Light Optimal Design For Machine Vision Based On Reflection

In recent years, the application of machine vision is more and more extensive withthe rapid development of artificial intelligence, computer technology, industryautomation and machine learning, etc. The versatility and flexibility of machine visionapplications have been greatly improved. Particularly, it is playing extremely importantrole in the industrial automation field. And the most important element to capture highquality images in machine vision systems is lighting, because the quality of the visualinformation which is the basis of further processing is directly affected by it.Traditionally, the aforementioned lightings are selected from the premanufacturedsamples based on experiences. The traditional methods are trail and error strategieswhich are not only laborious but also time-consuming. Generally, the lighting selectedby the trail and error methods cannot optimally match the surface property and can’t getthe desired high quality image. However, the personalized optimization design forapplications can avoid the blindness of selecting lightings and will not be limited to theexisted ones, more importantly better lightings than using traditional method can beachieved. At present, The LED illumination is extensively used in machine visionsystems with the low power consumption, high luminous efficiency and long life and soon, most importantly, the light of LED is composed with well arranged individual unit,that make LED be the best choice for us to optimize.In this thesis, a LED lighting optimization model is proposed on the basis of thereflection characteristics of object considering the factors and difficulties when lightingsare selected. Obtaining the biggest contrast between the target and back-ground istreated as the objective of the lighting optimization model. In the optimization model,the biggest contrast is obtained mainly through optimizing the wavelength of light,angle of incidence and the luminance of light. At the same time, the model ensure theuniformity of illumination by using the uniform illumination technology for non-normal incidence proposed in this thesis, improves the value of LED lightings forpractical application.The basic knowledge of optics and the characteristics about light reflection andabsorption of different objects, the definition of the contrast and its application inlighting design are studied in this thesis. The bidirectional reflectance distributionfunction (BRDF) based reflection models are generally used in the optimization oflightings. Then Cook’s model is improved to meet the need of light improving theaccuracy and efficiency. The algorithm and the simulation of its result are achieved by C++and Matlab. The illumination effect of optimized lightings is greatly improvedrelative to the non-optimized ones by comparing with the result.