Design And Implementation Of Calcium Carbide Furnace Inspection Robot System Based On Machine Vision

As an important chemical material,calcium carbide is widely used in many fields such as medicine,chemical industry and metallurgy.As a high energy consumption,rough traditional industry,the production process of calcium carbide industry is harsh environment,low level of equipment intelligence,high degree of safety risk of manual operation,so there is an urgent need for intelligent equipment to replace manual safety production.In the calcium carbide furnace inspection task,if only rely on the traditional manual on-site inspection inevitably exist mis-inspection leakage,inspection workers personal safety is difficult to guarantee and other problems,so this paper proposed a calcium carbide furnace intelligent inspection robot system based on machine vision and deep learning algorithm.The system firstly addresses the shortcomings of the traditional image recognition algorithm used for fixed-point monitoring,such as low recognition accuracy and failure to ensure real-time detection,and uses deep learning methods to achieve automatic real-time recognition of readings of pointer-type meters at the production site of calcium carbide furnace.This paper also investigates the image visibility enhancement of the calcium carbide furnace environment,and finally designs and implements an inspection robot uplink system.The system is used to make inspection plan,inspect the furnace environment instead of manual inspection,transmit the camera screen,the concentration of harmful gases in the environment and other information to the inspector through the host computer interface and save the inspection history.The main research contents of this paper are:(1)Overall design and hardware selection of inspection robot system: It is determined that the calcium carbide furnace inspection robot is mainly composed of robot body,wireless charging module,and upper monitoring module.The robot body is equipped with visible and thermal infrared binocular gimbal camera,gas sensor,wireless IP phone,and navigation equipment,and the robot is equipped with autonomous navigation,wireless charging,emergency obstacle avoidance,automatic alarm and other functions.Complete the hardware selection of the calcium carbide furnace inspection robot system,including industrial cameras,motor controllers and other important components,and finally complete the construction of the hardware system.(2)Pointer meter reading recognition: In order to make the calcium carbide furnace inspection robot system complete the task of pointer meter reading faster and more accurately,this paper proposes a pointer meter automatic recognition reading method based on the improved lightweight YOLOv5 for calcium carbide furnace production site.Firstly,it locates the dial area of the meter in the image captured by the inspection robot without any a priori information;secondly,it extracts the position of the pointer in the dial area,the center of the dial and the dial characters,and calculates the Euclidean distance to get the two characters closest to the pointer;then,it uses the improved lightweight YOLOv5 to recognize the characters,and uses the model with less number of parameters in YOLOv5 The Shufflenetv2 structure is used as the backbone network,and the channel attention mechanism of fused position is added after each backbone network to improve the expression ability of the network learning features;finally,the angle between the pointer and the line from the center of the dial to the adjacent characters is calculated to obtain the indication value of the pointer.Compared with other pointer meter reading methods,the proposed automatic pointer meter reading method has obvious advantages in terms of speed and accuracy.(3)Image enhancement algorithm: To address the phenomenon of poor visualization of captured images that may occur during calcium carbide production(e.g.,poor lighting conditions,nighttime,or thick smoke during production),this paper validates several image enhancement methods and improves the dark channel a priori algorithm by replacing the transmittance optimization step with a fast guide filter that processes images faster and without degrading the image processing effect.It is demonstrated that the improved dark channel a priori algorithm combined with the restricted contrast histogram equalization(CLAHE)algorithm is able to achieve better de-fogging and image enhancement of calcium carbide furnace images.(4)Research and Development of Robot Uplink Monitoring System: In this paper,the relevant architecture and process design,database design,human-machine interface design and specific development of the robot inspection uplink monitoring system are carried out to realize the functions of video monitoring,inspection,one-touch charging,environment detection,meter reading,etc.The system uses TCP/IP protocol to communicate with the lower computer to obtain real-time robot status information for remote inspection personnel to view.The research and development of real-time storage database system for inspection data as well as capture video data,which can print inspection reports with one click to facilitate personnel to view the abnormalities in calcium carbide production.