A deep learning-based automatic detection method and a deep learning-based automatic detection system for a thrown object on an expressway

Abstract

The invention provides a deep learning-based automatic detection method and a deep learning-based automatic detection system for a thrown object on an expressway. The method sequentially comprises a video image acquisition step, a target detection step, a target judgment step, an accident alarm step and an optimization upgrading step. The thrown objects in the monitoring video range are identified through a deep learning detection algorithm, and real-time detection can be carried out; a deep learning detection algorithm is continuously optimized and upgraded through a judgment strategy; and the detection accuracy is improved, so that accidents in the expressway are reduced.

Description

technical field [0001] The invention relates to the technical field of intelligent transportation, in particular to a deep learning-based automatic detection method and system for thrown objects on expressways. Background technique [0002] In recent years, the rapid development of expressways has brought convenience to people's lives. At the same time, accidents on expressways have become more frequent. The most important factor causing accidents is the thrown objects on the highway. big challenge. The traditional detection algorithm of spilled objects adopts the method based on image processing, which requires manual extraction of image features, which is not only time-consuming and labor-intensive, but also the detection effect is not ideal. The detection methods of spilled objects based on video sequences also use the traditional Gaussian model and the three-frame difference method to distinguish the foreground, and then further identify the spilled objects, but the rec...

AI Technical Summary

Problems solved by technology

[0003] The present invention aims at the problem that the accuracy rate of identifying spills in the prior art is not high and small spills cannot be identified. The present invention provides an automatic detection method for expressway spills based on deep learning. Through the deep learning detection algorithm, the monitoring The spilled objects within the video range are identified and can be detected in real time. Through the determination strategy and continuous optimization and upgrading of the deep learning detection algorithm, the detection accuracy is improved.

Images