Research On Real-time Detection Of Dynamic Target In Driverless System

The research content of this paper is mainly focused on the low-speed running environment with low flow density of the traffic participants,while taking into consideration the requirements of the unmanned system for the accuracy and real-time of the environmental target perception.Aiming at the traditional target sensing technology in the system,this paper studies the detection technology of static and dynamic targets in monocular vision from the perspective of image processing.First of all,this paper has carried on the study and the research to the lane line detection technology.By using pre-processing methods such as camera calibration and color space conversion,the interference information which is not related to the road in the image is removed,so as to reduce the data processing capacity and shorten the data processing time.This paper uses different images with salt and pepper noise to test and compare different filtering methods.Among them,the ideal median filtering time for noise reduction is about 1.018 s,while the Gaussian filtering takes about 0.325 s,which is shorter and can be used.And it can get a good balance between removing noise and preserving graphic information.Secondly,this paper compares the commonly used Sobel method and Canny method to observe the experimental results.On this basis,this paper adjusts the threshold parameters of the Canny edge detection method,and changes the threshold ratio to 3:1 as the detection parameter.Then,the lane edge is regarded as a triangle area,and the key information is extracted.Finally,this paper uses the improved Hough transform method to detect linear lane lines in images compared with the Hough transform with large computational complexity.The experimental results show that the method used in this paper can detect the road lanes well.Aiming at the target detection technology in dynamic scene,this paper first studies the static target detection technology in monocular vision,and uses Hog algorithm to extract target features.In order to reduce the rate of missed detection and improve the detection speed,Haar features are combined.Then the classifier is used to screen the extracted features in order to achieve a better detection effect.Finally,the research on dynamic target detection is carried out on the basis of static method.Firstly,the Hog feature of single channel is studied,Hog parameters are analyzed and modified,and the method of non-overlapping is realized to achieve the purpose of reducing the dimension of the algorithm.At the same time,the virtual machine is installed,the Ubuntu system is chosen as the experimental platform,and the algorithm of dynamic target detection is optimized by using the characteristics of high optimization and strong compatibility.