Image Processing Techniques For Tomato Segmentation Applying K-Means Clustering and Edge Detection Approach

摘要

The implementation of image processing techniques in the plantation field has been extensively researched and developed, for example, to identify fruit maturity and control fruit harvesting robots. The main procedure, termed segmentation, is required by those systems in order to determine the fruit and background area. This work aims to put into practice a method of tomato segmentation. The method consists of four main processes: region of interest (ROI) detection, pre-processing, segmentation, and post-processing. The resize and K-means clustering were applied in ROI detection. The color space conversion of RGB into HSV was applied in pre-processing, followed by implementing edge detection using the Canny operator. In post-processing, morphology operation was carried out to discard the remaining noise. The performance evaluation of the tomato segmentation method against 300 images showed the average value of segmentation accuracy, false positive and false negative obtained reached Sc,, and 91.43%, 2.84%, and 4.77%, respectively.