Detection of potato tubers using an ultraviolet imaging-based machine vision system

A machine vision system based on ultraviolet imaging was developed to detect potato tubers on the potato harvester. This vision system is expected to control a sorting mechanism which would remove clods and unwanted potato tubers, especially small tubers, flowing along with sound tubers before entering a storage container. The detection was based on the ultraviolet reflectance of the tubers compared to their background which includes pieces of clods. An algorithm for detecting threshold values between the tubers, the clods and the conveyor automatically was developed by smoothing the original intensity histogram until the ultimate peak for each kind of object was found, a procedure which could overcome the variation in the intensity values of the objects due to the difference in the lighting conditions and the water content of the clods. Since the algorithm depended on continuous comparison between the tubers and their background, a simple calibration method was applied to provide accurate detection of the objects by placing a stationary tuber in the field of view of the system. Besides, regulating the aperture of the ultraviolet camera lens was necessary to aid the detection of the tubers when their surfaces were covered with layers of dust or mud. Segmenting the tubers was accomplished by estimating their sizes through the calculation of their maximum length and width. The vision system was tested by taking a video of tubers and clods passing on the conveyor of a potato harvester using the ultraviolet camera. One video frame was taken from the video stream each second and the tubers within the frame were detected using the developed algorithm. 1171 video frames which included 2233 tubers and 1457 clods were segmented. The results showed that 98.79% of the tubers and 98.28% of the clods were detected successfully. Furthermore, the processing time required to segment the objects in each frame was approximately 94 ms which indicated the ability to apply the developed vision system at real time within the normal speed of the conveyor of the harvester. (C) 2009 IAgrE. Published by Elsevier Ltd. All rights reserved.