Road congestion condition detection method taking robustness vehicle target detection as core
A technology of vehicle detection and target detection, which is applied in the traffic control system of road vehicles, traffic flow detection, instruments, etc., and can solve the problems of damaging the road surface, poor real-time performance, and insufficient accuracy
Patent Information
- Authority / Receiving Office
- CN · China
- Current Assignee / Owner
- Publication Date
- 2021-07-30
Smart Images

Figure 1 
Figure 2 
Figure 3
Abstract
Description
technical field
[0001] The invention relates to a traffic jam detection method, in particular to a road congestion detection method with robust vehicle target detection as the core. Background technique
[0002] With the gradual improvement of people's living standards, the per capita occupancy rate of motor vehicles has increased significantly, and urban traffic congestion has become increasingly serious. Traffic congestion is a phenomenon in which traffic activities are slow and interrupted due to traffic volume exceeding road capacity, resulting in travel delays and environmental pollution causing massive economic losses. Therefore, it is of great significance to study traffic congestion detection methods to make targeted preventive measures for traffic congestion. For traffic congestion detection, the traditional method uses sensors to obtain traffic flow parameters such as the number of vehicles on the road and vehicle speed, which has the following disadvantages: [...
Examples
specific Embodiment
[0094] The detection of urban traffic road congestion needs to be real-time and reliable at the same time. The traditional method of judging congestion has the disadvantages of obtaining inaccurate road surface information and relying too much on historical data. In order to obtain accurate road vehicle information in real time and provide accurate road vehicle information for congestion detection, the vehicle must first be detected in real time and accurately. YOLOv3 is an end-to-end target detection algorithm with both speed and accuracy. It uses a large number of residual structures in the feature extraction network to ensure that the deep network can effectively extract the features of the target. Targets of different sizes are positioned, and the output network is divided into three layers, corresponding to three different sizes of targets: large, medium and small. At the same time, since a network with a fixed output size will limit the receptive field of the output neur...