Research And Implement Of Camera Calibration In Autobody Parts Vision Measurement

Auto Body In White is the major assembly of a vehicle and its quality has great effect on the whole function of the vehicle.Because of the complexity of the industrial field,tolerances are inevitable in the manufacturing process.To improve the product quality,quality control must be implemented in the process,while the key of quality control is measuring.As the level of image processing,computer technique and industrial camera manufacturing is improving,computer vision is developing fast,making the non-contact online inspection possible.Computer vision gets images of the scene through vision sensors,apperceives and renews the object's geometric properties,gestural configuration,movement,position,etc.It recognizes,describes,explains the scene,and gives judgments.For the excellence of rapidity,automation,precision and non-contact characters。computer vision is becoming the important solution and developing direction of manufacturing quality measuring.In the domain of computer vision,camera calibration is an absolutely necessary step,an important part and the precondition and foundation of obtaining 3D geospatial data.Once the camera is calibrated,the motion parameters between images will be solved easily and robustly.The precision of calibration directly affects the precision of 3D measuring and 3D reconstruction.And the real-time performance of calibration can satisfy the need of industrial computer vision such as navigation.So,research of camera calibration has significant theoretical research sense and practical application value.Under the background,some research about the theory,method and technique of camera calibration was done in this paper,and the content is as followed.The first chapter is an introduction,Based on quality control of body manufacturing and computer vision technique,high precise camera calibration has significant theoretical research sense and practical application value.The second chapter is about calibration.The basic knowledge of calibration is introduced,including the kinds of solutions,the derivation process and mathematical expressions of linear and non-linear calibration.The development and present situation of calibration is also summarized.After researching the traditional methods,their advantages and disadvantages are compared,and their feasibilities are analyzed.In chapter 3 which is about neural network based on fuzzy theory,the fuzzy neural network calibration method is introduced.First,introduces the neural network and BP algorithm.Then the fuzzy theory is introduced to region division making the transition between regions smoother.And then the reasoning process is completed based on the correlation between the right and left images' fuzziness.At last, a fuzzy neural network is built and a precise calibration result is gotten.Chapter 4 is an experiment.Use the fuzzy neural network calibration method in the dimension inspection of the front apron board's holes.The calibration result is well expressed with data and images . The result is good and the fitting precision is satisfactory.The possibility and effectiveness of this method is proved by theory research and experimental results.The adaptability and practicality of the calibration formality is proved also.Chapter 5 is part of summarize and expectation.The research result and innovation work of this paper are concluded,the deficiency is also summarized.An expectation of how to develop body dimension inspection rapidly and precisely is given in the end.Therefore,it is practically significant to research using fuzzy neural network for calibration in computer vision inspection of parts.And this paper does some useful studies and tries in camera calibration.