The Research Of Pedestrian Detection Technique Under Vehicle Autopilot System

The pedestrian detection technology has always been a research focus and difficulties in the field of digital image processing and computer vision.For now,the research of this technology has a major breakthrough,its research results have been widely applied in various fields.Not limited to the pedestrian detection,the technology was widely used in intelligent surveillance,robot vision and the auxiliary driving system.But by the rising as people demand,pedestrian detection technology has been used in the complex environment mostly.Therefore it's still lack of a rapid accurate and robust pedestrian lesson plan calculation method at present.This paper was focused on the study of pedestrian detection algorithm in the vehicle autopilot system on the basis of digital image processing and computer vision technology,combined with FPGA technology.In this paper,the first step is through the analysis and comparison of the image color space,we chose the CIE Lab color space to continue the image processing operations.And we introduced the concept of super pixels at the same time.Super pixels has faster processing speed and high border shooting rate.We also made a further improvement algorithm by optimizing the operation and extracting a small number of pixels of the original image to reduce computational complexity and realize the image segmentation algorithm on FPGA platform.The simulation results show that with similar accuracy,the improved algorithm is much faster than the SLIC algorithm when processing the same object.In terms of pedestrian detection algorithm,this paper studies the pedestrian detection algorithm based on single frame image.We study the principle of the classifier,such as the SVM,the decision tree and random forest.We also discussed the traditional pedestrian detection algorithm based on SVM+HOG model in this paper.And based on the above theory this paper proposes a faster pedestrian detection algorithm based on random forest+SVM model.The algorithm extracts the target area according to the random forest classification effect,and then use SVM to do further testing to realize the pedestrian detection function.The experimental results show that the algorithm is significantly improved on the detection time,and the detection error rate is also greatly reduced compared with traditional pedestrian detection algorithm.