A robust farmland edge detection method combining anisotropic diffusion smoothing and a Canny edge detector

Farmland mapping is relevant to food security. Recently, farmland information extraction has become a hot topic in agricultural research. The farmland pattern recognition from high-resolution remote-sensing imagery faces a big challenge caused by complex ground environment. In this work, a novel farmland border detection method based on the multi-scale, anisotropic diffusion algorithm and Canny edge detector is proposed. The derived anisotropic diffusion algorithm is presented as a preprocessor to generate the ideal scale space. By combining the canny edge information at fine scale with gradient magnitude at the coarse scale, this method performs better than other segmentation methods, as it can meet the location and noise removal demand. The effectiveness of this approach is demonstrated by several image examples.