Feature based Fruit Quality Grading System using Support Vector Machine

The conventional post-harvesting process of horticultural fruit and vegetables is exorbitant and highly subjective which susceptible to errors. This process is extremely unreliable in terms of accuracy, throughput, speed and cost. An automated fruit grading approach is desirable to reduce the manual efforts and increase the efficiency. In this paper, we are developing a computer vision based an image acquisition and classification system for fruit grading. This system is divided into three modules image pre-processing, image segmentation and Classifier. In the first phase, we are extracting the image features such as image quality, area, perimeter, mean, variance, color, intensity. Next Otsu algorithm works on automatic threshold calculation and image binarization part. In the second phase k-means clustering is used for image segmentation. In last phase the vectors of color, intensity, segments, and image quality feature were utilized for training of the support vector machine structure. The image k-means clustering approach is embedded with support vector machine classifier to improve the speed and accuracy of the classifier. As a result, the system will be able to classify the fruit into three different categories. The proposed algorithm is implemented with an example of tomato fruit grading. The experiments consider real tomato image dataset of 200 samples.75% of the total dataset is used for training and remaining 25% used for testing purpose. Training dataset is distinguished into three categories depending on the maturity of tomato.