Coordination Of Illumination And Camera In Computer Vision System

Traditional methods for machine vision to better interpret scene are usually focused on post-image processing (e.g. smoothing, filtering, contrast stretching, etc.). Post-image processing does NOT increase the inherent information content of data, but originally better image contains more information of object surface that will benefit further vision analysis and a lot of enhancement processing is evaded, which is very important in a machine vision system, especially for real-time purpose.The performance of the vision perception of a robot, and thus the quality of knowledge learnt, can be significantly affected by the properties of illumination and sensor such as intensity, color and relative placement. This paper presents a study on obtaining the optimal illumination condition for the vision perception in computer vision. The "comfortable" condition for a robot eye is defined that the image has a high signal-to-noise ratio and high contrast, is within the linear dynamic range of the vision sensor, and reflects the natural properties of the concerned object. In this paper we investigate the problem of relative placement between an object, a camera and a light source and propose appropriate methods to optimize the optical parameters of the luminaries and the sensor. Finally we apply this placement algorithm to optimize the vision perception of 3D surface measurement based on computer vision. It turns out that ensuring an optimal placement of the camera or of a light source is an essential step in the development of industrial vision systems.Indeed good lighting conditions ensure good image quality and thus enable to simplify or improve reliability of vision algorithms. The main contribution of this paper is to planning the camera and the light together as a whole based on traditional sensor panning. The purpose is to obtain an optimal light and camera combination of their positions. Then we extract surface feature such as smoothness using highlight inspection and specular exponent measurement in the coordinated stable visual system. Both numerical simulations and practical experiments are carried out. Results show very satisfactory consistency with human visual inspection. The implementation also demonstrates that the light-camera coordinated visual system is the indispensable step to improve qualities of industrial visual systems, with a wide range of Application.