Bus-mounted traffic civilization behavior monitoring system and method based on cloud-side cooperation

Abstract

The invention belongs to the technical field of intelligent traffic monitoring, and particularly relates to a bus-mounted traffic civilization behavior monitoring system and method based on cloud-side collaboration, and the monitoring system comprises an edge sensing unit at a bus end side and a cloud processing unit at a center cloud side. The edge sensing unit collects videos through a vehicle-mounted camera, an edge calculation box carries out real-time analysis by using an algorithm of modular pipeline design, various traffic uncivilized behaviors are identified, and an alarm event is generated and uploaded. And the cloud unit performs data aggregation and deep analysis. According to the invention, through the cloud edge collaborative architecture, the calculation load is reasonably distributed, and the bandwidth and center computing power cost is reduced; wide-area monitoring is realized by using the flow characteristics of the buses; the modular design enables the system to flexibly adapt to various detection scenes, and achieves the efficient, real-time and low-cost monitoring and evidence obtaining of traffic uncivilized behaviors.

Description

Technical Field

[0001] This invention belongs to the field of intelligent traffic monitoring technology, specifically relating to a cloud-edge collaborative bus-borne traffic civility monitoring system and method. Background Technology

[0002] Traffic civility is a crucial aspect of creating a civilized city. Existing methods for monitoring civilized traffic behavior mainly include on-site enforcement by traffic police, off-site enforcement using traffic monitoring technology, and public reporting of traffic violations. However, these methods all have limitations and are insufficient for comprehensively and effectively monitoring traffic violations. On-site enforcement by traffic police is costly in terms of manpower, and the monitoring range is constrained by limited police resources. Traffic monitoring technology is typically installed in fixed locations, also limiting its monitoring range and making it impossible to cover traffic violations across broad areas of general roads. Furthermore, drivers are easily aware of violations and can deliberately avoid them. Public reporting of traffic violations suffers from issues such as uploaded video materials not meeting evidence standards or missing violation clues, resulting in a low success rate and high manual review costs.

[0003] Therefore, designing a low-cost, high-efficiency, wide-coverage, and high-quality real-time mobile monitoring method and system for civilized traffic behavior is of great significance for improving the level of civilized traffic. To this end, this invention provides a cloud-edge collaborative bus-borne civilized traffic behavior monitoring system and method. Summary of the Invention

[0004] The purpose of this invention is to provide a cloud-edge collaborative bus traffic civility monitoring system and method, which has the advantages of low cost, high efficiency, wide coverage and high evidence quality.

[0005] The specific technical solution adopted by this invention is as follows:

[0006] A cloud-edge collaborative bus traffic civility monitoring system includes several bus terminals and a central cloud monitoring system. The bus terminals are connected to the core switch of the central cloud monitoring system via a 5G module and the Internet.

[0007] The bus end-side includes: an edge sensing unit deployed on the bus, used to collect video in real time, perform edge computing analysis and generate alarm events;

[0008] The central cloud monitoring system includes a cloud processing unit and a video surveillance system deployed in the management center, used to collect, store, display and analyze the alarm events;

[0009] The edge sensing unit includes:

[0010] At least one set of vehicle-mounted cameras installed on the bus;

[0011] An edge computing box, connected to the vehicle-mounted camera, is used to analyze video streams in real time, identify preset types of uncivilized traffic behaviors, and generate alarm event data. The edge computing box adopts a modular pipeline design and includes at least a raw data acquisition module and a target attribute acquisition module connected in sequence.

[0012] The communication module is used to upload alarm event data to the cloud processing unit.

[0013] Preferably, the modular pipeline design inside the edge computing box further includes:

[0014] An optional intermediate processing module is connected after the target attribute acquisition module;

[0015] An optional logic rule judgment module is connected after the intermediate process processing module or directly after the target attribute acquisition module;

[0016] By selecting and enabling different module combinations, multiple data processing pipelines can be constructed to support the detection of various types of uncivilized traffic behaviors.

[0017] Preferably, the target attribute acquisition module includes: a pedestrian target detection and attribute recognition submodule, a motor vehicle target detection and attribute recognition submodule, a non-motor vehicle target detection and attribute recognition submodule, and an other object target detection and attribute recognition submodule;

[0018] The intermediate processing module includes: a target trajectory analysis submodule and a target overlap relationship submodule;

[0019] The logic rule judgment module includes: a line mixing logic rule submodule and a region detection logic submodule.

[0020] Preferably, the edge sensing unit further includes a sound column connected to the edge computing box, used to provide voice prompts when uncivilized behavior is detected.

[0021] Preferably, the edge computing box is also used to receive the GPS positioning information of the bus and dynamically switch the enabled data processing pipeline combination according to the relationship between the vehicle's current location and the preset geofence.

[0022] Preferably, when the vehicle's position enters the preset range of a traffic intersection, the system switches to a data processing pipeline combination for detecting behaviors such as running red lights, failing to yield to pedestrians, driving against traffic, occupying motor vehicle lanes, or crossing the road; when the vehicle's position is not at an intersection, the system switches to a data processing pipeline combination for detecting behaviors such as pedestrians climbing over guardrails or not wearing helmets.

[0023] A cloud-edge collaborative method for monitoring civilized traffic behavior on public buses, applied to a monitoring system, includes the following steps:

[0024] S1: Collect road environment video streams via bus-mounted cameras;

[0025] S2: The edge computing box receives and parses the video stream, and performs target detection, attribute recognition and rule judgment through a modular pipeline;

[0026] S3: When uncivilized behavior that meets preset conditions is identified, an alarm event is generated and uploaded to the cloud;

[0027] S4: The cloud processing unit receives alarm events, performs data aggregation, statistical analysis, and visualization.

[0028] Preferably, in step S2, the specific steps for target detection, attribute recognition, and rule determination via a modular pipeline include:

[0029] Based on the type of uncivilized behavior to be detected, a combination of the corresponding target attribute acquisition submodule, intermediate process processing submodule, and logical rule judgment submodule is selected to form a specific data processing pipeline;

[0030] The video stream is processed sequentially through the pipeline, and the behavior recognition result is output.

[0031] Preferably, the vehicle-mounted speaker is triggered to provide on-site voice prompts at the same time the edge computing box generates an alarm event.

[0032] Preferably, the method further includes the step of: the edge computing box automatically selecting and switching the currently effective data processing pipeline combination based on the vehicle's real-time GPS location information to adapt to different road scene monitoring needs.

[0033] The technical effects achieved by this invention are as follows:

[0034] Bus-mounted video surveillance boasts advantages such as wide spatial and temporal coverage, low evidence collection costs, and high-quality leads, making it suitable as a new data source for monitoring traffic behavior. In terms of coverage, bus-mounted video surveillance can conduct high-frequency, mobile monitoring of traffic behavior along bus routes during operating hours, effectively eliminating blind spots of fixed traffic monitoring equipment.

[0035] This invention employs intelligent edge computing devices deployed inside buses to access bus video feeds. It monitors people, vehicles, non-motorized vehicles, roads, and the environment within the video feeds, identifies and detects uncivilized behaviors in traffic scenarios, and uses loudspeakers to promptly correct these behaviors. Simultaneously, it uploads the location of the uncivilized behavior to a central terminal, where uncivilized behavior is scored to determine high-incidence road sections, locations, and types of uncivilized behavior.

[0036] Intelligent algorithm edge devices include GPU chips, which are dedicated to image and graphics processing and excel at performing simple, repetitive operations on large datasets. Furthermore, compared to intelligent algorithm server backend deployment solutions, they enable simultaneous intelligent analysis and video surveillance, thus reducing backend analysis costs.

[0037] This invention adopts a "cloud-edge collaboration" design concept, where "cloud" refers to cloud computing and "edge" refers to edge computing. "Cloud-edge collaboration" means integrating cloud computing and edge computing to deliver overall computing power to each bus. By leveraging the rapid response, bandwidth saving, and compatibility with existing infrastructure of edge computing, the overall project investment cost is significantly reduced, while improving computing results and efficiency.

[0038] In this invention, edge computing detects traffic behavior through edge boxes, while the cloud center collects data and reports events through an AI service platform, enabling rapid and efficient detection, alarming, and handling of potential hazards. Attached Figure Description

[0039] Figure 1 This is a system block diagram of a cloud-edge collaborative bus traffic civility monitoring system according to the present invention;

[0040] Figure 2 This is a system block diagram of the vehicle delivery end side of the present invention;

[0041] Figure 3 This is a flowchart illustrating the actual operation of a cloud-edge collaborative method for monitoring civilized traffic behavior on public buses, based on the present invention. Detailed Implementation

[0042] To make the objectives and advantages of this invention clearer, the invention will be specifically described below with reference to embodiments. It should be understood that the following text is merely used to describe one or more specific embodiments of the invention and does not strictly limit the scope of protection specifically claimed by the invention.

[0043] like Figure 1 As shown, a cloud-edge collaborative bus traffic civility monitoring system includes several bus terminals and a central cloud platform. The bus terminals are connected to the core switch of the central cloud platform via 5G modules and the Internet.

[0044] The bus end-side includes: an edge sensing unit deployed on the bus, used to collect video in real time, perform edge computing analysis and generate alarm events;

[0045] The central cloud monitoring system includes a cloud processing unit and a video surveillance system deployed in the management center, used to collect, store, display and analyze the alarm events;

[0046] The edge sensing unit includes:

[0047] At least one set of onboard cameras installed on the bus, for example, each set of onboard cameras includes four onboard cameras installed at the front, rear, and sides of the bus, such as... Figure 2 As shown;

[0048] An edge computing box, connected to the vehicle-mounted camera, is used to analyze video streams in real time, identify preset types of uncivilized traffic behaviors, and generate alarm event data. The edge computing box adopts a modular pipeline design and includes at least a raw data acquisition module and a target attribute acquisition module connected in sequence.

[0049] The communication module is used to upload alarm event data to the cloud processing unit.

[0050] Preferably, the modular pipeline design inside the edge computing box further includes:

[0051] An optional intermediate processing module is connected after the target attribute acquisition module;

[0052] An optional logic rule judgment module is connected after the intermediate process processing module or directly after the target attribute acquisition module;

[0053] By selecting and enabling different module combinations, multiple data processing pipelines can be constructed to support the detection of various types of uncivilized traffic behaviors.

[0054] Preferably, the target attribute acquisition module includes: a pedestrian target detection and attribute recognition submodule, a motor vehicle target detection and attribute recognition submodule, a non-motor vehicle target detection and attribute recognition submodule, and an other object target detection and attribute recognition submodule;

[0055] The intermediate processing module includes: a target trajectory analysis submodule and a target overlap relationship submodule;

[0056] The logic rule judgment module includes: a line mixing logic rule submodule and a region detection logic submodule.

[0057] Preferably, the edge sensing unit further includes a sound column connected to the edge computing box, used to provide voice prompts when uncivilized behavior is detected.

[0058] Preferably, the edge computing box is also used to receive the GPS positioning information of the bus and dynamically switch the enabled data processing pipeline combination according to the relationship between the vehicle's current location and the preset geofence.

[0059] Preferably, when the vehicle's position enters the preset range of a traffic intersection, the system switches to a data processing pipeline combination for detecting behaviors such as running red lights, failing to yield to pedestrians, driving against traffic, occupying motor vehicle lanes, or crossing the road; when the vehicle's position is not at an intersection, the system switches to a data processing pipeline combination for detecting behaviors such as pedestrians climbing over guardrails or not wearing helmets.

[0060] A cloud-edge collaborative method for monitoring civilized traffic behavior on public buses, applied to a monitoring system, includes the following steps:

[0061] S1: Collect road environment video streams via bus-mounted cameras;

[0062] S2: The edge computing box receives and parses the video stream, and performs target detection, attribute recognition and rule judgment through a modular pipeline;

[0063] S3: When uncivilized behavior that meets preset conditions is identified, an alarm event is generated and uploaded to the cloud;

[0064] S4: The cloud processing unit receives alarm events, performs data aggregation, statistical analysis, and visualization.

[0065] Preferably, in step S2, the specific steps for target detection, attribute recognition, and rule determination via a modular pipeline include:

[0066] Based on the type of uncivilized behavior to be detected, a combination of the corresponding target attribute acquisition submodule, intermediate process processing submodule, and logical rule judgment submodule is selected to form a specific data processing pipeline;

[0067] The video stream is processed sequentially through the pipeline, and the behavior recognition result is output.

[0068] Preferably, the vehicle-mounted speaker is triggered to provide on-site voice prompts at the same time the edge computing box generates an alarm event.

[0069] Preferably, the method further includes the step of: the edge computing box automatically selecting and switching the currently effective data processing pipeline combination based on the vehicle's real-time GPS location information to adapt to different road scene monitoring needs.

[0070] In actual operation, the edge computing box for bus-mounted traffic civilization is internally structured into four modules: raw data acquisition (frame extraction), target attribute acquisition, intermediate process processing, and logical rule judgment. Data flows sequentially between these modules, and algorithm results are output through module combinations. The raw data acquisition and target attribute acquisition modules are mandatory, while the intermediate process processing and logical rule judgment modules are optional.

[0071] The raw data acquisition module includes a frame extraction module, which converts the video stream into an image stream for processing by the target attribute acquisition module.

[0072] The target attribute acquisition module consists of four sub-modules: pedestrian target detection and attribute recognition, motor vehicle target detection and attribute recognition, non-motor vehicle target detection and attribute recognition, and other object target detection and attribute recognition; the main categories include: helmets, traffic lights, and guardrails;

[0073] The intermediate processing module contains three sub-modules: target trajectory analysis and target overlap relationship. If the intermediate process is omitted, this module is not needed.

[0074] The logic rule judgment module contains five sub-modules: lane marking logic rule, counting logic, lane line judgment logic, and area detection logic. If the logic rule is omitted, this module is not needed.

[0075] The combinations of results generated by each algorithm are as follows:

[0076] Pedestrian crossing guardrail detection: such as Figure 3 As shown in the data flow line marked 1 in the detection process of the bus-mounted traffic civilization edge analysis box, when the pedestrian target detection and attribute recognition submodule in the target attribute acquisition module detects a pedestrian, the output result is sent to the target trajectory analysis submodule in the intermediate process processing module. The trajectory tracking result enters the line crossing logic rule judgment in the logic rule judgment module. When the condition is met, the result of the guardrail climbing algorithm is output.

[0077] Detection of non-motorized vehicles / delivery bikes without helmets: such as Figure 3 As shown in the data flow path labeled 2 in the bus traffic civilization edge analysis box detection process, when the non-motorized vehicle target detection and attribute recognition submodule in the target attribute acquisition module detects a non-motorized vehicle and the other object detection and attribute recognition submodule detects a helmet, the output results of the two modules are sent to the target overlap relationship submodule in the intermediate process processing module. When the overlap relationship meets the requirements, the judgment result data skips the logic rule judgment submodule and directly outputs the non-motorized vehicle / delivery vehicle helmet-wearing detection result.

[0078] Detection of non-motorized vehicles / delivery vehicles / courier vehicles driving against traffic / occupying motor vehicle lanes / crossing the road: such as Figure 3 As shown in the data flow line numbered 3 in the bus traffic civilization edge analysis box detection process, when the non-motorized vehicle target detection and attribute recognition submodule in the target attribute acquisition module detects a non-motorized vehicle / delivery vehicle / express vehicle, the output result is sent to the target analysis submodule in the intermediate process processing module. The trajectory obtained from the analysis enters the area detection judgment logic in the logic rule judgment module. When the conditions are met, the algorithm results of non-motorized vehicles / delivery vehicles / express vehicles going against traffic / running red lights / crossing the road are output.

[0079] Motor vehicles running red lights / failing to yield to pedestrians: such as Figure 3 As shown in the data flow path numbered 4 in the bus traffic civilization edge analysis box detection process, when the pedestrian target detection and attribute recognition submodule in the target attribute acquisition module detects a pedestrian, and the motor vehicle target detection and attribute recognition submodule detects a motor vehicle, the output result is sent to the target analysis submodule in the intermediate process processing module. The trajectory obtained from the analysis enters the area detection judgment logic in the logic rule judgment module. When the conditions are met, the algorithm result of motor vehicle running red light / failing to yield to pedestrians is output.

[0080] In actual use, for the four different onboard cameras installed on the bus, cameras 2 & 4 are used to detect pedestrians climbing over guardrails and non-motorized vehicles / delivery riders not wearing helmets. Cameras 1 & 3 are used to detect non-motorized vehicles / delivery riders / courier vehicles going against traffic / occupying motorized vehicle lanes / crossing the road, and motorized vehicles running red lights / failing to yield to pedestrians. Combining the vehicle's GPS location data, when the bus is approaching traffic intersections, the system switches to the detection logic for non-motorized vehicles / delivery riders going against traffic / occupying motorized vehicle lanes / crossing the road, and motorized vehicles running red lights / failing to yield to pedestrians. In other situations, the system uses the detection logic for pedestrians climbing over guardrails and non-motorized vehicles / delivery riders not wearing helmets. This makes full use of computing resources. When uncivilized behavior is detected, a trigger speaker can be configured to issue a warning.

[0081] In summary, the bus-mounted edge computing box for civilized traffic behavior includes the following algorithmic capabilities:

[0082] Fence-climbing detection: This algorithm detects individuals illegally climbing over fences. Once detected, the AI ​​algorithm immediately issues an alarm and outputs a bounding box diagram of the illegal event.

[0083] Motor vehicles / non-motor vehicles / pedestrians running red lights: Combining the traffic light status, the system uses AI algorithms such as image recognition and target detection to analyze the video stream in real time and identify motor vehicles running red lights. Once the AI ​​algorithm detects the violation, it immediately issues an alarm and outputs a captured image of the violation.

[0084] Motor vehicles failing to yield to pedestrians: AI algorithms such as image recognition and target detection are used to analyze video streams in real time to identify the behavior of motor vehicles passing through zebra crossing areas at traffic intersections. Once the AI ​​algorithm detects the violation, it immediately issues an alarm and outputs a captured image of the violation.

[0085] Non-motorized vehicles / delivery bikes without helmets: Based on road cameras, the system detects non-motorized vehicles without helmets, reports and issues warnings upon discovery of violations. Using urban road surveillance video, combined with traffic light status and delivery bike trajectory information, the system identifies delivery bikes running red lights and immediately issues warnings upon detection.

[0086] Non-motorized vehicles / delivery vehicles running red lights: Based on urban road surveillance video, combined with traffic light status and delivery vehicle trajectory information, it identifies non-motorized vehicles / delivery vehicles running red lights and immediately issues an alarm once detected.

[0087] Non-motorized vehicles / delivery vehicles traveling in the wrong direction: Based on urban road surveillance video, combined with road travel direction and delivery vehicle travel trajectory information, it identifies non-motorized vehicles / delivery vehicles traveling in the wrong direction and immediately issues an alarm once detected.

[0088] Non-motorized vehicles / delivery vehicles crossing the road: Based on urban road surveillance video and combined with the driving trajectory of delivery vehicles, an alarm is immediately issued when a non-motorized vehicle / delivery vehicle is detected crossing the road.

[0089] Non-motorized vehicles / delivery vehicles occupying motor vehicle lanes: Based on urban road surveillance video and lane type, an alarm will be immediately triggered when a non-motorized vehicle / delivery vehicle is detected in a motor vehicle lane.

[0090] Delivery vehicle red light detection: Based on traffic department electronic police monitoring video data, AI algorithms such as image recognition, target detection, and trajectory tracking are used to analyze the video stream in real time. Combined with the traffic light status, the system detects whether delivery vehicles within the designated area have run red lights. Once the algorithm detects a red light, it immediately issues an alarm and outputs an alarm event diagram.

[0091] Delivery vehicle crossing the road detection: Based on traffic department electronic police monitoring video data, AI algorithms such as image recognition, target detection, and trajectory tracking are used to analyze the video stream in real time to detect whether delivery vehicles are crossing the road. Once the algorithm detects it, it immediately alarms and outputs an alarm event diagram.

[0092] Traditional deployment methods using only a central endpoint can lead to wasted computing resources in small parks or mining areas with few video points. Therefore, for scenarios such as small and medium-sized parks, a cloud-edge collaborative approach can be adopted, utilizing edge computing boxes to fully leverage the value of computing resources. In this invention, to fully utilize bus-mounted video and reduce video transmission network bandwidth, onboard edge devices are used for parsing, enabling timely event detection and processing, while also reducing resource investment.

[0093] The algorithm is rich in scenarios and fully meets the needs of traffic civilization scenarios. A single box includes multiple algorithms such as pedestrian crossing guardrail detection, non-motorized vehicle / delivery vehicle helmet-wearing detection, non-motorized vehicle / delivery vehicle / express delivery vehicle going against traffic / occupying motor vehicle lane / crossing the road detection, and motor vehicle running red lights / failing to yield to pedestrians. It also supports mixed configuration of algorithms and supports full algorithm analysis of 4-channel 1080P video streams. The number of streams that can be analyzed can be specifically evaluated according to detailed algorithm requirements.

[0094] The above description is merely a preferred embodiment of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention. Structures, devices, and operating methods not specifically described or explained in this invention are implemented according to conventional methods in the art unless otherwise specified or limited.

Claims

1. A cloud-edge collaborative bus-borne traffic civility behavior monitoring system, characterized in that: It includes several bus-side terminals and a central cloud testing system. The bus-side terminals are connected to the core switch of the central cloud testing system via 5G modules and the Internet. The bus end-side includes: an edge sensing unit deployed on the bus, used to collect video in real time, perform edge computing analysis and generate alarm events; The central cloud monitoring system includes a cloud processing unit and a video surveillance system deployed in the management center, used to collect, store, display and analyze the alarm events; The edge sensing unit includes: At least one set of vehicle-mounted cameras installed on the bus; An edge computing box, connected to the vehicle-mounted camera, is used to analyze video streams in real time, identify preset types of uncivilized traffic behaviors, and generate alarm event data. The edge computing box adopts a modular pipeline design and includes at least a raw data acquisition module and a target attribute acquisition module connected in sequence. The communication module is used to upload alarm event data to the cloud processing unit.

2. The monitoring system according to claim 1, characterized in that: The modular pipeline design inside the edge computing box also includes: An optional intermediate processing module is connected after the target attribute acquisition module; An optional logic rule judgment module is connected after the intermediate process processing module or directly after the target attribute acquisition module; By selecting and enabling different module combinations, multiple data processing pipelines can be constructed to support the detection of various types of uncivilized traffic behaviors.

3. The monitoring system according to claim 2, characterized in that: The target attribute acquisition module includes: a pedestrian target detection and attribute recognition submodule, a motor vehicle target detection and attribute recognition submodule, a non-motor vehicle target detection and attribute recognition submodule, and an other object target detection and attribute recognition submodule; The intermediate processing module includes: a target trajectory analysis submodule and a target overlap relationship submodule; The logic rule judgment module includes: a line mixing logic rule submodule and a region detection logic submodule.

4. The monitoring system according to claim 3, characterized in that: The edge sensing unit also includes a sound column connected to the edge computing box, used to provide voice prompts when uncivilized behavior is detected.

5. The monitoring system according to claim 4, characterized in that: The edge computing box is also used to receive GPS positioning information from the bus and dynamically switch the enabled data processing pipeline combination based on the relationship between the vehicle's current location and the preset geofence.

6. The monitoring system according to claim 1, characterized in that: When the vehicle enters the preset range of a traffic intersection, the system switches to a data processing pipeline combination for detecting behaviors such as running red lights, failing to yield to pedestrians, driving against traffic, occupying motor vehicle lanes, or crossing the road. When the vehicle is not located at an intersection, the system switches to a data processing pipeline combination for detecting behaviors such as pedestrians climbing over guardrails or not wearing helmets.

7. A method for monitoring civilized traffic behavior on public buses based on cloud-edge collaboration, characterized in that, The method applied to the monitoring system as described in any one of claims 1-6 includes the following steps: S1: Collect road environment video streams via bus-mounted cameras; S2: The edge computing box receives and parses the video stream, and performs target detection, attribute recognition and rule judgment through a modular pipeline; S3: When uncivilized behavior that meets preset conditions is identified, an alarm event is generated and uploaded to the cloud; S4: The cloud processing unit receives alarm events, performs data aggregation, statistical analysis, and visualization.

8. The monitoring method according to claim 7, characterized in that, In step S2, the specific steps for target detection, attribute recognition, and rule determination via a modular pipeline include: Based on the type of uncivilized behavior to be detected, a combination of the corresponding target attribute acquisition submodule, intermediate process processing submodule, and logical rule judgment submodule is selected to form a specific data processing pipeline; The video stream is processed sequentially through the pipeline, and the behavior recognition result is output.

9. The monitoring method according to claim 7 or 8, characterized in that, While the edge computing box generates an alarm event, it triggers the vehicle-mounted speaker to provide on-site voice prompts.

10. The method according to claim 7 or 8, characterized in that, It also includes the following steps: The edge computing box automatically selects and switches the currently active data processing pipeline combination based on the vehicle's real-time GPS location information to adapt to different road scene monitoring needs.

Images