Navigation line detection based on support vector machine for automatic agriculture vehicle

Agriculture image segmentation plays an important role in the agriculture vehicle navigation. Robust segmentation gives a better influence on the extraction of navigation parameters. The color image was converted into gray scale image, and in order to obtain more crop row information, the average image of the gray scale one based on rectangle weight module were realized. The standard deviation of every pixel was computed too, to remain the origin crop row width. The average and the deviation value were fused and as a new input factor of support vector machine to segment image. In order to save running time of algorithm, all the operation were executed on the low resolution image obtained through wavelet analysis. The result of segmentation obviously conquered the influence of broken ridges, weeds and others high frequency disturbs. The running time is less than 0.6s when the program was written in Matlab.