基于改进 YOLOv5s的烟叶育苗盘生育状态评估算法OA

In order to solve the problem of continuous intelligent monitoring of the growth state of the entire tobacco seed-ling tray as a community throughout the entire growth period,an algorithm for evaluating the growth status of tobacco seedling trays based on the improved YOLOv5s object detection model from the perspective of"individual"to"popula-tion"was proposed.Firstly,tobacco leaf area of the whole tray was extracted based on the normalized Excess greenness index(ExG)and Otsu's thresholding(Otsu),and the linear mapping relationship between the leaf area rate and the progress of whole tray seedlings was analyzed.Then,the leaf area within each block was extracted,and the degree of uni-formity was estimated by analyzing the distribution of leaf area rate using the Gini coefficient.Secondly,by introducing the input image segmentation strategy to solve the problem of small target in large image detection,by introducing SimAM at-tention mechanism into the YOLOv5s model structure to enhance the aggregation ability between image features,by intro-ducing SPD-conv downsampling strategy into the YOLOv5s model structure to reduce the loss of fine-grained information and enhance the feature representation of low-resolution images,then train the improved YOLOv5s target detection mod-el.The test results showed that the performance was significantly improved after the input image segmentation strategy,and mAP@0.5∶0.95 was further improved to 80.7%after the model structure was improved,which was 6.9%higher than that of the YOLOv5s model before the structural improvement.The improved YOLOv5s algorithm can effectively monitor the growth status of tobacco seedlings in seedling trays,and provide technical support for the realization of intelligent management of tobacco seedlings.