Highly Efficient Machine Learning Approach for Automatic Disease and Color Classification of Olive Fruits

Abstract

After analyzing and assessing a large number of previous studies of olive fruit classifications, the following conclusions are derived: first, most of these studies relied on their technology being applied to a database, which often contains few images and is not enough to confirm the efficacy of the suggested technology, in order to verify the accuracy of their technology. Moreover, a large number of these studies depend on the utilization of an unrelated image library. All of the images are prepared for testing since they each show a single fruit with a background that is completely different from the color of the fruit. As mentioned earlier, this problem has nothing to do with reality. When using it practically, one has to cope with a frame that contains hundreds of fruits. The fruits are placed on a conveyor with several channels to keep them steady. Moreover, it is noteworthy that most of this study included recommendations for functional technologies that remain amenable to development. Finally, it is important to emphasize that processing speed data is essential in this type of application and has not been collected in many of these experiments. The study presented in this presentation dealt with a new strategy for classifying the colors and diseases of olive fruits. The proposed strategy was based on dividing the separation process into more than one stage, for the purpose of scrutinizing the selection of the most appropriate feature extraction for each stage according to the type of separation process, and also checking the selection of the most appropriate classifier for each stage according to the type of separation process as well. Also checking the accuracy and efficiency of the separation process at each stage, by training, validating, and testing each separation process that took place at each stage. Also, consideration was given to choosing rapid processing techniques in order to complete the identification of the type of fruit in the shortest possible time. Also, this study relied on verifying the efficiency of the proposed technology by using a powerful database that contained about 15,000 images of olive fruits taken directly from the fruit conveyor. Efficiency was also confirmed by comparing our results with the results of related technologies. The results of the test used to compare the suggested technique’s accuracy to those of other methods for classifying eight different kinds of olive fruits showed that it was incredibly effective at classifying the fruits in the least amount of time. When the fruits were arranged on a white backdrop, the suggested method had an effectiveness of 99.26% for fruit classification. The most significant finding was that, in contrast to other methods, it was able to categorize fruits while they were being placed on a fruit conveyor with an efficiency of 97.25%. Additionally, the results show that fruit identification with the suggested method only takes a few seconds. All experiments in this paper were conducted using MATLAB (2016 b) on a computer running Windows 10 64-bit with an Intel 2.7 GHz processor and 8 GB of RAM. The high-performance correctness of the suggested strategy, after training it on about 9000 samples of healthy and damaged olive fruit images (real frames) and then testing it on about 6000 samples, indicates the high possibility of using it in computer vision applications.