摘要
针对带式输送机落料口堆煤检测存在准确性、实时性和可靠性差的问题,提出一种基于机器视觉的带式输送机落料口堆煤检测方法,利用工业相机采集落料口图像,通过以太网传输给堆煤检测器,堆煤检测器利用堆煤检测算法对图像进行处理与分析,实现堆煤检测,并通过以太网将检测结果传输给上位机。改进了ShuffleNetV2网络模型,用空洞卷积替代模型中的标准卷积,在不增加计算量的前提下增大感受野,在基本单元中增加了高效通道注意力模块ECA,提高特征提取能力。提出了改进ShuffleNetV2网络模型的堆煤检测算法,利用暗通道的图像去雾增强算法对采集的图像进行去雾和增强处理,构建了落料口图像数据集,利用该数据集训练改进ShuffleNetV2网络模型,再利用训练好的网络模型进行堆煤检测。采用Cortex-A57架构4核JetsonNano开发板设计了堆煤检测器的硬件和软件。研究结果表明:该方法能够实现带式输送机落料口堆煤的实时检测,准确率达到98.34%,图像处理速度为23帧/s。
Aiming at the problems of poor accuracy,real-time and reliability of coal pile detection at the belt conveyor drop port,a machine vision-based coal pile detection method for the belt conveyor drop port is proposed.An industrial camera is used to capture the image of the drop port and transmit it to the coal pile detector through Ethernet,and the heap coal detector processes and analyzes the image to achieve coal pile detection by using the coal pile detection algorithm,and transmits the detection results to the host computer via Ethernet;the ShuffleNet V2 network model is improved by replacing the standard convolution with the null convolution in the model,which increases the perceptual field without increasing the computation,and the efficient channel attention module ECA is added to the basic unit to improve the feature extraction capability;the heap coal detection algorithm with improved ShuffleNet V2 network model is proposed,and the image dehazing enhancement algorithm of the dark channel is used to dehaze and enhance the collected images,the image data set of drop port is constructed.The improved ShuffleNet V2 network model is used to train the improved ShuffleNet V2 network model,and then the trained network model is used for coal pile detection.The heap coal detector hardware and software were designed using a Cortex-A57 architecture 4-core Jetson Nano development board.The experimental results show that the method can achieve real-time detection of heap coal at the belt conveyor dropout with an accuracy of 98.34%and an image processing speed of 23 frames/second.
作者
苗长云
李佳
MIAO Changyun;LI Jia(School of Electronic and Information Engineering,Tiangong University,Tianjin 300387,China)
出处
《辽宁工程技术大学学报（自然科学版）》
北大核心
2023年第5期617-624,共8页
Journal of Liaoning Technical University (Natural Science)
基金
国家自然科学基金项目(NSFC51274150)
天津市重点研发计划科技支撑重点项目(18YFZCGX00930)。
