Research Of Vehicle Detection System Based Gabor Filter Features

The research of computer vision promotes modern automatic drivingtechnology, and the vehicle detection is the key and the premise. Recently, moreand more researches use Gabor filters to detect vehicle being its successapplying on pattern recognition. For being of less effective design measure andits imperfect on efficiency and real-time, these has been an obstacle todevelopment of the vehicle detection.For practical need of vehicle detection, this paper will study how todetection vehicle with Gabor filters and SVMs. To get best robustness andreal-time, it put forwards a highly efficient and practical detecting system.Firstly, this paper analyses the character of the Gabor filter. In order to getthe character that has the best representation, its parameters are optimized. Thispaper put forward a Gabor filter optimizing method which can conform geneticalgorithm by increasing cluttered number and combine optimizing parameterand choosing filter, which not only optimizes the parameter but also minify thefilters, and also reduce the high dimensional about Gabor eigenvector, and solvethe large computing and storage problem.Secondly, an experiment is put in practice for compare effect of classifierbased SVMs and NNs. The result reveals that classifier based SVMs has mostefficiency.At last, a model is constructed about vehicle detection system, and we alsodiscuss effect of optimized Gabor filters and classification of classifier basedSVMs.