Vehicle safety warning device detection method based on space prior

Abstract

The invention relates to the field of vehicle safety warning device detection, in particular to a vehicle safety warning device detection method based on space prior. The invention provides a vehicle safety warning device detection method based on deep learning and space prior. The method comprises the following steps: S1, an existing data set being labeled; S2, training a YOLOv3 target recognition framework by using the marked data set; S3, calculating a distance reference quantity and an area reference quantity by utilizing the training set and the label thereof; S4, identifying the picture to be predicted by using a YOLOv3 framework, and then obtaining an array in which warning sign information is stored; S5, calculating a distance score and an area score of the warning device in the picture to be predicted by using the information in the array; S6, if the sum of the distance score and the area score is larger than a preset threshold value, outputting and marking as a key target at the same time.

Description

technical field [0001] The invention relates to the field of vehicle safety warning device detection, in particular to a space prior-based detection method for vehicle safety warning devices. Background technique [0002] In recent years, the number of motor vehicles has continued to increase at a high level, which has brought significant traffic pressure and challenges in vehicle management. Among them, the vehicle annual inspection system is an important means to eliminate hidden dangers of motor vehicle safety and reduce traffic accidents. [0003] According to my country's traffic laws and regulations, if a vehicle breaks down during driving, it is necessary to install a triangle warning sign to remind the vehicle coming from behind within a certain distance; and the "Technical Conditions for Motor Vehicle Operation Safety" stipulates that medium-sized and above passenger vehicles should be equipped with fire extinguishers. There are no mandatory regulations, but there a...

AI Technical Summary

Problems solved by technology

[0005] In the existing target detection process, the maximum frame method is often used to determine the key target of detection. The maximum frame method is simply to use the target frame to mark the target on the picture during detection, and which target uses the largest target frame, which target is Focusing on the target, it is very easy to make wrong judgments, because in the actual vehicle inspection scene, the camera does not necessarily face the front of the target vehicle, but often tilts 45 degrees to shoot the vehicle, so when other non-target vehicles are on the right Even if the safety warning devices are placed on the other side, even if these devices are not fully captured, the target frame of the warning device on the non-target vehicle may be larger than the target frame of the warning device on the target vehicle due to the perspective relationship during shooting, so that The degree of shooting of non-target vehicles is greater than that of target vehicles, thus affecting the accuracy of the results

Images