Maize tassels play a crucial role in maize yield and excellent breeding.To improve the detection accuracy and detec-tion speed of corn tassels during the tasseling period in complex field environments,this paper proposes a lightweight object detec-tion model YOLOv5-EMM that integrates efficient multi-scale attention mechanism.By embedding an efficient multi-scale attention mechanism(EMA)module,replacing the MobileNetV3 lightweight backbone network,optimizing the MPDIoU loss function,and using BiFPN to optimize the neck network,the YOLOv5 model is improved.The experimental results show that the YOLOv5 EMM network has an average accuracy of 97.8％,which is 2.7％higher than before improvement.The number of model parameters is as low as 3.31M,and the amount of computation is also reduced by 51％.The YOLOv5 EMM model has a good detection effect on maize tassels in complex field environments,and can provide technical support for monitoring maize growth.
关键词
目标检测/深度学习/玉米雄穗/复杂田间环境
Key words
object detection/deep learning/corn tassels/complex field environment
维普
万方数据