SDCM: An Efficient Camellia Oleifera Fruits Detection Algorithm in Dense Occlusion Scenes

To realize intelligent management in the Camellia oleifera fruit industry and foster the efficient growth of its supply chain, the accurate identification and detection of Camellia oleifera fruits are essential. However, in natural environments, these fruits exhibit challenges such as clustering, mutual occlusion, and obstruction by leaves. This paper proposes a Camellia oleifera fruits detection algorithm based on an improved YOLOv7 deep convolutional neural network, which is referred to as the SDCM network model, and add the function of counting the number of fruits. We introduce an enhancement to the backbone layer of the YOLOv7 model by incorporating a module called SDCM (Shortcut and Dilation Convolution based Module). This module enriches multiscale edge information, enabling the network to extract features more effectively, thus improving detection accuracy. The experimental results show that the detection accuracy of the SDCM network model is significantly improved compared with the original YOLOv7 model, and the detection performance of the model in this paper is superior compared with other classical detection models. By enhancing the detection accuracy of the model and adding the function of fruit counting, the model in this paper can provide more reliable and accurate detection performance and the function of predicting fruit yield for the development of the Camellia oleifera fruit industry, improve the production efficiency, and reduce the artificial resources.