The Design and Evaluation of an Orange-Fruit Detection Model in a Dynamic Environment Using a Convolutional Neural Network

Agricultural robots play a crucial role in ensuring the sustainability of agriculture. Fruit detection is an essential part orange-harvesting robot design. Ripe oranges need to be detected accurately orchard so they can successfully picked. Accurate fruit significantly hindered by challenges posed illumination and occlusion fruit. Hence, it important detect dynamic environment based on real-time data. This paper proposes deep-learning convolutional neural network model for orange-fruit using universal dataset, specifically designed complex environment. Data were annotated dataset was prepared. A Keras sequential prepared with layer-activation function, maximum pooling, fully connected layers. The trained then validated test assessed image acquired from Kinect RGB-D camera. run its performance evaluated. proposed CNN shows accuracy 93.8%, precision 98%, recall 94.8%, F1 score 96.5%. mainly affected leaves orchard’s trees. Varying another factor affecting results. Overall, orange-detection presents good results effectively identify environment, like orchard.