CLASSIFICATION OF TOMATO QUALITY BASED ON COLOR FEATURES AND SKIN CHARACTERISTICS USING IMAGE PROCESSING BASED ARTIFICIAL NEURAL NETWORK

Abstract

Tomato (Solanum Lycopersicum) is a plantation commodity in Indonesia with a production rate that tends to increase every year. With a high economic value, maintenance is important so that the quality is getting better. The problems that arise at this time are related to the determination of the quality of tomatoes which is still done manually and depends on humans so classification using technology is considered important to be developed. Previously there has been researching related to the classification of tomatoes. However, accuracy and computation time still need to be improved. Therefore, in this research, a method of classification of tomatoes was carried out using Artificial Neural Network (ANN) Backpropagation algorithm by utilizing color features and skin characteristics based on image processing. This research followed several stages, from acquiring 300 tomato images with 3 class levels to the classification process using ANN Backpropagation. Several training scenarios and tests were conducted to select the feature combined with the highest accuracy and fastest computation time. The combination of 3 best features used is RGB color feature with shape and texture features as skin characteristic parameters. Based on training results with 210 training images, an accuracy of 100% was obtained with a computation time of 2.58 seconds per image. While test results with 90 test images, accuracy reaches 95.5% with a computing time of 1.39 seconds per image. So it can be concluded that the method used has gone well in classifying tomato image quality based on color features and skin characteristics.