Tobacco Maturity Recognition Based on an Improved YOLOv8 Algorithm

Tobacco maturity is an important factor in determining tobacco baking, and the existing means of tobacco maturity detection are usually based on manual detection, lacking objective and uniform detection standards. In this paper, a new object detection algorithm GSSS-YOLO is proposed based on the YOLOv8 algorithm. The GSSS-YOLO introduces the CPCA attention mechanism to raise the precision of the detection, and the loss function SIoU is used to speed up the model's convergence speed. Finally, the tobacco maturity recognition model is trained using nearly 3700 tobacco leaves collected in a real environment, and the results of experiment indicate that the proposed detection algorithm has good detection accuracy and detection efficiency. The proposed GSSS-YOLO algorithm can be deployed on an embedded platform to realize on-line tobacco maturity recognition, which is very important for improving the efficiency of tobacco harvesting.