Recognition method for apple fruit based on SUSAN and PCNN

This study proposes a recognition method for apple fruit based on SUSAN (Smallest univalues segment assimilating nucleus) and PCNN (Pulse coupled neural network) to accurately identify and locate fruit targets. First, homomorphic filtering is used to conduct image enhancement by considering the influence of different lighting conditions on the segmentation effect, thus achieving light compensation. After an image is processed by R-G color differences in RGB color space, the apple image is segmented using the PCNN image segmentation method based on minimum cross entropy. In terms of prior knowledge of the maximum and minimum radius of the apple fruit, an improved random Hough transform method is used to detect the characteristic circle of the apple target; according to the edge of the apple target obtained by the SUSAN edge detection algorithm. Comparative experiments with different segmentation algorithms confirm that the algorithm of this study has outstanding performance in reducing the influence of insufficient light on the segmentation result. In 50 images, 93% of apples were accurately identified, which proves the effectiveness of the algorithm in this study.