Improved Watershed Segmentation for Apple Fruit Grading

Abstract

Fruit grading technique will rank the images using their shapes, color and outer look. For that we present an improved watershed segmentation method based on multi-scale edge detection. Multi-scale edge detection method which is the weighted combination of morphological operations and it applied to fruit Image. First the image is given to multi-scale edge detection method and the result is given as input for watershed segmentation. Normal watershed segmentation will result in over segmentation and the fruits cannot be graded with the segments. Multi-scale edge detection will improve the small edges and convert the edge regions and other regions with different colors. Then region merging is performed to get the final result. Result show that this method can segment all part of defects in the fruit image. The fruits are classified into ripe and unripe bunches which are reflected by red, green and blue (RGB) color information. The fruits will then be sorted accordingly. This automated grading system increases accuracy, quality and consistency of fruit grading which can standardize the grading criteria.