Research On Image Online Detecting Algorithm For Laser Writing Parts

With the large number of applications of laser writing technology in actual production,the demand for laser writing and laser marking on line quality detection increase with each passing day,but the study of laser writing detection method for the surface of parts under complex background is relatively few.The machine vision technology meets the requirements of quality inspection in the process of laser writing parts because of its good adaptability and accuracy.The main content of laser writing graph is width of contour,connectivity and burr information of the filtered image.The real-time performance,accuracy and robustness are the key performance indicators of the detection algorithm.Due to the uniformity of cutting material and the stability of laser power,there are more burrs on the edges of the lines.The metal material may have scratches and the background of wood plate is influenced greatly by wood grains and knurls.The image is characterized by complex background noise.At the same time,Because of large size of laser writing products,it is easy to cause uneven illumination,which makes conventional image segmentation difficult to achieve the desired results.In this regard,traditional machine vision detection methods are not suitable for laser writing.In view of the actual situation of the feature detection of the laser writing parts and the requirement of the machine vision detection,this paper presents a quality detection method based on the guide line,which uses the vector track file of the laser cutting tool.The main research contents are as follows:Firstly,The hardware and software design and image preprocessing of the system were completed.In addition to routine hardware selection,a targeted light source solution is designed for image acquisition of large size of laser writing products.In order to facilitate operation and debugging,the visual software with friendly interface and convenient operation was coded.In addition,the distortion of the collected samples is corrected to ensure the accuracy of contour detection.In view of the background of the image,the complex noise and the indistinct target contour,bilateral and nonlinear stretching are used to enhance the image.Secondly,the vector file of laser cutting path is analyzed and reconstructed to get the sampling points needed for detection.The vector coordinate data is converted into a line segment,and the redundant data are eliminated;Then corner points of the line segment are marked and are avoided in the generation step of the sampling point;at the same time,distinguish between line segment and face area of line set where sampling points is;and the package is collected.Then,vector data including sampling points and corner points are projected onto the image by affine transformation.Finally,the gradient based growth detection template and the SUSAN operator are used to detect the various parameters of the contour on the local images of the sampling points and the corner points.The sample test and comparison experiment with other methods have been carried out.The results show that the method shows a good robustness under the complex background condition,while ensuring high detection accuracy and good real-time performance,which meets the quality inspection requirement in the laser writing field.