A robust fruit image segmentation algorithm against varying illumination for vision system of fruit harvesting robot

Abstract

Vision system is the crucial component of fruit harvesting robot for recognising fruit, however, which is seriously affected by varying illumination when the robot works in real natural environment. A robust fruit segmentation algorithm against varying illumination for vision system was proposed with the aim of effectively extracting fruit object in the natural environment. The method involved the application of improved wavelet transform to fruit image to normalise illumination of object surface. Then Retinex-based image enhancement algorithm was used to highlight fruit object of illumination normalised image. Finally fruit image was segmented by implementing K-means clustering. Three kinds of fruit images of different colour under sunny and cloudy days were segment using the proposed method respectively and the experimental results showed that the proposed algorithm could be robust against the influence of varying illumination and precisely segment different colour fruits.