CN-121999611-A - Obstacle early warning system and obstacle early warning method

CN121999611ACN 121999611 ACN121999611 ACN 121999611ACN-121999611-A

Abstract

The application discloses an obstacle early warning system and an obstacle early warning method, and relates to the technical field of Internet, wherein the system comprises a vehicle terminal and a cloud platform; the method comprises the steps of generating an obstacle rich media early warning package of a road obstacle when a vehicle terminal detects that the road obstacle exists on a current driving road, transmitting the obstacle rich media early warning package to a cloud platform through a vehicle road communication channel, performing risk assessment on the road obstacle by the cloud platform based on the target image data, determining a risk grade corresponding to the road obstacle, and early warning the road obstacle based on the risk grade. According to the application, the cloud end platform can determine the risk level of the obstacle based on the image data of the obstacle sent by the vehicle terminal, and perform obstacle early warning based on the risk level, so that the technical problems of low instantaneity and accuracy of the existing obstacle early warning scheme relying on manual reporting are solved.

Inventors

YANG JIEKUN
LIU XIANGHENG
HUANG SHAOMANG
PAN JIANFENG

Assignees

三六零数字安全科技集团有限公司

Dates

Publication Date: 20260508
Application Date: 20260127

Claims (10)

1. The obstacle early warning system is characterized by comprising a vehicle terminal and a cloud platform; The vehicle terminal is used for generating an obstacle rich media early warning packet of the road obstacle under the condition that the road obstacle exists on the current driving road, and the obstacle rich media early warning packet comprises target image data of the road obstacle; the vehicle terminal is further used for transmitting the obstacle rich media early warning packet to the cloud platform through a vehicle-to-vehicle communication channel; The cloud platform is used for performing risk assessment on the road obstacle based on the target image data and determining a risk level corresponding to the road obstacle; the cloud platform is further used for early warning the road obstacle based on the risk level.
2. The system according to claim 1, wherein the vehicle terminal is configured to collect multi-modal environmental data of a current driving road in real time, and perform obstacle detection on the current driving road based on the multi-modal environmental data; The vehicle terminal is further used for acquiring the obstacle information of the road obstacle under the condition that the road obstacle exists on the current driving road; the vehicle terminal is further used for generating an obstacle rich media early warning package of the road obstacle based on the obstacle information.
3. The system of claim 2, wherein the vehicle terminal is configured to collect in real time multi-modal environmental data of a current driving road, the multi-modal environmental data including vehicle image data and vehicle radar data; the vehicle terminal is further used for fusing the vehicle image data and the vehicle radar data to generate target environment data; The vehicle terminal is further used for detecting the obstacle on the current driving road based on the target environment data through a preset obstacle recognition algorithm.
4. The system of claim 3, wherein the cloud platform is configured to finely identify the road obstacle based on the target image data to obtain target obstacle detection information of the road obstacle; The cloud platform is further used for determining a risk level corresponding to the road obstacle according to the target obstacle detection information.
5. The system of claim 4, wherein the cloud platform is configured to input the target image data to a preset image recognition engine, and the preset image recognition engine is provided with a feature extraction module and an obstacle detection module; The cloud platform is further used for carrying out feature extraction on the target image data through the feature extraction module to generate a semantic feature map of the road obstacle; the cloud platform is further used for carrying out fine recognition on the road obstacle based on the semantic feature map through the obstacle detection module to obtain target obstacle detection information of the road obstacle.
6. The system of claim 5, wherein the cloud platform is further configured to determine, by the obstacle detection module, candidate obstacle regions from the semantic feature map, the candidate obstacle regions being regions that are likely to contain an obstacle; the cloud platform is further used for determining the obstacle types and the obstacle positions of all road obstacles in the candidate obstacle region; The cloud platform is further configured to perform example segmentation on the candidate obstacle region to obtain a segmentation mask of the candidate obstacle region, where the segmentation mask is used to characterize a contour of the road obstacle; the cloud platform is further configured to generate target obstacle detection information of the road obstacle based on the obstacle type, the obstacle position, and the segmentation mask.
7. The system of claim 6, wherein the cloud platform is further configured to classify the candidate obstacle region, and detect whether a suspicious obstacle in the candidate obstacle region is a real obstacle according to a classification result; The cloud platform is further configured to determine the suspicious obstacle as a road obstacle if the suspicious obstacle is a road obstacle, and acquire an obstacle type of the road obstacle; the cloud platform is further used for carrying out frame regression processing on the candidate obstacle areas to obtain adjusted candidate obstacle areas; The cloud platform is further used for determining the obstacle position of the road obstacle according to the adjusted candidate obstacle area.
8. The system of any one of claims 4 to 7, wherein the cloud platform is further configured to obtain an obstacle reporting event corresponding to the road obstacle, where the obstacle reporting event carries multi-source obstacle detection data of the road obstacle; the cloud platform is further used for mapping the multi-source obstacle detection data to a three-dimensional space-time space to obtain obstacle space-time information of the road obstacle; the cloud platform is further used for detecting whether the obstacle reporting event is a real obstacle event or not based on the obstacle space-time information; and the cloud platform is further used for determining the risk level corresponding to the road obstacle according to the target obstacle detection information if the road obstacle is detected.
9. The system of claim 8, wherein the cloud platform is further configured to perform cluster analysis on the obstacle spatiotemporal information by using a preset spatial clustering algorithm, and determine a high-density cluster of points in the three-dimensional spatiotemporal space; The cloud platform is further used for determining event confidence and event space-time characteristics of the event reported by the obstacle according to the high-density point clusters; The cloud platform is further configured to detect whether the obstacle reporting event is a real obstacle event based on the event confidence and the event space-time feature.
10. An obstacle pre-warning method based on the obstacle pre-warning system according to any one of claims 1 to 9, characterized in that the method is applied to a cloud platform, the method comprising: Receiving an obstacle rich media early warning packet generated by a vehicle terminal based on a road obstacle, wherein the obstacle rich media early warning packet comprises target image data of the road obstacle; Fine recognition is carried out on the road obstacle based on the target image data, and target obstacle detection information of the road obstacle is obtained; Determining a risk level corresponding to the road obstacle according to the target obstacle detection information; And carrying out early warning on the road obstacle based on the risk grade.

Description

Obstacle early warning system and obstacle early warning method Technical Field The application relates to the technical field of Internet, in particular to an obstacle early warning system and an obstacle early warning method. Background With the rapid development of intelligent traffic systems, the improvement of road safety and traffic efficiency is an important issue. In modern traffic environments, road sudden obstacles (such as falling objects, traffic accidents, road surface damage caused by bad weather, etc.) pose a serious threat to traffic safety. At present, the existing early warning scheme for the road sudden obstacle mainly relies on manual reporting, namely, after a driver or a road administrator finds the obstacle, the driver or the road administrator reports the obstacle to a traffic management department manually, and then related departments conduct information transmission and processing. However, the obstacle early warning process is generally long in time consumption, and deviation or inaccuracy may exist in the manually reported information, so that the real-time performance and the accuracy of the obstacle early warning are not high. Disclosure of Invention The application mainly aims to provide an obstacle early warning system and an obstacle early warning method, and aims to solve the technical problems that the real-time performance and the accuracy of the existing obstacle early warning scheme which relies on manual reporting are not high. In order to achieve the above purpose, the application provides an obstacle early warning system, which comprises a vehicle terminal and a cloud platform; The vehicle terminal is used for generating an obstacle rich media early warning packet of the road obstacle under the condition that the road obstacle exists on the current driving road, and the obstacle rich media early warning packet comprises target image data of the road obstacle; the vehicle terminal is further used for transmitting the obstacle rich media early warning packet to the cloud platform through a vehicle-to-vehicle communication channel; The cloud platform is used for performing risk assessment on the road obstacle based on the target image data and determining a risk level corresponding to the road obstacle; the cloud platform is further used for early warning the road obstacle based on the risk level. In an embodiment, the vehicle terminal is configured to collect multi-mode environmental data of a current driving road in real time, and detect an obstacle on the current driving road based on the multi-mode environmental data; The vehicle terminal is further used for acquiring the obstacle information of the road obstacle under the condition that the road obstacle exists on the current driving road; the vehicle terminal is further used for generating an obstacle rich media early warning package of the road obstacle based on the obstacle information. In an embodiment, the vehicle terminal is configured to collect, in real time, multi-modal environmental data of a current driving road, where the multi-modal environmental data includes vehicle image data and vehicle radar data; the vehicle terminal is further used for fusing the vehicle image data and the vehicle radar data to generate target environment data; The vehicle terminal is further used for detecting the obstacle on the current driving road based on the target environment data through a preset obstacle recognition algorithm. In an embodiment, the cloud platform is configured to perform fine recognition on the road obstacle based on the target image data, and obtain target obstacle detection information of the road obstacle; The cloud platform is further used for determining a risk level corresponding to the road obstacle according to the target obstacle detection information. In an embodiment, the cloud platform is configured to input the target image data to a preset image recognition engine, where a feature extraction module and an obstacle detection module are disposed in the preset image recognition engine; The cloud platform is further used for carrying out feature extraction on the target image data through the feature extraction module to generate a semantic feature map of the road obstacle; the cloud platform is further used for carrying out fine recognition on the road obstacle based on the semantic feature map through the obstacle detection module to obtain target obstacle detection information of the road obstacle. In an embodiment, the cloud platform is further configured to determine, by using the obstacle detection module, a candidate obstacle region from the semantic feature map, where the candidate obstacle region is a region that may include an obstacle; the cloud platform is further used for determining the obstacle types and the obstacle positions of all road obstacles in the candidate obstacle region; The cloud platform is further configured to perform example segmentation on the candidate obstacl