Apple tree branch segmentation from images with small gray-level difference for agricultural harvesting robot

Abstract

Aiming at identifying for tree branches obstacle in navigation automatically and picking process of agricultural harvesting robots, an iterative threshold segmentation of apple branch images based on contrast limited adaptive histogram equalization (CLAHE) is proposed. Firstly, the RGB color space of the branch images are converted to the XYZ and I1I2I3 color space by transformation formula, and the X-Y and I2 color factor of the apple branch images are extracted to analyze their gray-level difference. Then, The CLAHE is applied to the images whose gray-level difference is not intensity before iterative threshold. Finally, the apple branches are segmented from the original images by the iterative threshold. To verify the validity of the proposed method, 100 testing images in size of 1280×960 pixels under different illumination are utilized to compare the proposed method with other famous approaches, such as OTSU and histogram algorithm. Experimental results show that 94% of the apple branch images are correctly recognized and the segmentation quality of the proposed method is better than other approaches, which implies that the proposed method is effective for the tree branch segmentation.