Research On Visual Inspection System For Vehicle Lamp Bulb Appearance Defect Inspection

As the industrial field continues to carry out automated production technology innovation,the lights are no longer artificially fired,but have been producted in assembly line.As a critical part of vehicles,Vehicle Lamp Bulb(VLB)and its manufacture quality are closely related to traffic safety.And since the firing process is complicated,the quality inspection of Vehicle Lamp Bulb Appearance Defects(VLB-ADs)has been a major concern in the industry.At present,the quality inspection of VLBs still depends on manual word in China.Manual inspection has a lot of problems,such as strong subjectivity,inconsistent standards,low efficiency,and high labor costs.Therefore,manual inspection has already not meet the needs of high-speed and large-scale automated production.Therefore,the market has demanded the development of a visual inspection system for VLB-ADs inspection.In the visual inspection system for defect detection,a reasonable illumination imaging scheme and an effective defect inspection method are important factors that affect the inspection result.At present,the illumination imaging scheme scheme and inspection method for other objects is not suitable for VLB-ADs.In this paper,the widely used illumination imaging schemes are designed for the bayonet VLBs according to their material and structure,and the causes and locations of typical appearance defects.And a visual inspection systems for the bayonet VLB-ADs inspection.The main contents of this paper are as follows:1.According to the material and structural characteristics of the detection region,the illumination imaging schemes are designed,and a matching device structure is designed for these schemes.The structure can be combined with the processing and control module and the pipeline mechanical device to realize automatic inspection and classification of the product.2.According to the formation cause and the appearance position of the typical VLB-ADs,the detection region of a VLB includes the top and side area of the lamp cap.For the top area localization,the Canny edge detector is adopted to detect the edges of the lamp cap object,and then a least square method based on a priori constraint is proposed to fit the outer boundary circle of the top area,which reduces the interference of the unrelated edges on the fitting process and reduces iteration times required by the traditional least squares method.The method more satisfies the requirements for accuracy and real-time in actual production.3.This paper analyzes typical VLB-ADs,including: insufficient tin saturation,smudge and deformation on top area,lamp ear deficiency,smudge and deformation on lamp ear,lamp ear position error,smudge and deformation on side area,crooked lamp cap,product label absence.And the eight typical VLB-ADs can be effectively detected.4.A superpixel fine fusion method based on region constraint is proposed for the inspection of the Insufficient Tin Saturation(ITS)and Smudge and Deformation on Top Area(TASD)defects.SLIC is adopted to generate super-pixels.This paper firstly model the similarity distance between super-pixels as Weibull distributions to preliminarily cluster superpixels to obtain clustering patches.Then,the second clustering process is constrained according to the gray feature of the detection zone where the clustering patche is located,so as to find out defect blobs and achieve the fine fusion of the non-defective patches and their detection zones.5.This paper converts the image information of the target area into a one-dimensional signals according to the imaging feature of the side area.The inspection of Smudge and Deformation on Side Area(SASD)defects is achieved through adopting multi-scale gaussian shape detector to detect ridge/valley of the one-dimensional signals.