Detection Technology And System Of Hardware’s Surface Defects

In the production process of hardware,the hardware’s surface defects can be produced due to the poor processing equipment or worker’s mistake.In the automatic line of industrial products,when the industrial products are made by this hardware which has surface defects,the quality of industrial products will be seriously affected.The traditional detection method of hardware’s surface defects is mainly detected by the experienced workers for each hardware.However,the subjective factors and the tired working state of workers will reduce the efficiency of detection.Therefore,the realization of efficient automatic detection of hardware’s surface defects has great significance for the production of hardware.Machine vision detection technology is an efficient non-contact detection technology,and machine vision technology can be used to detect hardware’s surface defects.This thesis studies the problems existing in the hardware’s surface defects detection technology that is based on machine vision and the key technologies to solve the problems.The main innovations of this thesis are as follows:Firstly,the hardware’s surface defect detection technology based on machine vision mainly relies on camera to take pictures of hardware.The system transfers the image data to the computer for image processing,and detects the hardware ’s surface defects.In the process of camera shooting of hardware,camera calibration is needed because of the general distortion of the camera lens.However,it is needs to shoot multiple calibration images to calibrate the camera by using the traditional camera calibration method.It is not convenient for the workers to operate and reduces the convenience of the machine vision detection system by using traditional camera calibration method.Therefore,the mathematical model and principle of Zhang Zhengyou plane c alibration method is analyzed and the camera calibration method based on K-SVD dictionary learning is proposed in the this thesis.The camera calibration method based on K-SVD dictionary learning uses a sparse dictionary to solve the initial value of t he camera’s internal parameters quickly,and camera external parameters can be solved by internal parameters.The distortion coefficient is obtained by optimization,and the camera’s internal and external parameters are used in this process.Finally,the optimal internal parameters and external parameters and distortion coefficients of the camera are obtained by maximum likelihood estimation.This method is more efficient than the traditional calibration method,and the calibration accuracy is high.The camera calibration can be completed by only one calibration image,and it reduces the operation of camera calibration.Secondly,image alignment is one of the key technologies for hardware ’s surface defects detection.The angle deviation,The angle deviation which is between the detected hardware and the standard hardware,and it is not conducive to detect the hardware’s surface defects,and the correct rate of defect detection is greatly reduced.In order to solve the alignment problem between the standard hardw are and the detected hardware,the image alignment method based on contour features is proposed in this thesis.To calculate the Hu moments of the standard hardware’s contours and the detected hardware’s contours,and let the 7 invariant moments of the Hu moment as feature vector for each contour.The contours is matched by the feature vectors,and two pairs of matching contours will be obtained.To find the two pairs of center points in the two pairs of contours,and the two angles of the two pairs of contours can be calculated by the two pairs of center points.Those two angles are subtracted so as to find the angle between the standard hardware and the detected hardware.Through the image rotation and translation,the detected hardware’s image and the standard hardware’s image can be aligned quickly.The requirement of image alignment based on this method can be met the process of hardware’s surface defects detection.Finally,according to the above theoretical research achievements,a system of hardware surface’s defect detection has been designed and has been applied to hardware manufacturing industry.It has already successfully detected the common defects in the hardware surface in practice.This not-contact,timely and efficient detection system has great practical values and economic benefits in engineering applications.