Binocular vision vehicle environment collision early warning method based on machine learning. (27th July 2020)

• Record Type: Journal Article
• Authors: Mi, Hong; Zheng, Ying
• Is Part Of: International journal of vehicle information and communication systems. Volume 5:Number 2(2020)
• Page Start: 219
• Page End: 230
• Publication Date: 2020-07-27
• Subjects: machine learning -- binocular vision -- vehicle environment -- camera -- classifier -- threshold value; Automobiles -- Electronic equipment -- Periodicals; 629.27
• Languages: English
• ISSNs: 1471-0242

Abstract

Because the existing early warning methods do not assign weights, it is easy to cause collisions in the vehicle driving process, and the prediction accuracy is low. Therefore, this paper proposes a binocular vision vehicle environment collision early warning method based on machine learning. The comparison of experiments on high-speed sections shows that the number of vehicle collisions decreases by about six times when using the proposed method in this paper is used, which is significantly less than that of the existing methods. Moreover, the distance error between the target vehicle and the running vehicle measured by the method in this paper is small, and the error rate is between 0.005 and 0.041. Therefore, it can accurately warn of the occurrence of vehicle collisions, and its application advantages are obvious.