Efficient active monocular fixation using the log-polar sensor

This paper applies a sensor which has been developed based on a model of primate vision to corner fixation. Fixation is critical for the log-polar sensor as many of the sensor’s advantages are demonstrated only under fixation. This paper presents a fast corner fixation algorithm using the log-Hough transform to find the dominant centre-most corner and aligning the optical axis with the corner. The algorithm has been implemented in a real-time closed-loop control system on a pan-tilt platform, This paper demonstrates an important benefit of the log-polar sensor with vote-based algorithms, an emphasis of foveal pixels. This removes the need for searching explicitly through all detected corner positions in order to locate the most central. In addition, since the sensor has a large field of view for a given number of pixels compared to space-invariant cameras, it shows increased computational performance, while still supporting accurate fixation.