Research On The Surface Character Inspection System Of Automobile Front Bar Bracket Based On Machine Vision

As people’s requirements for the quality of life continue to increase,cars have gradually become a necessity for every family to travel in China.The automobile front bumper bracket is one of the important parts of automobile products,and the characters on its surface have important meanings.During the manufacturing process of parts,the characters on the surface will generate various production defects,such as character adhesion,character incompleteness,and character area pollution.If products with character defects enter the consumer market,relevant manufacturers will inevitably fall into disputes between manufacturers and consumers,and will be subject to greater negative impacts and financial fines.Therefore,the character recognition and defect detection on the surface of the front bumper bracket of the automobile is an important part of the production process.However,most manufacturing enterprises still adopt manual detection methods,which has the problems of high labor intensity,low production efficiency,and easy omissions.This paper adopts the visual inspection method,and proposes a vision-based character detection system for the surface of the automobile front bumper bracket.The research content includes the following aspects:(1)According to the mechanical structure characteristics of the parts to be inspected and the technical requirements that need to be met during the inspection,the overall mechanical structure design of the inspection machine and the electrical control design of the motion state of the object to be inspected during the inspection process are completed.The content of mechanical design mainly includes the selection of power source,the design of motion transmission mechanism and motion actuator.The design of the motion control part mainly includes stepper motor,single chip microcomputer and control program.(2)For the characteristics of overlapping characters on the part surface,the selection of the image acquisition hardware part of the vision system is completed and explained in detail.The study compares the effect of image acquisition under different light source illumination,and finally selects the quadrant light source and stroboscopic light source controller as the light source combination method,which provides a new lighting method for separating the chromaticity-free concave and convex characters in overlapping characters.(3)In the image pre-processing of automobile front bar bracket,a polynomial gray value fitting method is proposed to optimize the phenomenon of uneven character brightness and reduce the complexity of subsequent image processing for the uneven character brightness of the image after camera acquisition.In order to avoid the overlapping of black spray code characters and colorless characters affecting the character recognition accuracy,the separation of two overlapping characters is achieved by further processing the gradient data based on the photometric stereo method.(4)Character localization and segmentation are investigated separately based on character characteristics.For the precise localization of chromaticity-free characters,a morphological improvement algorithm based on the scanning algorithm is proposed.The algorithm first coarsely locates all character regions with the scanning algorithm,and finally determines whether the target character is precisely located by matching the character region aspect ratio threshold range.For the problem of character adhesion after character localization,a projection dichotomy improvement algorithm based on the connected-domain method is proposed to enable accurate character segmentation,and the processing of the algorithm is also described in detail.The Blob analysis method is used to complete the character defect detection and reduce the false recognition rate.For the study of character recognition,by comparing the recognition methods of support vector machine,Le Net-5 network model and improved Le Net-5 network model,the improved Le Net-5 network model is selected for character recognition after comprehensive comparison.At the same time,the design of the software interface of the detection system is completed,which is beneficial to the users.This paper designs and develops a character detection system for the surface of automobile front bar bracket,and collects a large number of physical images for testing.The experimental results show that the character recognition accuracy of the detection system reaches 97.3% and the detection time is short,which can meet the requirements of actual industrial production inspection and be applied to actual production inspection.