Research On Image Monitoring Technology Of Bridge Crane Machinery Based On DNN And Edge Cloud

The existing measurement and monitoring system of large-scale lifting machinery relies on complex sensors installed on the body to measure real-time parameters so as to judge the running state of the crane,which is not easy to install and dismantle on site.Therefore,this paper takes"Research on image monitoring technology of bridge crane running state based on DCNN and edge cloud"as the topic,designs the overall framework of image monitoring of crane running state,studies DCNN recognition and visual measurement method of key components of bridge crane,designs edge cloud system for image monitoring,studies DCNN acceleration lifting technology for image monitoring,and develops bridge crane running state monitoring system Heavy machinery running state image monitoring platform,carry out relevant error analysis experiment.This research has important academic value and practical significance for promoting the development of equipment health and safety detection and intelligent detection technology.This paper is supported by the 2019 Science and Technology Project（2020JD09）of the Guangdong institute of Special Equipment Inspection and Research and the 2020 Research&Development Projects in Key Areas of Guangdong Province（2019B010154003）.This paper studies the key technology of image monitoring of bridge crane operation status,analyzes the research progress at home and abroad from the aspects of safety monitoring technology of bridge crane equipment,image sensing technology of operation status monitoring,heterogeneous image transmission technology and so on,and determines the research content.The specific work includes:⑴Overall architecture design of bridge crane running state image monitoring.This paper puts forward the requirements of crane image monitoring,and designs the overall functional framework of the monitoring system,including image acquisition module,key component identification module,state measurement and early warning module,remote service provision module,remote user interaction module and real-time promotion module,so as to realize the remote and timely warning of the running state of multiple cranes in the plant,and remote setting of camera parameters machine monitoring system.⑵Research on DCNN recognition and visual measurement method for key components of bridge crane.A DCNN model for identification of key components of bridge crane is proposed;a method for reconstruction of world coordinates based on key point identification of DCNN model and Zhang’s camera calibration technology is proposed.According to the geometric relationship of components and reconstruction coordinates of key points,the world coordinates of key points of components,horizontal displacement of trolley l_car,inclined angle of ropeθ_ropeand inclined angle of slingθ_sling are calculated;the calculation error of world coordinates of key points is estimated.⑶Design of edge cloud system for image monitoring.Select the hardware parameters of the edge cloud system,design the edge cloud software system based on Azure Io T Edge based on the OPNET software for image monitoring edge cloud system requirements analysis and Azure Io T Edge computing software framework principle.According to the status of server resources and the number of crane monitoring units in the plant,the computing resources and network resources are automatically allocated to improve the real-time performance of crane monitoring.⑷Research on DCNN acceleration lifting technology for image monitoring.Using FPGM pruning+weight quantization method,pruning Multi-task Mask R-CNN high redundancy convolution weights,quantization retraining convolutional layers’parameters,improve the model operation speed by 18.6%,reduce the model storage space by 57.2%;using tensorrt structure optimization technology,vertically and horizontally integrate the detectnet model perception module and the longitudinal fusion network backbone"convolution layer+offset+Re LU activation function layer"composite structure,improve the model operation speed by 216%.⑸Development and test of image monitoring platform for operation state of bridge crane.Develop the image monitoring hardware and software platform for monitoring the running state of the bridge crane and controlling the parameters of the camera;carry out the measurement error experiment of the horizontal displacement of the trolley and the tilt angle of the spreader,and verify the effectiveness of the image monitoring technology and the correctness of the theory used in this paper.The experimental results show that the average relative error of the horizontal displacement of the trolley is 2.80%;the relative error of the tilt angle measurement of the spreader is not more than 36.27%,and the absolute error is not more than 4°.