Study On Safety Exit Warning System Based On The Rear View Camera

In recent years, with the rapid growth of car ownership in China, the total amount of traffic accidents also increased year after year. How to reduce traffic accidents and reduce the loss have become an issue of growing concern caused by traffic accidents. Automotive safety driver assistant technologies, as an effective means to reduce traffic accidents, reduce accident losses, have got the governments’ more and more attention, and they have become a hot research because many scholars from various countries and the auto companies are studying on them.Safety exit warning system based on the rear view camera studied in this paper is not only an important part of automotive safety driver assistance technologies, but the first case in our country. It combines advanced image processing and computer vision technology, detects the moving object in the back area covered by both sides of the opening-doors and car body, and warns the drivers of the potential danger before opening the door, and not to open the door to prevent the rubs or collisions caused by the bicycles, motorcycles, and cars which suddenly pass closely. It can not only meet the urgent demand for advanced driver assistance technologies, but also meet consumers’ increasing requirements for car safety.According to the relevant principles of image processing and computer vision, this paper focuses on the research and application of the following technologies:Image postprocessing technology. This paper not only uses the traditional image preprocessing techniques such as mean filter and median filter, but combines optical flow threshold and morphological denoising algorithm based on the characteristics which optical flow is sensitive to noise. This technology can reduce the effect of noise to the minimum, and improve efficiency and accuracy of the moving object detection algorithm.Moving object detection technology. The techniques which are widely used in the moving object detection include: inter-frame difference method, background subtraction method and optical flow. Through experimental comparison of these three methods, the paper adopts the optical flow method to detect moving objects, and introduces two classic optical flow calculation methods: HS algorithm and LK algorithm and finally choose LK algorithm.Warning strategy. Warning strategy is another key step in this system, the paper chooses distance model as a basis for alarm and defines four dangerous levels according to the distance, and are satisfied with the results of simulation.In addition, the paper defines the dangerous zone, selects cameras and other aspects of the research, and builds a simulation platform in MATLAB, so that the system can meet the requirements of this study.