The Research On Global Path Planning And Warning System Of Intelligent Vehicle

Aiming at urban traffic congestion and rear end accidents,a multi-objective path planning algorithm based on improved ant colony algorithm is proposed from a global point of view,which not only takes the shortest distance as the evaluation standard,but also considers the smooth condition of the road,to find a reasonable path under the condition of avoiding congestion path,improves the applicability of ant colony algorithm.Appling machine vision technology from the consideration of local driving environment to prevent the rear end collision,the developed front vehicle collision warning system determines the target area of interest through lane line detection and identifies the vehicles in front of the lane in the area which are close to each other.Experimental results show that the scheme can provide technical support in reducing time cost and assisting drivers.The main research contents and results are as follows:(1)Optimization of the ant colony algorithm.The improvement process mainly includes the optimization and adjustment of heuristic information,and the environmental factors are employed to enhance the ability of reading map.The results suggest that the ant deadlock rate is reduced by 78%.The computing efficiency,planned path distance,and inflection point are obviously improved.At the same time,according to the congestion degree,the grid map is redrawn,and the optimal probability transfer algorithm is combined to achieve the shortest path acquisition under the congestion avoidance path.(2)Method of lane line recognition.The perspective of the lane in front of the vehicle is transformed by inverse perspective transformation.The lane lines are extracted in different color models according to the color characteristics of the lane lines.In order to improve the stability of lane line detection and reduce the impact of environmental interference on lane line recognition,this thesis proposes a second detection of lane line edge information.Through the fusion of the above two detection results,a binary map of lane lines is obtained.By the sliding window,the white pixels in the binary map containing lane lines are counted,and the white pixels are fitted to the lane lines using the least square method.Finally,the detected lane lines are mapped to the original image based on the inverse perspective transformation matrix.The experimental results show that the algorithm in this thesis can accurately detect lane lines on straight and curved roads with high detection rate and good robustness.(3)Front vehicle detection based on target area.On the basis of lane linerecognition,a reasonable front vehicle recognition area is selected according to the lane line.In the process of training,according to the frequent environmental information in the region of interest,the negative samples are divided into focus negative samples and no focus negative samples are introduced to the front vehicle detection training algorithm based on Haar and Adaboost,namely,the first step focus on the positive samples and the key negative samples of training,and the second step trains positive samples and non-key negative samples,and then cascades the training results to generate detection files.(4)Verification of anti-tailing warning algorithm.Through the experimental analysis on urban roads,the results show that the algorithm can effectively mark the front vehicle at a short distance to remind the driver to prevent rear-end collisions.