OPTIMIZATION OF APPLE EDGE DETECTION USING THE CANNY ALGORITHM IN DIGITAL IMAGE PROCESSING

Abstract

Digital image processing has rapidly developed in various fields, one of which is the analysis of visual objects such as fruits. Edge detection is an important technique for identifying object boundaries in images, which is highly needed in the process of assessing the quality of apples. Apples have high economic value and are widely consumed, making accuracy in their visual analysis crucial, especially for automatic sorting and packaging purposes. However, variations in color, lighting, and ripeness levels pose challenges in the edge detection process. This study aims to optimize apple edge detection using the Canny algorithm, combined with Manual Thresholding and Otsu Thresholding methods. The preprocessing stages used include grayscale conversion, contrast stretching, and sharpening. The research was conducted on 200 apple images, consisting of 100 fresh apples and 100 rotten apples. The results show that the combination of Canny with Otsu Thresholding provides better detection performance compared to Manual Thresholding. Perfect detection accuracy for fresh apples increased from 27% to 51%, and for rotten apples from 51% to 81%. The average accuracy of the Otsu method was 66%, still below the 75% hypothesis, which was influenced by lighting variations during image acquisition. In conclusion, the Canny method with Otsu Thresholding is quite effective, especially for detecting the edges of rotten apples. Further suggestions include testing with controlled lighting and integrating additional methods for more stable results.