Saccades and Fixating Using Artificial Potential Functions

This paper presents a mathematical model for saccadic motion and fixations. We relate this issue to the problem of motion planning and show that a family of artificial potential functions can be used for creating saccadic motion. The advan- tage of this approach is that finding the next fixation point does not require an explicit visual search - which is computationally costly and may be problematic in real-time applications. Rather, the system naturally 'slides' from the current fixation into the next. Thus real-time performance on cheap hardware can easily be achieved. Experimental results serve to provide insight into the performance of a robot APES implementing this approach. I. INTRODUCTION Inspired by human vision, an attentive robot works by allocating its limited computational resources to only the interesting parts of a visual scene. This is done by saccades, rapid eye movements that direct the optical axis from the current fixation to the next such that the fovea which is the high resolution area around the fixation point - overlaps with this interesting area. It has been proposed that this is achieved through visual search on a saliency map - a two-dimensional map that encodes the "interestingness" of the objects in the visual scene. Hence, the problem of how to select new fixation points is treated separately from the problem accomplishing the saccadic motion necessary for moving to this new point. In this paper, we present an alternative approach - in which the two stages are integrated in a unified framework. In this approach, a family of artificial potential functions - each of which encodes the saliency map features around the current fixation point - is defined and given the current fixation point, the next fixation is generated simply by 'sliding' into the equilibrium point of the associated surface. The advantage of the approach is twofold: there is no need for an explicit visual search and the camera actuator commands are automatically generated. In turn, the camera finds a sequence of fixation points without exhaustive search and in real-time. Of course, the 'psychophysical correctness' of the fixation points thus generated is prone to experimentation and is beyond the scope of the paper. Furthermore, since different saliency features - including top down features - can be used to define an artificial potential function, shifts of attention can be made to occur in a 'programmable' manner. In this paper, we show the construction of potential function using one of the most common saliency features - image gradients. However, in general they could easily be constructed based on other features such as color, texture or previously memorized top- down features. The organization of this paper is as follows: In the rest of this Section, we briefly overview related literature. In Section 2, the theoretical framework of fixation generation is presented. In Section 3, we describe the construction of our particular family of artificial potential functions. Experimental results in real scenes are then discussed. The paper concludes with a summary.