Study On Technology Of Ground Automatic Identification Of Intelligenced Vehicle

Technology of ground automatic identification of intelligenced vehicle in this paper is the technology of independent recognition for image information based on vision sensor.It has broad application prospects in the intelligent mobile robots field and the military industry field.In this design, the test vector is the educational intelligent robot Leobot-Edu of the Shanghai Zhong wei company. Daheng DH-HV2003UC-T vision sensor is installed on the top of the body and do some acquisition of the ground image information for the five common driving road (cobbled road, concrete road, dirt road, grass road, tile road). In this design, taking harvest for six times for each road, taking one of the best quality image information as the training sample and taking the other five groups of image information as the testing sample. Then taking compression code, re-image restoration, image smooth, image sharpen, image enhancement, image feature extraction and other related processes in turn for the road image information of the five common image in the training sample in the use of MATLAB image processing module.In this design, linear predictive coding mode is applied in the image compression coding part; wiener filter reconstruction is applied in image restoration part; low pass filtering and median filtering method is applied in image smoothing part; image fuzzy approach is applied in image sharpening part; histogram modification approach is applied in image enhancement part; extracting six typical image features of 8 connected boundary characteristics, 4 connected boundary characteristics, characteristics of the area ratio, features of euler number, corner features, moments and others in the image feature extraction part. And then taking pattern recognition in the use of the BP neural network module of MATLAB. In this design, BP neural network pattern recognition module is divided into three parts: Network establishing based on the newff function; Network training based on the train function; Network testing based on the sim function.On the basis of the experimental studies above, the design will design a set of autonomous vehicle prototype machine vision system including image acquisition module, image processing module and pattern recognition module.The analysis of the results of pattern recognition shows that: the network objective training error is 20 percent,and the road recognition rate has reached the intended requirement in the system,and you can universally accesse in the smart vehicle, robots, teaching experiments and other related fields.