Mining area operation vehicle anti-collision early warning system and early warning method based on partition early warning

Abstract

The invention provides a mining area operation vehicle anti-collision early warning system based on partition early warning. The system comprises a sensing layer, a processing layer and an application layer. The sensing layer comprises an AI camera module, a model processing module, an ultrasonic sensor and a point cloud data processing module; the processing layer comprises a cooperative processing module, an early warning analysis module and a state management module which are connected in sequence; the application layer comprises a visual display module, a sound-light alarm module and a data recording module; the invention further provides an early warning method of the system. By means of the system, the mining area vehicle can adapt to the complex environment of the mining area, the multi-level safety protection capacity is achieved, and all-around safety early warning protection can be conducted on mining area vehicle operation.

Description

Technical Field

[0001] The present invention relates to an anti-collision warning system and warning method for mining operation vehicles based on zonal warning, belonging to the technical field of intelligent safety warning. Background Art

[0002] Nowadays, mineral resources are an important foundation for industrial development. During the operation process in mining areas and underground, the safety protection problem of mining vehicles has become increasingly prominent. Traditional mining vehicle operations mainly rely on manual observation and basic video monitoring systems. This method has problems such as many monitoring blind spots, poor real-time performance, and great potential safety hazards in practical applications. Existing video monitoring technologies mainly rely on embedded technologies and use network and communication technologies as platforms. Basic image analysis algorithms are embedded in monitoring cameras to form a digital and networked monitoring system, which has improved the automation level of safety monitoring to a certain extent. However, in a complex mining area environment, especially in harsh conditions such as insufficient light and large dust, the reliability and real-time performance of such basic visual monitoring systems are difficult to guarantee.

[0003] In recent years, with the rapid development of computer vision and artificial intelligence technologies, object detection methods based on deep learning have shown great application potential in the field of safety warning due to their powerful feature extraction and object recognition capabilities. At the same time, ultrasonic ranging technology has unique advantages in detecting obstacles at close range due to its stable and reliable distance measurement characteristics. However, existing safety protection systems often adopt a single sensor solution and lack a scientific and reasonable warning zoning strategy, and cannot provide multiple levels of protection measures according to the safety requirements of different distance ranges of operation vehicles. In practical applications, such a single protection solution often has problems such as untimely warning and high false alarm rate, and cannot effectively guarantee operation safety, and at the same time poses a potential safety hazard to the safety of operation workers.

[0004] Therefore, there is an urgent need for a warning system that can adapt to the complex mining area environment and has multi-level safety protection capabilities. Through the complementary advantages of AI visual detection and ultrasonic ranging technologies, combined with a scientific and reasonable zonal warning strategy, it can achieve comprehensive safety warning protection for mining vehicle operations to overcome the above problems. Summary of the Invention

[0005] The purpose of the present invention is to overcome the deficiencies of the prior art and provide an anti-collision warning system and warning method for mining operation vehicles based on zonal warning, which can adapt to the complex mining area environment and has multi-level safety protection capabilities. Through the complementary advantages of AI visual detection and ultrasonic ranging technologies, combined with a scientific and reasonable zonal warning strategy, it can achieve comprehensive safety warning protection for mining vehicle operations.

[0006] The present invention is implemented as follows: The present invention provides an anti-collision warning system for mining operation vehicles based on zonal warning, which system includes a sensing layer, a processing layer and an application layer; The sensing layer includes an AI camera module, a model processing module, ultrasonic sensors, and a point cloud data processing module; The processing layer includes a collaborative processing module, a warning analysis module, and a status management module that are connected in sequence; The application layer includes a visualization display module, an audible and visual alarm module, and a data recording module; Among them, the AI camera module transmits the video information detected in the far field to the model processing module, and the model processing module is connected to the coordination processing module; the ultrasonic sensors transmit the data detected in the near field to the point cloud data processing module, and the point cloud data processing module is connected to the coordination processing module; the status management module is respectively connected to the visualization display module, the audible and visual alarm module, and the data recording module.

[0007] The present invention also provides an anti-collision warning method for mining operation vehicles based on zonal warning, including the following steps: Step 1, deploy sensors and collect data: Install an AI camera module and ultrasonic sensors on the mining operation vehicle, and use the AI camera module to perform video processing on the targets at the front, rear and both sides of the vehicle; use the ultrasonic sensors to detect the close-range obstacles at the front and rear of the vehicle; Step 2, data processing and analysis: including AI vision processing, dual-stream processing, ultrasonic data processing and multi-sensor data processing. Among them, the multi-sensor data processing is to perform collaborative processing on the AI vision detection results and the ultrasonic module distance data to achieve hierarchical anti-collision warning; Step 3, warning area division and determination: Define the first-level warning area as a high-risk area within 0 - 2.5 meters from the vehicle, which is monitored by the ultrasonic sensors; define the second-level warning area as a warning area within 2.5 - 7.5 meters from the vehicle, which is processed by the AI camera module; Step 4, the warning decision-making layer makes an intelligent judgment: When the target enters the first-level warning area, immediately trigger an emergency warning signal, highlight and display the position of the dangerous target on the warning display, emit an audible and visual alarm prompt, and record the warning event information; when the target enters the second-level warning area, emit a warning prompt signal, change the detection frame of the target from green to highlighted red on the display interface, update the target dynamic information in real time, and record the change of the warning status; Step 5, output warning information: Display the warning information through the visualization display module, the audible and visual alarm module, and the data recording module.

[0008] Further, in Step 1, 4 AI cameras are arranged at the front of the mining area operation vehicle, and 2 AI cameras are arranged at the rear. Their horizontal field of view angles are both ≥100°, respectively responsible for target detection and early warning of the front, rear, left and right sides of the vehicle; 2 ultrasonic sensors are arranged at the front and rear of the vehicle, and their horizontal field of view angles ≥60°, responsible for short-range obstacle perception; Among them, the AI camera module adopts an edge computing platform equipped with an NPU embedded processing unit, with local inference capabilities. It includes a high-definition image sensor and a large-aperture lens, an NPU neural network processor is built in, and an infrared fill light module is integrated.

[0009] Further, in Step 2, the specific process of AI vision processing is as follows: perform size normalization and format conversion on the collected video stream; input the preprocessed image into the vision target detection model for inference to generate target position, category, and confidence detection results; The dual-stream processing process is as follows: The main stream is mainly used for real-time display of early warning results. Its resolution is the original acquisition resolution, responsible for visual drawing of target frames and warning areas, used for real-time display of the results of AI vision detection, and real-time streaming display through the RTSP protocol; The sub-stream is mainly dedicated to the inference analysis of the AI model. The resolution is optimized and scaled to fit the size of the AI vision model, only used for target detection, generating model detection result data, and no display processing is performed; The ultrasonic data processing process is: real-time processing of the distance information returned by the ultrasonic sensor, calculating the specific position of the obstacle, and performing noise filtering and data smoothing processing.

[0010] Further, in Step 3, The first-level warning area is monitored and processed by ultrasonic sensors for obstacles within the defined range. The specific calculation is as follows: ; Among them, D is the distance of the obstacle, v is the speed of sound wave propagation in the air, and t is the time interval from transmitting the signal to receiving the echo; The determination of whether there is a target in the second-level warning area adopts the ray method. The specific steps are as follows: Obtain the center point coordinates of the target detection frame based on the target detection algorithm of deep learning: ; Draw a ray equation horizontally to the right from this point: ; The equation of the boundary line segment of the warning area polygon. Assuming there are n sides, for the i-th boundary line segment, the two endpoints are respectively and The line segment equation: , < < ; For each boundary line segment, the x - coordinate of the intersection point of the ray and the boundary line segment satisfies the following equation: ; When ≤ ≤ and ≥ this intersection point is valid; The ray - method determination rule is as follows: if the number of intersection points is even, including 0, the point is outside; if the number of intersection points is odd, the point is inside: ; Determine whether the target is in the early - warning area according to the parity of the number of intersection points. When it is detected that the target enters the early - warning area, immediately trigger the corresponding - level early - warning response.

[0011] Furthermore, in step five, the specific process of the visualization display module is: real - time display the position of the detected target on the early - warning display, display the status of the early - warning area, and directly perform image annotation and display based on the main code stream.

[0012] Furthermore, in step five, the specific process of the acoustic - optical alarm module is: trigger the corresponding acoustic - optical warning signals according to different early - warning levels, and support warning methods with different intensities and frequencies.

[0013] Furthermore, in step five, the specific process of the data recording module is: automatically record early - warning events and system status information, support data playback and analysis functions, and generate early - warning statistical reports.

[0014] The present invention has the following beneficial effects: The anti - collision early - warning system and method for mining operation vehicles based on zoned early - warning provided by the present invention have the following beneficial effects compared with the prior art: 1. The reliability advantage of multi - sensor collaboration: The present invention adopts a scheme combining an AI camera and an ultrasonic sensor to form a near - and - far - field collaborative monitoring mechanism; the ultrasonic sensor is responsible for accurate distance measurement in the near - field (within 2.5 meters), with high reliability; the AI camera is responsible for target detection and early - warning in the far - field (2.5 - 7.5 meters), with a large coverage range; the dual - guarantee mechanism significantly improves the detection reliability of the system in a complex mining area environment.

[0015] 2. The real - time improvement brought by edge computing: The present invention deploys the AI algorithm on the front - end device for local inference, without cloud transmission, reducing network latency and ensuring real - time performance; the embedded processing unit platform equipped with an NPU accelerates to ensure the deployment optimization of the AI vision model, improving the detection speed; significantly reducing the system response time and improving the real - time performance of early - warning.

[0016] 3. Scientific nature of zoning warning strategy: The present invention adopts a multi-level warning mechanism based on distance, scientifically divides the warning area, and avoids false alarms and missed alarms; different detection methods and warning levels are used in different areas, which improves the accuracy and practicality of the system.

[0017] 4. High efficiency of dual-stream processing: The present invention innovatively adopts a dual-stream processing mechanism to separate detection and display functions, thereby improving processing efficiency; visual drawing is performed directly on the original image data, reducing format conversion overhead, optimizing system resource utilization, and improving overall performance.

[0018] 5. Advanced engineering implementation: The present invention has significant advantages in practical engineering applications. The AI ​​camera adopts an integrated design with a compact size, which is easy to install and deploy. It also has an infrared fill light function and can adapt to various lighting environments. The AI ​​camera surgery adopts a protective design to meet the harsh environment requirements of the mining area and supports localized storage and information backtracking functions.

[0019] 6. Significant application value: In practical applications, the present invention can significantly improve the safety of mining operations, reduce manpower input, and lower operating costs; provide early warning data analysis, assist in safety management, and has good promotion and application value.

[0020] The present invention organically combines advanced technologies such as artificial intelligence, multi-sensor fusion and edge computing to provide a complete solution for vehicle safety warning in mining areas, which has important practical value and promotion significance. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0022] Figure 1 A schematic diagram of an anti-collision warning architecture provided by an embodiment of the present invention; Figure 2 A schematic diagram of the AI ​​camera module provided by the present invention; Figure 3 A schematic diagram of the installation of the sensor provided by the present invention; Figure 4 A hierarchical architecture diagram of the anti-collision warning system for mining area operation vehicles based on zone warning provided by the present invention; Figure 5 A flow chart of the warning method of the anti-collision warning system for mining area operation vehicles based on zone warning provided by the present invention. Specific embodiments

[0023] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0024] As Figures 1 to 5 shown, the present invention provides an anti-collision warning system for mining operation vehicles based on zone warning, and the system includes a sensing layer, a processing layer, and an application layer; The sensing layer includes an AI camera module, a model processing module, ultrasonic sensors, and a point cloud data processing module; The processing layer includes a collaborative processing module, a warning analysis module, and a status management module that are connected in sequence; The application layer includes a visualization display module, an audible and visual alarm module, and a data recording module; Among them, the AI camera module transmits the video information detected in the far field to the model processing module, and the model processing module is connected to the coordination processing module; the ultrasonic sensors transmit the data detected in the near field to the point cloud data processing module, and the point cloud data processing module is connected to the coordination processing module; the status management module is respectively connected to the visualization display module, the audible and visual alarm module, and the data recording module.

[0025] As Figures 1 to 5 shown, the present invention also provides an anti-collision warning method for mining operation vehicles based on zone warning, including the following steps: Step 1: Deploy sensors and collect data: Install an AI camera module and ultrasonic sensors on the mining operation vehicle, and use the AI camera module to perform camera processing on the targets at the front, rear, and both sides of the vehicle; use the ultrasonic sensors to detect the close-range obstacles at the front and rear of the vehicle. Specifically, 4 AI cameras are arranged at the front of the mining operation vehicle, and 2 AI cameras are arranged at the rear of the vehicle, and their horizontal field of view angles are all ≥100°, respectively responsible for target detection and warning at the front, rear, and both sides of the vehicle; 2 ultrasonic sensors are arranged at the front and rear of the vehicle, and their horizontal field of view angles ≥60°, responsible for perceiving close-range obstacles; Among them, the AI camera module adopts an edge computing platform equipped with an NPU embedded processing unit, with local inference capabilities. The AI camera module, as shown in Figure 2, adopts an edge computing platform (RKNN model processing) equipped with an NPU embedded processing unit, with local inference capabilities, mainly including: 1. High-definition image sensor and large-aperture lens; 2. Built-in NPU neural network processor, supporting local real-time inference; 3. Integrated infrared fill light module, automatically compensating for light, adapting to harsh environments such as low light or even no light; 4. Adopting a reinforced metal shell with an IP65 protection level to meet the protection requirements of the mining area. The data collected by all sensors are preprocessed and stored through the data acquisition layer (perception layer).

[0026] Step 2: Data processing and analysis: including AI vision processing, dual-stream processing, ultrasonic data processing, and multi-sensor data processing. Among them, the specific process of AI vision processing is as follows: perform size normalization and format conversion on the collected video stream; input the preprocessed image into the vision target detection model for inference to generate target position, category, and confidence detection results; The process of dual-stream processing is as follows: The main stream is mainly used for real-time display of the warning results, with its resolution being the original acquisition resolution, responsible for visualizing the drawing of target boxes and warning areas, used for real-time display of the results of AI vision detection, and performing real-time streaming display through the RTSP protocol; The sub-stream is mainly dedicated to the inference analysis of the AI model, with its resolution optimized and scaled to match the size of the AI vision model, only used for target detection, generating model detection result data, and not performing display processing; The process of ultrasonic data processing is: real-time processing of the distance information returned by the ultrasonic sensor, calculating the specific position of the obstacle, and performing noise filtering and data smoothing processing. Multi-sensor data processing is to perform collaborative processing on the AI vision detection results and the ultrasonic distance data to achieve hierarchical anti-collision warning; Multi-sensor data processing: Perform collaborative processing on the AI vision detection results and the ultrasonic distance data to achieve the hierarchical anti-collision warning function, ensuring the real-time and reliability of the data.

[0027] Step 3: Warning area division and determination: Define the first-level warning area as a high-risk area within 0 - 2.5 meters from the vehicle, monitored by the ultrasonic sensor; The first-level warning area is monitored and processed by the ultrasonic sensor for obstacles within the defined range. The specific calculation is as follows: ; Among them, D is the distance of the obstacle, v is the speed of sound wave propagation in the air, and t is the time interval from transmitting the signal to receiving the echo.

[0028] Define the secondary warning area as a warning area 2.5 - 7.5 meters away from the vehicle, which is processed by the AI camera module; first, divide the polygon warning area (secondary warning area) at the image level. The determination of whether there is a target in the secondary warning area adopts the ray method, and the specific steps are as follows: 1. Obtain the center point coordinates of the target detection frame based on the object detection algorithm of deep learning: ; 2. Draw a ray equation horizontally to the right from this point: ; 3. For the boundary line segment equation of the warning area polygon, assuming there are n sides, for the i-th boundary line segment, the two endpoints are respectively and The line segment equation: , < < ; 4. For each boundary line segment, the x coordinate of the intersection point of the ray and the boundary line segment satisfies the following equation: ; When ≤ ≤ and ≥ , this intersection point is valid; 5. The ray method determination rule is as follows: the number of intersection points is even, including 0, the point is outside; the number of intersection points is odd, the point is inside: ; Among them, N%2 represents the modulo operation on the number of intersection points N (that is, calculate the remainder of N divided by 2): When N%2 = 1, that is, the number of intersection points is odd, it is determined that the target point is inside the warning area; When N%2 = 0, that is, the number of intersection points is 0 or even, it is determined that the target point is outside the warning area.

[0029] Determine whether the target is in the warning area according to the parity of the number of intersection points. When it is detected that the target enters the warning area, immediately trigger the corresponding level of warning response.

[0030] Step 4. The warning decision layer makes an intelligent judgment: when the target enters the first-level warning area, immediately trigger an emergency warning signal, highlight and display the position of the dangerous target on the warning display, emit an audible and visual alarm prompt, and record the warning event information; when the target enters the secondary warning area, emit a warning prompt signal, change the target detection frame from green to highlighted red on the display interface, update the target dynamic information in real time, and record the change of the warning status; Step 5, Output warning information: Display the warning information through the visual display module, the audible and visual alarm module, and the data recording module.

[0031] The specific process of the visual display module is as follows: The position of the detection target is displayed in real time on the warning display, the status of the warning area is displayed, and image annotation and display are directly performed based on the main stream.

[0032] The specific process of the audible and visual alarm module is as follows: Corresponding audible and visual warning signals are triggered according to different warning levels, and warning methods with different intensities and frequencies are supported.

[0033] The specific process of the data recording module is as follows: Automatically record warning events and system status information, support data playback and analysis functions, and generate a warning statistical report.

[0034] The following will describe in detail the best implementation manners of the anti-collision warning system and method for mining area operation vehicles of the present invention with reference to the accompanying drawings: Embodiment 1: Implementation manner of the basic anti-collision warning system As Figure 1 Shown in the anti-collision warning architecture diagram, the anti-collision warning device in this embodiment mainly consists of an AI camera, an ultrasonic radar, and a warning controller. Among them: a) Composition of the AI camera module (as Figure 2 Shown in the composition of the AI camera module): The infrared fill light is used for compensation lighting in environments with insufficient light. The lens and the image acquisition module are responsible for real-time image acquisition. The AI calculation module is integrated on the image acquisition main board and is responsible for target detection and recognition; The specific parameters of the AI camera are as follows: It uses a Rockchip RV1109 / RV1126 processor, is equipped with 2GB DDR4 memory and 32GB eMMC storage, uses a SONY IMX307 image sensor, supports low-light image acquisition, video encoding supports H264 / H265, can output 1080P@30fps, the operating temperature range is -30°C to 65°C, and the protection level is IP67.

[0035] b) Composition of the ultrasonic radar module: It consists of 1 ultrasonic radar control box and 4 ultrasonic probes. The working parameters of the probes are: ranging range of 3 - 450 cm, detection angle of about 60°, accuracy of ±1 cm, and RS485 output is supported.

[0036] Embodiment 2: System installation and deployment method As Figure 3 Shown in the sensor installation schematic diagram, the following installation scheme is adopted in this embodiment: a) Sensor layout: Install 2 AI cameras (horizontal field of view angle ≥ 110°) and 2 ultrasonic sensors (horizontal field of view angle ≥ 60°) at the front of the vehicle, install 1 AI camera on each side of the vehicle body, and install 2 AI cameras and 2 ultrasonic sensors at the rear of the vehicle.

[0037] b) Early warning area division: Level 1 early warning area (red area): within 0 - 2.5 meters from the vehicle; Level 2 early warning area (orange area): within 2.5 - 7.5 meters from the vehicle.

[0038] Embodiment 3: Implementation method of data processing This embodiment adopts a dual - stream processing mechanism: a) Main stream processing: Maintain the original acquisition resolution for real - time display and warning visualization, and push the stream for display through the RTSP protocol.

[0039] b) Sub - stream processing: Reduce the resolution to adapt to the AI model, which is specifically used for target detection and analysis to generate model detection data.

[0040] Embodiment 4: Implementation method of warning response This embodiment executes corresponding responses according to different warning levels: a) Level 1 warning response: Triggered when the ultrasonic sensor detects that the distance to the obstacle is < 2.5 meters, the display interface highlights the dangerous target and emits a high - frequency sound and light alarm.

[0041] b) Level 2 warning response: Triggered when the AI detects that the target distance is within 2.5 - 7.5 meters, marks the detection box as red on the display interface and emits a warning prompt sound.

[0042] Through the organic combination of the above - mentioned implementation methods, the present invention can achieve all - round safety protection for mining area vehicles and achieve good results in practical applications.

[0043] In summary, the anti - collision warning system and warning method for mining area operation vehicles based on zoned warning provided by the present invention have the following beneficial effects compared with the prior art: 1. Reliability advantage of multi - sensor collaboration: The present invention adopts a scheme combining AI cameras and ultrasonic sensors to form a near - far - field collaborative monitoring mechanism; the ultrasonic sensor is responsible for accurate distance measurement in the near - field (within 2.5 meters) with high reliability; the AI camera is responsible for target detection and warning in the far - field (2.5 - 7.5 meters) with a large coverage range; the dual - guarantee mechanism significantly improves the detection reliability of the system in a complex mining area environment.

[0044] 2. Real-time improvement brought by edge computing: The present invention deploys AI algorithms on front-end devices for localized reasoning, eliminating the need for cloud transmission, reducing network latency, and ensuring real-time execution; the embedded processing unit platform equipped with NPU accelerates and ensures the deployment optimization of AI vision models, improving detection speed; significantly reducing system response time and improving early warning real-time performance.

[0045] 3. Scientific nature of zoning warning strategy: The present invention adopts a multi-level warning mechanism based on distance, scientifically divides the warning area, and avoids false alarms and missed alarms; different detection methods and warning levels are used in different areas, which improves the accuracy and practicality of the system.

[0046] 4. High efficiency of dual-stream processing: The present invention innovatively adopts a dual-stream processing mechanism to separate detection and display functions, thereby improving processing efficiency; visual drawing is performed directly on the original image data, reducing format conversion overhead, optimizing system resource utilization, and improving overall performance.

[0047] 5. Advanced engineering implementation: The present invention has significant advantages in practical engineering applications. The AI ​​camera adopts an integrated design with a compact size, which is easy to install and deploy. It also has an infrared fill light function and can adapt to various lighting environments. The AI ​​camera surgery adopts a protective design to meet the harsh environment requirements of the mining area and supports localized storage and information backtracking functions.

[0048] 6. Significant application value: In practical applications, the present invention can significantly improve the safety of mining operations, reduce manpower input, and lower operating costs; provide early warning data analysis, assist in safety management, and has good promotion and application value.

[0049] The present invention organically combines advanced technologies such as artificial intelligence, multi-sensor fusion and edge computing to provide a complete solution for vehicle safety warning in mining areas, which has important practical value and promotion significance.

[0050] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. An anti-collision warning system for mining operation vehicles based on partition warning, characterized in that, The system includes a perception layer, a processing layer, and an application layer; The perception layer includes an AI camera module, a model processing module, an ultrasonic sensor, and a point cloud data processing module; The processing layer includes a collaborative processing module, a warning analysis module, and a status management module connected in sequence; The application layer includes a visualization display module, an audible and visual alarm module, and a data recording module; Among them, the AI camera module transmits the video information detected in the far field to the model processing module, and the model processing module is connected to the coordination processing module; the ultrasonic sensor transmits the data detected in the near field to the point cloud data processing module, and the point cloud data processing module is connected to the coordination processing module; the status management module is respectively connected to the visualization display module, the audible and visual alarm module, and the data recording module.

2. A collision warning method for mining operation vehicles based on partition warning, characterized in that, It includes the following steps: Step 1, Deploy sensors and collect data: Install an AI camera module and an ultrasonic sensor on the mining area operation vehicle, and use the AI camera module to perform video processing on the targets at the front, rear, and both sides of the vehicle; use the ultrasonic sensor to detect the close-range obstacles at the front and rear of the vehicle; Step 2, Data processing and analysis: It includes AI vision processing, dual-stream processing, ultrasonic data processing, and multi-sensor data processing. Among them, multi-sensor data processing is to perform collaborative processing on the AI vision detection results and the ultrasonic distance data to achieve hierarchical anti-collision warning; Step 3, Warning area division and determination: Define the first-level warning area as a high-risk area within 0-2.5 meters from the vehicle, which is monitored by the ultrasonic sensor; define the second-level warning area as a warning area within 2.5-7.5 meters from the vehicle, which is processed by the AI camera module; Step 4, The warning decision-making layer makes an intelligent judgment: When the target enters the first-level warning area, an emergency warning signal is immediately triggered, the dangerous target position is highlighted and displayed on the warning display, an audible and visual alarm prompt is issued, and the warning event information is recorded; when the target enters the second-level warning area, a warning prompt signal is issued, the detection frame of the target object is changed from green to highlighted red on the display interface, the target dynamic information is updated in real time, and the warning status change is recorded; Step 5, Output warning information: Display the warning information through the visualization display module, the audible and visual alarm module, and the data recording module.

3. The anti-collision warning method for mining operation vehicles based on partition warning according to claim 2, wherein: In Step 1, 4 AI cameras are arranged at the front of the mining area operation vehicle, and 2 AI cameras are arranged at the rear. Their horizontal field of view angles are both ≥100°, which are respectively responsible for target detection and warning at the front, rear, and both sides of the vehicle; 2 ultrasonic sensors are arranged at the front and rear of the vehicle respectively, and their horizontal field of view angles ≥60°, which are responsible for close-range obstacle perception; Among them, the AI camera module adopts an edge computing platform equipped with an NPU embedded processing unit, has local inference capabilities, includes a high-definition image sensor and a large-aperture lens, has an NPU neural network processor built-in, and integrates an infrared fill light module.

4. The anti-collision warning method for mining operation vehicles based on zonal warning according to claim 2, characterized in that: In Step 2, the specific process of AI vision processing is: Perform preprocessing on the normalization of the size and format conversion of the collected video stream; input the preprocessed image into the vision target detection model for inference to generate target position, category, and confidence detection results; The dual-stream processing process is as follows: The main stream is mainly used for real-time display of early warning results. Its resolution is the original acquisition resolution, which is responsible for visualizing the drawing of target boxes and warning areas, used for real-time display of the results of AI vision detection, and real-time streaming display through the RTSP protocol; The sub-stream is mainly dedicated to the inference analysis of the AI model. The resolution is optimized and scaled to fit the size of the AI vision model, only used for target detection, generating model detection result data, and no display processing is done; The ultrasonic data processing process is as follows: Real-time process the distance information returned by the ultrasonic sensor, calculate the specific position of the obstacle, and perform noise filtering and data smoothing processing.

5. The anti-collision warning method for mining operation vehicles based on partition warning according to claim 2, wherein: In step three, The first-level warning area is monitored and processed by the ultrasonic sensor for obstacles within the designated range. The specific calculation is as follows: ； Where D is the distance of the obstacle, v is the speed of sound wave propagation in the air, and t is the time interval from transmitting the signal to receiving the echo; The determination of whether there is a target in the second-level warning area uses the ray method. The specific steps are as follows: Obtain the center point coordinates of the target detection box based on the object detection algorithm based on deep learning: ； Draw a ray equation horizontally to the right from this point: ； The equation of the boundary line segment of the warning area polygon. Suppose there are n sides. For the i-th boundary line segment, the two endpoints are respectively and Line segment equation: ， < < ； For each boundary line segment, the x coordinate of the intersection point of the ray and the boundary line segment satisfies the following equation: ； When ≤ ≤ and ≥ the intersection point is valid; The ray method determination rule is as follows: The number of intersection points is even, including 0, the point is outside; The number of intersection points is odd, the point is inside: ； Determine whether the target is in the warning area according to the parity of the number of intersection points. When it is detected that the target enters the warning area, immediately trigger the corresponding level of warning response.

6. The anti-collision warning method for mining operation vehicles based on partition warning according to claim 2, wherein: In step five, the specific process of the visualization display module is as follows: Real-time display the detection target position on the warning display, display the warning area status, and directly perform image annotation and display based on the main stream.

7. The anti-collision warning method for mining operation vehicles based on partition warning according to claim 2, wherein: In step five, the specific process of the audible and visual alarm module is as follows: Trigger the corresponding audible and visual warning signals according to different warning levels, and support warning methods with different intensities and frequencies.

8. The anti-collision warning method for mining operation vehicles based on partition warning according to claim 2, wherein: In step five, the specific process of the data recording module is as follows: Automatically record warning events and system status information, support data playback and analysis functions, and generate warning statistical reports.

Images