A ROBUST APPROACH TO OBTAIN A CROP EDGE BASED ON WAVELET FILTER AND FUZZY RECOGNITION

摘要

This paper intended to investigate an innovative method of detecting crop cut-uncut edge for agricultural vehicle guidance systems. Unlike traditional image segmentation-based approaches, this approach tries to recognize a step at the edge position in pixel intensity curve. It includes thresholding strategy by a histogram statistic, pixel intensity curve smoothness by wavelet filter algorithm, step recognition by a fuzzy pattern recognition algorithm, and edge-line extraction by a linear regression method. The approach has been used to extract edge-line of paddy field with a RMS intercept error of 7 pixels, and the RMS slope error of 1.11°. The validation results show that this method is capable of getting crop cut-uncut edge reliably.