Research Of Recognition Of Laser Making Defect Based On Image Detection Technology

Computer image detection technology is an important means to realize industrial automation,which can effectively improve speed and accuracy of the detection.And image detection can be applied to automatic recognition of surface defects.In this paper,the detected object is the mobile phone keyboard of laser marking.This paper studied the laser keyboard surface marking character recognition method of defects,and developed the hardware and software system for laser marking detection of mobile phone keyboard defects,and focused on the surface of the keyboard image preprocessing,segmentation,tilt correction,registration,feature extraction and recognition algorithm.This paper studied the following contents:In the preprocessing stage,the external and internal parameters of the camera are calibrated by the checkerboard calibration plate.Then,the filtering,the adaptive histogram equalization with limited contrast,and the local adaptive threshold method are adopted.The original image with noise interference is processed,and the target area is significantly enhanced.In this paper,we study the image location and correction algorithm.Firstly,the background and the target region are separated by the diffuse filling algorithm of morphology,and the edge of the edge is detected by the canny operator.Finally,the angle of the image is corrected by using the Hough transform.At the same time we put forward the algorithm of detection and segmentation of the keyboard,the center and radius are detected by the circle Hough transform algorithm.after we get accurate positioning of each target,every single image is segmented by ROI.Because the characteristic of the keyboard image registration is vulnerable to scale changes,rotation changes,images blur,illumination variations,so the image registration algorithms are compared through performance comparison test.A keyboard image registration algorithm(SURF+BRISK)is best.After registration,the feature points are matched with RANSIC and nearest neighbor Hamming distance to eliminate the error matching feature points,and then the affine model is used to obtain the mapping parameters.Experimental results show that compared with SIFT and SURF,the algorithm not only has the same registration accuracy and robustness,but also improves the registration speed by 1.5 times and 1 times,respectively.Finally,the defect area of the image after registration is obtained by template subtraction.In this paper,we study the keyboard image feature and support vector machine.We extract the image feature and the whole keyboard defect area features,and then these the global and local features are entered the classification system of the multi classifier cascade,The classification of keyboard image is realized.The experimental results show that the correct recognition rate of the system can reach more than 90%.In order to verify the laser marking computer character recognition algorithm,we simulated the production site and developed image acquisition hardware system of high performance ARM processor,and analysed the production site with different optical effects on imaging,designing a set of conditions to meet the lighting light source device,and gave the pipeline control method.