Research On Road Detection And Recognition Method For Unmanned Vehicle

According to the visual information, the unmanned vehicle can identify the driving area of a road correctly and the road detection and recognition system is the core of the vision system, after selecting the current frame road image from the car-camera's video streaming to do static analysis, to identify the posture of the vehicle and predict the path of the next moment. According to the issue above, this thesis has studied a road recognition algorithm which has strong adaptability, better stability and higher recognition rate, that has been a correct decision-making basis to the terminal control of unmanned vehicle.The main research works are as follows:(1) Studied an improved wavelet filter algorithm to deal with road images.Because road images collected from on-board camera contain a lot of noise were,for the sake of identifying the multi-features rules of the road rapidly and accurately, the thesis utilizes an improved wavelet filtering algorithm for image filtering processing and reduced the interference of noise effectively;(2) Studied an improved Canny edge detection algorithm to conduct road images edge detection. In allusion to defects of the traditional Canny edge detection algorithm which high and low threshold filter parameters depend on the manual setting and lacks of adaptability and real-time, used an improved Canny detection algorithm based on Otsu method of images processing and realized the high and low threshold value can be selected automatically, obtained(3) Studied an improved Hough transform detection algorithm. By dividing the image detection area, each sub areas proceeded Hough transform, and then connected the straight segment which adjacent sub region distance is less than the set value based on Hough parameter Settings rules, according to the screening of the Hough parameters, fitted these parameters using piecewise linear model fitting roads, to judge the type of the road and provide the basis of determining where the road ahead to next moment.Through a lot of simulation and experiment, the results and the data show that by the road recognition algorithm the thesis introduced and Houghparameters of road model can well perform the basic information of the roads,and under the complicated condition, the algorithm can be recognize the straight roads and the curved roads effectedly.