Detection of corn stalk granulator's blockage with modified YOLOv4
This study aimed to develop a new method based on modified YOLOv4 network for detecting blockage fault with corn stalk granulator machinery in view of low precision and efficiency of traditional method.First,with YOLOv4 network as a fundamental target detection algorithm,a Mixed Attention Module(MA)was introduced to develop a YOLOv4-MA network featuring strengthened weights on small-target channel and spatial characteristics.Then,loss function in YOLOv4 network was replaced with Focal-EIOU function to raise target detecting precision.Finally,the modified YOLOv4 network was applied to de-tecting blockage in corn stalk granulator.The results showed the algorithm's MAP value in detecting different stalk granules was 98.51％,which was 16.33％and 20.05％higher than the traditional SSD algorithm and Faster-RCNN algorithm,respec-tively.Furthermore,this algorithm took only 0.024 s in detecting the stalk granule target,faster than the other two models.It can be concluded that the algorithm can improve minute stalk granule detecting precision and stalk granulator blockage fault,and satisfy the system detection demand with high practicability and efficiency.
万方数据
维普
