Autonomous lemon grading system by using machine learning and traditional image processing

Fruits like lemons are always harvested in huge quantities. However, currently, the assessment of lemon quality as best, good and bad, which based on the color change in the surface of the lemon, is still being used manually by humans. This leads to low productivity as well as uneven grading quality. This study will present an effective method that can help the classifying has high accuracy and fast grading time that could meet industrial needs by using a conveyor belt which makes lemons move and rotate at the same time and machine learning, traditional image processing. By moving and rotating continuously on the conveyor belt, the machine learning algorithm using Yolov4 network and traditional image processing will examine the whole surface of the lemon to classify the current state of the lemon. This methodology has the advantage of being able to check the entire lemon surface and the number of lemons to be classified at the same time is not affected too much by the image processing time. Beside the number of image collection for YoloV4 also reduced significantly. These advantage are very important in real implement at the factory. The efficiency of the system will be proven through series of experiments on real machine by comparing the performance of the current machine learning, traditional method with the proposed method.