Research And Development Of Surface Quality Inspection System For Automotive Hoses Based On Machine Vision

Automotive hoses are widely used in automotive cooling systems,lubrication systems,and fuel supply systems.They serve to transport oil,gas,water and power,and are important parts of automobiles.During the production process of the hose,some defects or quality problems such as offset or misprinting of the assembly mark may occur on the surface of the outer rubber layer.At present,the inspection of the surface quality of the hose is mainly completed by manual visual inspection,which can no longer meet the needs of modern industrial production.Machine vision inspection is the use of cameras instead of human eyes to complete the inspection of products,which has advantages that cannot be matched by humans.To this end,this subject developed an online visual inspection system to realize the surface quality inspection of the hose.In this paper,the overall design of the hose surface quality inspection system is carried out.Through the analysis of the hose production process,three types of surface defects are defined: cracks,white spots,and protrusions.Combined with the lighting mode of the light source,the gray level distinction of the upper and lower edges of the raised defect is improved,and the dimensionality reduction is realized for the three-dimensional defect.After the hose image is collected,it is preprocessed and edge detected.This paper proposes the APS-Canny adaptive parameter selection edge detection algorithm to completely extract the edge of the defect,and perform morphological processing on the detected image and remove the boundary connected domain and so on.According to the shape characteristics of the defect,parameters such as the aspect ratio were selected to extract the characteristics of the defect.Through a large number of experiments,the characteristic value range of the three types of defects was determined,and the classification of surface defects was completed.In the surface mark recognition and positioning stage,because the system requires high real-time detection,the current fastest ORB algorithm is selected for feature matching,but the stability of the ORB is relatively poor.This paper proposes an improved ORB algorithm based on the invariance of the rotation scale.Fusion of FAST corner detection and SURF descriptor algorithm,combined with RANSAC and FLANN matching algorithm to complete the mark matching and positioning.By detecting the presence or absence of the mark and the deviation angle,and then comparing with the threshold value,the judgment of the mark defect is completed.Compared with the traditional ORB algorithm,the improved ORB algorithm matching accuracy has increased by more than 20%.This article has developed a complete set of machine vision inspection system for automotive hoses,which has certain practical value,and its efficiency and accuracy are far superior to manual inspection.The system's algorithm recognition accuracy rate is over 95%,and the detection time is less than 1s,which proves the feasibility of applying machine vision to the surface quality detection of hoses.