Defect Detection for Tomato Grading by use of Six Color CCD Cameras

Abstract

Grading systems to inspect various agricultural products using imaging technologies have been increasingly introduced and utilized in Japan recently. Among many kinds of fruits, tomato is one of the most difficult fruits to be inspected by image processing because of following reasons: 1) tomato fruits whose peel colors vary from almost green to completely mature red are shipped depending seasons. 2) defects whose colors are similar with fruit peel and whose sizes are very thin like “stitching” also should be detected. 3) defects, which are similar with color of immature fruit peel at shipping time, sometimes become conspicuous by its different color from mature fruit peel color at purchasing. 4) calyxes are sometimes have similar colors with cracks around them and their shadows may be misunderstood as defects. 5) defect type depends on fruit side and each defect level is different.
In this paper, a tomato fruit inspection system with six color TV cameras is described and its experimental results to grade tomato fruits with many types of defects are discussed. From the top view, “blossom end” and general defects were recognized by color and detected edges. Those defects were properly judged through their maximum lengths and area. From the side views, “stitching” and other defects detected, but it was considered that an algorithm to match side views with their top view should be investigated for a further precise measurement of the defects. From the bottom view, it was observed that calyx recognition and crack detection sometimes were difficult because the calyx size, color, shape and freshness changed from time to time and from fruit to fruit. To understand defect type and level, the resolution of image sensor and gray level should be increased and the recognition algorithm should be improved.