Research On Safety Belt Detection System Based On Machine Vision

Automobile safety belt is an important device for protecting drivers and passengers.In recent years,with the rapid growth of China's automobile market,the market size of the automobile safety belt industry is also growing rapidly.In 2017,the market reached 3.8 billion yuan,up 8.57 percent year-on-year.In the process of production of safety belt,due to the unstable process,there will be some size deviation,color difference,smudge,burr,silk and other defects.These defects not only affect the vision beauty and feel of consumers,but also affect the luxury and comfort of the car.Various defects of automobile safety belts result in huge market demand for defect detection of automobile safety belts.Manual detection is greatly affected by subjective factors,and has disadvantages such as slow speed,high cost and low accuracy.Studies show that manual detection can only detect about 85%of defects,and the detection speed is no more than 15m/min.There are many researches on defect detection of textiles at home and abroad,mainly focusing on cloth,but there are few domestic researches on vision detection of safety belts.In order to reduce enterprise cost and improve productivity,it is urgent to research and develop automatic safety belt detection system based on machine visionTherefore,in view of the current adverse situation,the main research content of this paper is as follows:First,vision detection problem against defects in safety belt,according to the production process of safety belts,defects,and test requirements,system overall design scheme is put forward,from the mechanical structure,electrical control,vision inspection and PC software for overall scheme design,the scheme involved in the selection of hardware and design,build a testing platform.Secondly,in the process of safety belt detection,the safety belt must be under constant tension state of vision detection,real-time tension control algorithm will be studied.Thirdly,vision detection of safety belt defects includes two parts:(1)the size measurement algorithm of safety belt and the design and verification of edge defect detection algorithm;(2)design and verification of surface defect algorithm.As for the vision detection algorithm in part(1):due to the distortion of camera lens,the relationship between pixel coordinates and world coordinates needs to be clarified in the dimension measurement,so the real-time image needs to be calibrated and the camera imaging model,distortion model and calibration method need to be studied.After calibration,safety belt size measurement and edge defect detection will be carried out respectively.In order to make the algorithm have high precision and high robustness,various image algorithms are used to explore the advantages and disadvantages of various algorithms through experiments.In view of the(2)part of the vision detection algorithm:the directivity of the safety belt and the periodic appearance of the texture have an adverse impact on the detection,the spectrum characteristics are studied,that is,the spectrum image after Fourier transform is analyzed,and the texture interference is filtered to extract the defects.For some non-obvious defects that are difficult to be extracted,the deep learning model was studied for analysis and extraction,and finally the experimental verification was carried out.Finally,system integration and overall system experiment analysis are carried out.Based on C#programming language and Halcon image processing library,the system software is integrated and developed through modular programming to build a complete detection system.Experiments were carried out to verify the overall time consumption and accuracy of the system.The results show that the system can meet the requirement of safety belt detection and achieve the expected effect in the practical trial.It provides beneficial exploration and improvement for vision inspection of safety belt defect.