Automation in agriculture enhances productivity, sustainability, and the national economy. The main emphasis of this study was on fruit sorting. The fruits selected for the project included lemon, peach, mango, banana, apple, and orange. The initiative aimed to implement automation in the sorting and categorising process to minimise post-harvest human errors. The assembled mechanical system comprised a conveyor belt, actuator, and image-capturing chamber. A camera was utilised to capture the digital photos. Image blur could be eliminated by developing a system capable of capturing the image before the multi-class fruits are inserted into the conveyor belt. The study employed image processing techniques to gather valuable attributes for the purpose of fruit classification. ResNet and transfer learning algorithms were applied to acquired images. The RGB colour model was employed to choose the colours. The noise was diminished after eliminating the foreground. The following data was obtained: the area of the fruit, the average skin colour, the global standard deviation of the individual colour channels (Red, Green, and Blue), differences in contrast enhancement, and the local standard deviation of the three colour channels. A Back-Propagation Neural Network (BPNN) was provided with nine inputs. To train the neural network, a total of 250 samples were collected from 10 distinct categories, including ripe, healthy, as well as faulty and unripe samples. The approach achieved a classification accuracy of approximately 97%.

Index Terms- About four key words or phrases in alphabetical order, separated by commas. Keywords are used to retrieve documents in an information system such as an online journal or a search engine. (Mention 4-5 keywords)