Optimum illumination for machine vision using optical scatter data

The objective of this research is to establish a method of obtaining optimum illumination for machine vision based on knowledge about how light is scattered upon impingement on different surfaces. Proper illumination is critical to image quality and hence to a vision system's ability to extract information from a scene. Surprisingly, the state of the art of obtaining proper illumination in machine vision applications remains largely a time-consuming ad hoc trial-and-error process which usually results in sub-optimum lighting. It is shown that an understanding of the way in which light is reflected from surfaces in a 2-D scene can be used to predict and/or control how the image of the scene will be formed on an image sensor.; This study begins with an empirical look at how light is scattered by various surfaces. An instrument is built that measures scatter over an entire hemisphere above the scattering surface when the surface is subjected to a collimated light arbitrarily positioned in space. Details, limitations and design considerations for this scatterometer are provided.; The plethora of data accumulated by the scatterometer necessitates the use of a reflectance model that can characterize surface reflective properties. A model based on a modification of the Torrance-Sparrow model is developed for this purpose. This model characterizes surface reflectivity, and hence perceived brightness, as the sum of a diffuse, specular and retro-reflective term. The model eliminates many of the complexities of the Torrance-Sparrow model at the cost of sacrificing some of the latter's accuracy. This trade-off between simplicity and accuracy is acceptable in many vision applications where scene brightness is generally discretized into 256 shades of gray. The model is parameterized by six constants obtained from data collected using the scattermometer. Validation of the model is presented.; Using this reflectance model, a definition of contrast is developed based on the model-predicted brightness differences between surfaces in a scene. This mathematical expression of contrast is maximized with respect to a set of coordinates used to describe the orientation of a collimated light source and imaging system with respect to the scene. From this, the orientation of camera and light that results in the image of greatest contrast is numerically determined. The reflectance model is also used to predict how images will appear at a given orientation of camera and light source.