Detection method and validation analysis based on the improved YOLOX-s apple blossom growth state

In order to realize intelligent blossom thinning in apple orchards, this paper proposes an improved YOLOX-s apple blossom growth state detection method. First, the apple blossom dataset was collected and established for the training and verification of the network model, and then the YOLOX-s model was built, and the backbone network was improved. The Convolutional Block Attention Module（CBAM） attention mechanism module was added to the two layers after the output features, EIOU was used as the regression function of the model and the Focal Loss function was introduced in the post-processing stage to improve the model’s ability to detect clustered apple blossoms and improve the average accuracy of the model. The results showed that the accuracy of the improved YOLOX-s model was 91.75%, which was 0.5% higher in precision, 6.19% higher in recall rate, and 4.28% higher in average precision than before. This research provides technical support for the realization of intelligent detection of apple blossom growth status to guide the accurate decision of intelligent blossom thinning.