Real-time adaptive regulation of a visual camera for automatic investigation of changing environments

An adaptive strategy for regulating the intrinsic parameters of a CCD visual camera is proposed. The parameter regulation procedure is based on the computation of the "quality" of an acquired signal by means of a set of functions which estimate the goodness of the datum. Each function performs a quantitative evaluation by computing several features derived from the high-frequency content of an image and from the histogram analysis. The regulation strategy uses such values to compute the actual values of the camera parameters, i.e., focusing distance, aperture diameter, electronic gain, and black level. Moreover, the same strategy can be applied both to the whole image and to smaller subareas, so providing a focus-of-attention mechanism. This is useful for the automatic investigation and control of environments subject to dynamically changing conditions (e.g., the illumination and the structure of a scene), whenever it is not possible to interact directly with such environments. The system has been implemented on a Sun Spark Station II linked with a CCD camera (with the capability of computer-controlled parameters) and with an image-acquisition system, which permits real time performances.