Fruit Detection and Classification Using Computer Vision and Machine Learning Techniques

Within the recognition, classification, and determination of the quality of the fruits, the methods to process the information and the observed characteristics acquire importance, especially if it is sought to reduce the time, the work invested and reduce human bias. To cover these factors, the application of artificial intelligence and artificial vision within the processes is necessary. The objective of this work is to develop an artificial vision system through the OpenCV library and image database, which enables an accurate and quickly fruit classification, trained with machine learning algorithms. The region of interest (ROI) inside the image is located using a Canny edge detection algorithm, and the features of the fruits are then extracted. A k-nearest neighbor (KNN) algorithm is then trained to classify the fruit using its circularity and color features. A K fold cross validation is then used to test this method. It has a high degree of precision, the F1 score depending on the fruit, can range between 83% and 98%, which shows that positive results can be obtained using these technologies for agricultural applications.