Study On The Quality Of Glass On-line Detection System Based On Machine Vision

With the rapid growth of the demand for glass products in the home and abroad market, the production of glass has changed not only the quality and variety but also the production process. Particularly the development of modern production technology, the demand of the quality of glass original plate is higher and higher for high-end product. So it is very important to improve their grades and fully guarantee the quality of glass. At present, the method of testing of the quality of domestic glass is man-made testing, which has heavy workload, low efficiency and low level of automation. For this issue, this paper according to the theory of vision technology and image processing technology, through analysis and research on the online detection system of key technologies of the quality of glass, reach and develop the detection system of the quality of glass, which based on the machine vision. On the analysis of the basic principles of visual technology, combined with the physical environment of glass production,and application characteristics, the overall scheme of the detection system of the quality of glass has been given, and the basic principle of detection system of quality of glass has been analyzed. A hardware and software system has been constructed. The main component of hardware system has been discussed. The algorithm of the software system has been researched. The major technical indicators of the system have been proposedAccording to the application characteristics of glass, acquisition system of glass image was developed; affecting the main factors of the image quality of glass defects was obtained from comparative analysis of the glass defects original drawing. In order to reduce influence of image quality in the external environment, the system adopted image point arithmetic processing as linear transformation, gray histogram, filtering and noise reduction, etc. it was to get improved gray glass defect image and to lay the foundation for the subsequent image processing.The motion-blurred images disappear technology is analyzed, the method of filtering processing based on mathematical morphology to be adopted, and increasing image definition, and then it has the difference image operation between the image and the template image, at last it can obtain the characteristics of glass defects.In combination with the characteristics of glass defects by themselves, an adaptive threshold algorithm is proposed by the analysis of threshold segmentation algorithm. System uses the opening and closing operation in the mathematical form, which removed the noise effect on image when binary processing. It has done comparative analysis on various classic edge detection operators, the glass edge detection methods of canny operator has been given, and to extract edge feature of glass defects. Through interpolation calculated for the characteristics of image, it can get defect image of the sub-pixel, and make characteristic parameters of the image form vector, provide characteristic information for glass defect pattern recognition.According to the characteristics of glass defects, neural network classifier is devised by the analysis of advantages and disadvantages of the traditional BP algorithm. Then Improved BP neural network algorithm for identification is proposed, and it be applied to classification of glass defects and character recognition. The experiments show that compared with the traditional BP algorithm, the improved algorithm has convergence speed and identifies the acceptable failure rate lower. Experimental results show that the system is stable, the anti-interference is strong, the system detection accuracy is 0.3mm, defect recognition rate is 91.75%, achieving 100% Continuous detection.With the improvement of machining automation, more and more requirements are proposed in surface roughness measurement, which mainly includes measurement methods improving, improvement of measurement accuracy and assessment of roughness parameters. Around these requirements, this paper study measurement method in comparison, the measuring principle of laser triangulation is analyzed and the theory of trigonometric functions is used to derive measuring principle; Principle of the measuring instrument overall structure is studied, the types and parameters of component is selected, then measuring instrument test bench is established, Three-dimensional roughness assessment parameters theory is analyzed, On the basis of theory, experimental measurements of surface roughness is conducted, and verify the theoretical analysis with experimental data.By the depth analysis and research of laser triangulation measuring principle, and though the analysis of direct laser triangulation and symmetric oblique optical laser triangulation, CCD pixel displacement is changed by the trigonometric relationship, which reach the height information formula of the measured micro-surface; Comparison between direct laser triangulation and symmetric oblique-type triangulation, direct-style triangulation obtain that simple structure, larger measuring range, but lower resolution. The analysis of triangulation measurement on surface roughness measurements, obtain the direct-type triangulation as surface roughness measurement method.Though the study of measurement methods and component structure principle, it is established that a new type of surface roughness measuring instrument, CCD laser probe uses the combination both Ernostar optical system and Li-CCD, which have good linearity, XY work platform is combined with static flotation rail and linear motors, which achieve high-precision positioning and smooth movement, Column is matched by the stepper motor and ball screw, which meet the exact lift mobile in height.Study of the three-dimensional roughness evaluation parameters, reaches least squares base level established method and the eight of basic roughness evaluation parameters. The processed plane is studied, then the parameters of both two-dimensional and three-dimensional are assessed on experimental data, then the assessment results are compared, which obtain that evaluation of two-dimensional parameters with the limitations and evaluation of three-dimensional parameters with integrated features.