Autofocus Technology Research, Based On Digital Image Processing

Vision measurement is a hot technology in the field of precision measurement. It's been widely applied in many occasions, and can be used to solve some problems that common measurement methods couldn't do. Auto-focus is one of the key technologies in the whole vision measurement system. With the development of the focusing technology, it is also becoming a non-contact displacement measuring method, and as a combination with the traditional 2-D CCD measuring, it will have wide application potential in the future as a 3-D non-contact measurement technique. Focusing based on the digital image processing technology is an important development trend, and research on this technology is of great significance to the development of vision measurement.Auto-focus based on the digital image processing technology for precision measurement, it is necessary to study focusing resolution, efficiency and reproducibility etc. that have significant influence on measuring. Specifically it need to solve the selection and design of the high-resolution focusing evaluation algorithm and high efficiency & high-precision search algorithm (strategy), as well as the study on the factors that affect the accuracy of focusing repeatability. In view of these points, the paper has done a more in-depth theoretical analysis and experimental research.In a more comprehensive analysis of the focusing evaluation functions based on the digital image processing, the experiment using the functions based on the edge detection algorithms was carried out. A confirmatory test results shows that the performance of focusing functions varied from one to another when the evaluation faced on the different objectives; The feasibility of wavelet transform in application to image definition criterion was verified and analyzed, the research shows that the function's capability is significantly related to the spatial texture distribution of the image, and image gray gradient direction matrix texture analysis methods can greatly improve its performance. Due to the prevalence of machined parts' surface texture, the image texture analysis has practical applications. An image definition criterion using wavelet transform based on the texture analysis was proposed in this paper through the statistical analysis to the image texture characteristic. Experiments shows that the evaluation function aimed at the target texture features has superior performance, and with a higher resolution (better than 2μm).On the basis of analyzing and comparing with several auto-focus searching algorithms, an ameliorative method based on mountain climb servo (MCS) was proposed. Currently the search precision is in conflict with the search efficiency. The paper used threshold control, limited range curve fit, rough & elaborate evaluation functions and other schemes to improve the accuracy of MCS. The simulation experiment about modified MCS shows that the searching accuracy is not below the minimum searching step (2μm in experiment). Compared with the original method, it usually has a very good improvement.To apply in the practical engineering, the factors that affect focusing accuracy and repeatability were studied in detail. Considered the noise in the process of focusing, the paper used filtering algorithms for image preprocessing. Experiments show median filtering is a better way; In order to overcome the troubles caused because of the focusing planar surface is not parallel with the target surface, and the interference between the main image of the target and the background image, the paper used the layouts of focusing window size, location, and focusing regional selection etc. Experiments show that preferable focusing window layout can solve these problems in part, and in addition the focusing functions' performance can be improved effectively, also can reduce the computational complexity; Experimental studies on the imaging target's material, surface roughness, brightness of imaging surface and texture direction of the imaging target were carried out, the conclusion is imaging target characteristics need to be considered to improve the focusing accuracy and consistency. The experiment provided some valuable references for the selection and design of focusing algorithms.