Tracking natural trails with swarm‐based visual saliency

Abstract

This paper proposes a model for trail detection and tracking that builds upon the observation that trails are salient structures in the robot's visual field. Due to the complexity of natural environments, the straightforward application of bottom‐up visual saliency models is not sufficiently robust to predict the location of trails. As for other detection tasks, robustness can be increased by modulating the saliency computation based on a priori knowledge about which pixel‐wise visual features are most representative of the object being sought. This paper proposes the use of the object's overall layout as the primary cue instead, as it is more stable and predictable in natural trails. Bearing in mind computational parsimony and detection robustness, this knowledge is specified in terms of perception‐action rules, which control the behavior of simple agents performing as a swarm to compute the saliency map of the input image. For the purpose of tracking, multiframe evidence about the trail location is obtained with a motion‐compensated dynamic neural field. In addition, to reduce ambiguity between the trail and trail‐like distractors, a simple appearance model is learned online and used to influence the agents' activity. Experimental results on a large data set reveal the ability of the model to produce a success rate on the order of 97% at 20 Hz. The model is shown to be robust in situations where previous models would fail, such as when the trail does not emerge from the lower part of the image or when it is considerably interrupted. © 2012 Wiley Periodicals, Inc.

A Machine Learning Approach to Visual Perception of Forest Trails for Mobile Robots

“…Simple Saliency-based Model. We compute saliency maps of the input frame using Itti's model [33], as in Santana et al [12]. This map is computed on the image hue only, which preliminary experiments shown to be the configuration where saliency is most correlated to trail location.…”

“…Several previous works [11], [12] dealing with trail perception solved a segmentation problem, i.e., aimed at determining which areas of the input image correspond to the image of the trail. In order to solve this task, one needs to explicitly define which visual features characterize a trail.…”

“…[11] relied on appearance contrast, whereas Santana et al adopted image conspicuity [12]; both features are conceptually similar of image saliency [13]. For every pixel of an image, saliency quantifies how much such pixel visually "stands out" from the rest; for example, the pixels belonging to a small colored object on an uniform background will be characterized by an high saliency with respect to the saliency computed for the background pixels.…”

“…If we assume that the trail image exhibits some sort of marked visual difference with respect to its surroundings, then saliency will be high for trail pixels, and low elsewhere. This information, which by itself is expected to be very noisy, was aggregated [14] with a number of simple (symmetric, triangular shape [11]) or complex (spatialtemporal integration based on virtual ants [12]) geometry priors in order to infer the trail position and direction in the image, thus producing a rough segmentation of the trail. A similar segmentation system using data from stereo omnidirectional cameras and a laser rangefinder was also implemented [15] to steer a wheeled robot along a trail.…”

A Review of Deep Learning Methods and Applications for Unmanned Aerial Vehicles

“…In [40], a DNN model was trained to map image to action probabilities (turn left, go straight, or turn right) with a final softmax layer and tested on board by means of an ODROID-U3 processor. The performance of two automated methods, SVM and the method proposed in [76], is latterly compared to that of two human observers.…”