Classifying the Size of Shallots Using Watershed Transformation Algorithm

This study aims to detect shallots that are close together and their classification based on diameter. The proposed system uses images taken from a moving conveyor belt with a distance between the camera and the object of 25 cm and a speed of 40 rpm with a frame rate of 30 fps and a resolution of 1920 × 1080 pixels. The total data used in this study is 267 frames, researchers divide it into 70% for training data and 30% for testing data. This research implements watershed transform segmentation for segmenting shallot images that are close together, then implements the bounding box method to detect shallots based on the results of segmentation and feature extraction using the pixel per centimeter ratio formula to calculate the diameter of shallots then classify them according to their diameter. The test results of this research provide a limitation for the bounding box area to be used. Blob areas that are less than 100 will not be labeled for the bounding box of the detected object. The training data value for RMSE is 0.292, and MAPE is 6.290%, while the testing results with video data produce an RMSE of 0.176 and MAPE of 3.507%. The MAPE value below 10% demonstrates that the model exhibits excellent performance in measuring shallots diameter.