Research And Implementation Of Vheicle Detection Algorithms Based On Fisheye Image

With the rapid development of society and economy, people's living standard has got great improvement and the number of vehicles has quickly increased. The transportation hugs, traffic accidents and environment pollution increasingly threaten our environment and economy development. So how to enhance traffic efficiency and avoid traffic accidents effectively has received considerable attention in the field of ITS (Intelligent Traffic Systems). Vision-based vehicle detection is one of the most important and fundamental techniques. Therefore, research on it is very essential and meaningful.Fisheye camera is a means of acquiring external environment information. A vehicle detection system based on fisheye image is designed from the practical point of view in the thesis. This paper analyzes the current situation of vehicle detection algorithms based on visual system. By means of comparison of various algorithms, the existent issues of algorithms are expounded. A calibration algorithm based on cylindrical model for fisheye image is proposed and vehicle detection algorithm will be achieved on the calibrated image. Then the multi-features fusion is proposed after summarizing the applicability of some vehicle detection algorithms. The contrast sensitivity of shadow underneath vehicle is used for vehicle segmentation in the image and the features of a vehicle such as shadow underneath vehicle, edges of a vehicle, symmetry are used to locate accurate position of a vehicle. The algorithm is mainly used to detect vehicles in the daytime. Whereas light ray in the night is dark and features of a vehicle are fuzzy. The vehicle detection algorithm based on headlights of a vehicle is introduced and the headlight detection algorithm based on morphology is adopted. The algorithm is mainly used to detect vehicles at nighttime.The vehicle detection algorithm is evaluated in different video sequences. Experimental results have shown that the algorithm has effective performance and excellent robustness on vehicle detection.