Vanishing point conducted diffusion for crop rows detection

Abstract

A partial differential equation based diffusion is presented for crop rows detection. In the diffusion, the evolving direction is estimated through the vanishing point, which is one of global feature of row-crop images. According to the vanishing point, we generate the orientations of row crop textures, and then integrate the induced field of directions into an oriented diffusion. After processing the row-crop image with the new diffusion, we extract the crop rows from its black-white version using a morphological operation. Experiments on the real row-crop image data show the proposed diffusion can suppress the undesired interference more efficiently than the other diffusion when extracting crop rows.