Sensor geometry for dynamic vergence: characterization and performance analysis

Real-time control architectures for active vision impose strong constraints on the computational complexity of image processing algorithms. In many cases, fast and reactive visual routines are achieved by drastically reducing the image size, at the cost of degrading the accuracy of the algorithms. The non-uniform sampling of the images, such as the log-polar mapping, is an eecient way of reducing the amount of data to process 7, 17], while maintaining a good resolution in the center of the visual eld. Additionally, log-polar images have several advantages over cartesian images, namely for the control of vergence in stereo active vision devices. In the work described here, two diierent vergence control strategies are presented. Both are based on the correlation of low-resolution log-polar images and the control law is a function of direct image measurements, leading to fast visuo-motor processes. In this paper we focus on the problem of characterizing the performance of the vergence system relative to various log-polar sensor layouts and control strategies. The quality of a given sensor layout is assessed by analyzing the shape of the correlation function for various values of the vergence angle. Therefore, we introduce a set of quality criteria, depending on the sensor layout, that ultimately determine the performance of the control algorithm. The innuence of the log-polar sensor parameters on the shape of the correlation function is analyzed, and criteria to choose adequate layouts are established. Experiments on real image data, in closed loop control, are shown.