In order to realize fast,accurate,and reliable detection of coal quantity on belt conveyors and improve the reliability and safe-ty of belt conveyors,vision-based non-contact coal quantity detection technology has been widely used.Aiming at the detection task of belt conveyor coal quantity,combined with specific industrial application scenarios,a vision-based coal quantity detection method was designed.The process of image capture,image processing,and image classification was used to predict the coal quantity.Design an im-proved MobileNet classification algorithm for image classification problems,use the space and the channel attention mechanism to sup-press the interference of invalid information,and use the CSP(Cross Stage Partial)structure to improve the expressive ability of the network.To solve the problem of uneven distribution of data labels,the down-sampling method was used to maintain the balance of input picture labels during training.Finally,the comparative experiment was conducted on a real coal quantity image dataset.The results show that the proposed improved algorithm has better performance in terms of accuracy,floating-point operations,and inference time.
维普
万方数据
