition module wirelessly.

[0022] 2. Video footage collection, AI algorithm training, and core logic: Collect video surveillance footage of construction vehicles such as concrete mixer trucks, excavators, and dump trucks passing through sharp turns and steep slopes in front of and behind the office area. After converting the footage into image footage, label the front and rear of the vehicles. Use the labeled image footage to train and optimize the AI ​​algorithm. Then, deploy the trained AI algorithm to an AI edge computing box and set up electronic fences at a distance of 10 meters from the sharp turns and steep slopes in front of and behind the office area.

[0023] The core logic of AI algorithms can be divided into three main stages: target recognition and analysis, event verification and confirmation, and alarm signal response.

[0024] The first phase involves target recognition and analysis, including image standardization, core AI engine analysis, and preliminary result refinement. Preliminary structural refinement includes confidence threshold filtering and redundancy detection merging. Image standardization involves frame-by-frame parsing and standardization of the raw video stream captured by the high-definition camera. Without distorting the image content or maintaining the original aspect ratio, the images are uniformly scaled and adjusted to the standard size W×H (640×640) required by the AI ​​analysis engine through center cropping. Core AI engine analysis involves feeding the standardized images into a deep learning neural network model (AI engine) for core analysis. This enables rapid and accurate scanning, positioning, and identification of preset engineering vehicle targets in images. Confidence threshold filtering ensures that when the confidence level assigned to each identified engineering vehicle target by the system is higher than the preset recognition confidence threshold (0.6), the target is considered to be effectively identified. Redundant detection merging automatically starts and optimizes the program to analyze these overlapping detection results after the AI ​​engine generates multiple highly overlapping detection boxes for the same engineering vehicle target. When the overlap of multiple detection boxes is greater than the set non-maximum suppression threshold (0.45), the single detection box with the highest confidence level is retained as the final identification result.

[0025] Phase Two:

[0026] The event verification and confirmation include spatial rule verification, time persistence verification, event recording and cooling-off. Spatial rule verification involves setting up electronic fences in the monitoring screens of the upper and lower cameras respectively. When an engineering vehicle touches the electronic fence set by the upper camera and enters the key monitoring area, the subsequent time persistence verification logic is triggered. Time persistence verification is determined to be a real and valid event when the same engineering vehicle is continuously and uninterruptedly observed in the key monitoring area for a specified number of video frames, which is greater than the set number of consecutive frame confirmations N (15). Event recording and cooling-off is that once an action passes all the above verifications, the system will officially record it as an alarm event (store it in the database, generate a snapshot with annotations, etc.). At the same time, in order to avoid repeatedly recording the same continuous event, the system will enter an alarm recording cooling-off period (15s) for the event type. During the cooling-off period, even if the action continues, the system will not generate a new alarm record, ensuring the simplicity and efficiency of the event log.

[0027] Phase Three:

[0028] Alarm signal response includes continuous status confirmation and an alarm silence period. Continuous status confirmation involves the system performing a high-level continuous assessment before signal output. If the proportion of engineering vehicle frames detected within the continuous confirmation time window (3s) exceeds the set alarm trigger ratio (0.5), the system determines it as a stable and continuous event and outputs an alarm signal. The alarm silence period is a period after a complete round of signal output. To prevent "alarm fatigue" and provide a window for on-site handling, the alarm algorithm enters a silence period (or "refractory period"). During this period, even if the AI ​​algorithm still detects events that meet the conditions, it will not output a signal until the silence period ends.

[0029] 3. Intelligent early warning algorithm:

[0030] Intelligent early warning algorithm for the construction access road section in front of the office area: When the AI ​​edge computing box detects, through its trained AI algorithm, that the front of a construction vehicle is touching the electronic fence preset by the camera above, it immediately generates a trigger signal and transmits it to the integrated controller. Upon receiving the signal, the integrated controller controls the IP network speaker to activate a voice prompt, playing three times at a set volume (85dB) and interval (2s) stating that a construction vehicle is approaching from above and passage is prohibited. When the algorithm detects that the front of a construction vehicle below touches the electronic fence preset by the camera below, it immediately generates a trigger signal and transmits it to the integrated controller. After receiving the signal, the integrated controller controls the IP network speaker to start a voice prompt, playing three times in a loop at a set volume (85db) and interval (2s) that a construction vehicle is approaching from below and passage is prohibited; it controls the projector to execute a display content switching program, changing the ground projection pattern from a zebra crossing style to a no-pass sign; and it controls the multi-frequency flashlight to start a high-frequency flashing mode, using alternating red and blue strong light to warn passing personnel.

[0031] Intelligent early warning algorithm for construction access road section ahead of the office area: When the AI ​​edge computing box identifies, through the trained AI algorithm, that the front of an overhead construction vehicle touches the electronic fence preset by the camera below, or vice versa, it immediately generates a trigger signal and transmits it to the integrated controller. Upon receiving the signal, the integrated controller controls the IP network speaker to mute the voice program; controls the projector to execute the display content switching program, changing the ground projection pattern from a no-entry sign to a zebra crossing pattern; and controls the activation of the multi-frequency flashlight to deactivate the high-frequency flashing mode.

[0032] Intelligent early warning algorithm for power outage or communication interruption: When power outage or communication interruption occurs, the integrated controller controls the IP network speaker to start a voice program to prompt the device malfunction in a loop at a set volume (85db) and interval (2s); controls the projector to execute a display content switching program, switching the ground projection pattern from a zebra crossing style to a no-entry sign; controls the multi-frequency flashlight to start a high-frequency flashing mode, using alternating red and blue strong light to warn passing personnel.

[0033] Intelligent early warning algorithm for power or communication restoration: When power or communication is restored, the integrated controller controls the IP network speaker to shut down the voice program; controls the projector to execute the display content switching program, switching the ground projection pattern from a no-entry sign to a zebra crossing pattern; and controls the activation of the multi-frequency flashlight to disable the high-frequency flashing mode.

[0034] The above description is not intended to limit the present utility model, nor is the present utility model limited to the above examples. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the present utility model should be protected by the present utility model.

Claims

1. A smart zebra crossing early warning system for blind spots in construction access roads based on video recognition, characterized in that: The system includes a video data acquisition module, a data parsing module, and an on-site execution module. The video data acquisition module includes one or more high-definition cameras (3) for monitoring the blind spots of the construction access road (1), sharp turns, and steep slopes. The data parsing module includes an AI edge computing box (8). The on-site execution module includes an integrated controller (7) and an IP network speaker (5). The integrated controller (7) and the AI ​​edge computing box (8) are installed in the office area (2). The IP network speaker (5) is fixed at the top of the office area entrance. The AI ​​edge computing box (8) and the high-definition camera (3) are connected by signal matching to complete video material acquisition and AI algorithm training. The AI ​​edge computing box (8) and the integrated controller (7) are connected by a wire to generate trigger signals and transmit them to the integrated controller (7). The integrated controller (7) is equipped with a wireless transmission module that connects to the IP network speaker (5). The integrated controller (7) controls the IP network speaker (5) to work.

2. The intelligent zebra crossing early warning system for blind spots in construction access roads based on video recognition as described in claim 1, characterized in that: The on-site execution module also includes a projector (4). The integrated controller (7) is connected to the projector (4) via a wired connection. The projector (4) is installed in a suitable location in the office area (2). The projector (4) projects the dynamic zebra crossing onto the pedestrian access area of ​​the construction access road.

3. The intelligent zebra crossing early warning system for blind spots in construction access roads based on video recognition as described in claim 1 or 2, characterized in that: The field execution module also includes a multi-frequency flashlight (6), which is fixed to the top of the entrance of the office area (2) by bolts.

Images