A novel pure-vision egomotion estimation algorithm is presented, with extensions to Unmanned Aerial Systems (UAS) navigation through visual odometry. Our proposed method computes egomotion in two stages using panoramic images segmented into sky and ground regions. Rotations (in 3DOF) are estimated by using a customised algorithm to measure the motion of the sky image, which is affected only by the rotation of the aircraft, and not by its translation. The rotation estimate is then used to derotate the optic flow field generated by the ground, from which the translation of the aircraft (in 3DOF) is estimated by another customised, iterative algorithm. Segmentation of the rotation and translation estimations allows for a partial relaxation of the planar ground assumption, inherently increasing the robustness of the approach. The translation vectors are scaled using stereo-based height to compute the current UAS position through path integration for closed-loop navigation. Outdoor field tests of our approach in a small quadrotor UAS suggest that the technique is comparable to the performance of existing state-of-the-art vision-based navigation algorithms, whilst also removing all dependence on additional sensors, such as an IMU or global positioning system (GPS).

In macaque, it has long been known since the late nineties that the medial parieto-occipital sulcus (POS) contains two regions, V6 and V6A, important for visual motion and action. While V6 is a retinotopically organized extrastriate area, V6A is a broadly retinotopically organized visuomotor area constituted by a ventral and dorsal subdivision (V6Av and V6Ad), both containing arm movement-related cells active during spatially directed reaching movements. In humans, these areas have been mapped only in recent years thanks to neuroimaging methods. In a series of brain mapping studies, by using a combination of functional magnetic resonance imaging methods such as wide-field retinotopy and task-evoked activity, we mapped human areas V6 (Pitzalis et al., 2006) and V6Av (Pitzalis et al., 2013d) retinotopically and defined human V6Ad functionally as a pointing-selective region situated anteriorly in the close proximity of V6Av (Tosoni et al., 2014). Like in macaque, human V6 is a motion area (e.g., Pitzalis et al., 2010, 2012, 2013a,b,c), while V6Av and V6Ad respond to pointing movements (Tosoni et al., 2014). The retinotopic organization (when present), anatomical position, neighbor relations, and functional properties of these three areas closely resemble those reported for macaque V6 (Galletti et al., 1996, 1999a), V6Av, and V6Ad (Galletti et al., 1999b; Gamberini et al., 2011). We suggest that information on objects in depth which are translating in space, because of the self-motion, is processed in V6 and conveyed to V6A for evaluating object distance in a dynamic condition such as that created by self-motion, so to orchestrate the eye and arm movements necessary to reach or avoid static and moving objects in the environment.

There is evidence that neurons in medial superior temporal area (MST) respond to rotation in depth of textured planes. MST neurons project to the ventral intraparietal area (VIP) and the question arises whether VIP neurons are responsive to rotation in depth as well. In the present study on awake monkeys, we have simulated movement of a flat board, covered with dots, by a computer. The two-dimensional images corresponded to the projection of structured planes rotating around a fronto-parallel axis. In the literature this stimulus is called fanning. Fanning effectively induced responses in VIP neurons. Most often the responses were nearly as strong as for translation, expansion/contraction, or rotation, indicating that there was no special sensitivity for rotation in depth. For neurons, sensitive to expansion, the response to fanning could often be explained by the positioning of the expanding part of the fanning stimulus over the area which was most responsive to expansion. For neurons which were direction selective to translation, the optimal direction of fanning was usually the same as the preferred direction for translation. It is concluded that VIP neurons may be sensitive to movement of structured planes but they are not specialized for the detection of such movement.

We recorded optokinetic eye movements of the crab, Carcinus maenas, in split-drum experiments. The patterns were either oscillated in antiphase on both sides mimicking translational image flow or they were oscillated in phase producing rotational image flow. Eye movements elicited by the rotational stimulus were larger than those produced by the pseudotranslational pattern movements. The smaller response to the latter is mainly a consequence of binocular interaction, the strength of which depends on both the phase-shift and amplitude of pattern oscillation. We develop two hypotheses to explain our results: either (1) signals from each eye modify the gain of the linkage signals coming from the other eye, or (2) the signals coming from the other eye modify the gain of the control loop itself. Quantitative evaluation of the data favors the second of these two hypotheses, which comprises the models of Barnes and Horridge (1969) and Nalbach et al. (1985). In addition, we found that it is the signals from the two slow channels of the crab's movement-detecting system that are transferred from one eye to the other, while signals of the fastest channel act almost exclusively ipsilaterally. We discuss our results as an adaptation by which an animal with panoramic vision compensates exclusively the rotational component of image flow during locomotion. The fact that freely walking crabs distinguish the two components of image flow better than restrained crabs indicates that further visual and nonvisual signals help to disentangle image flow.

During locomotion, observers respond to objects in the environment that may represent obstacles to avoid or landmarks for navigation. Although much is known about how visual cortical neurons respond to stimulus objects moving against a blank background, nothing is known about their responses when objects are embedded in optic flow fields (the patterns of motion seen during locomotion). We recorded from cells in the lateral suprasylvian visual area (LS) of the cat, an area probably analogous to area MT. In our first experiments, optic flow simulations mimicked the view of a cat trotting across a plain covered with small balls; a black bar lying on the balls served as a target object. In subsequent experiments, optic flow simulations were composed of natural elements, with target objects representing bushes, rocks, and variants of these. Cells did not respond to the target bar in the presence of optic flow backgrounds, although they did respond to it in the absence of a background. However, 273/423 cells responded to at least one of the taller, naturalistic objects embedded in optic flow simulations. These responses might represent a form of image segmentation, in that cells detected objects against a complex background. Surprisingly, the responsiveness of cells to objects in optic flow fields was not correlated with preferred direction as measured with a moving bar or whole-field texture. Because the direction of object motion was determined solely by receptive-field location, it often differed considerably from a cell's preferred direction. About a quarter of the cells responded well to objects in optic flow movies but more weakly or not at all to bars moving in the same direction as the object, suggesting that the optic flow background modified or suppressed direction selectivity.

It has been proposed that one area of extrastriate cortex in the cat, the lateral suprasylvian area (LS), plays an important role in visual analysis during locomotion (Rauschecker et al., 1987). Cells in LS reportedly tend to prefer directions along a trajectory originating at the center of gaze, and passing outward through the receptive-field center. Such directions coincide with the directions of image motion in an optic flow field, the pattern seen by locomoting observers when they fixate the point towards which they are heading (Gibson, 1950). We re-examined this issue for cells in LS with receptive fields in the lower visual field. Cells recorded posterior to Horsley-Clarke A2 showed a clear correlation between preferred direction and receptive-field location, but not that predicted: preferred directions were generally orthogonal to “optic flow” directions. Since these cells were all located posterior to those in studies showing a bias for “optic flow” directions, we hypothesized that there are two cell populations within LS, an anterior population that tends to prefer radial-outward directions, and a posterior population that tends to prefer directions orthogonal to radial. Data from earlier mapping experiments (Sherk & Mulligan, 1993) supported this idea.

We have examined the distribution of preferred directions of motion for neurons in the middle temporal visual area (MT) of the macaque. We found a marked anisotropy favoring directions that are oriented away from the center of gaze. This anisotropy is present only among neurons with peripherally located receptive fields. This peripheral centrifugal directionality bias corresponds well to the biased distribution of motions characteristic of optic flow fields, which are generated by displacement of the visual world during forward locomotion. The bias may facilitate the processing of this common form of visual stimulation and could underlie previously observed perceptual anisotropies favoring centrifugal motion. We suggest that the bias could arise from exposure of modifiable cortical circuitry to a naturally occurring form of selective visual experience.

Humans discriminate approaching objects from receding ones shortly after birth, and optic flow associated with self-motion may activate distinctive brain networks, including the human MT+ complex. We sought evidence for evoked brain activity that distinguished radial motion from other optic flow patterns, such as translation or rotation by recording steady-state visual evoked potentials (ssVEPs), in both adults and 4–6 month-old infants to direction-reversing optic flow patterns. In adults, radial flow evoked distinctive brain responses in both the time and frequency domains. Differences between expansion/contraction and both translation and rotation were especially strong in lateral channels (PO7 and PO8), and there was an asymmetry between responses to expansion and contraction. In contrast, infants' evoked response waveforms to all flow types were equivalent, and showed no evidence of the expansion/contraction asymmetry. Infants' responses were largest and most reliable for the translation patterns in which all dots moved in the same direction. This pattern of response is consistent with an account in which motion processing systems detecting locally uniform motion develop earlier than do systems specializing in complex, globally non-uniform patterns of motion, and with evidence suggesting that motion processing undergoes prolonged postnatal development.

The visual responsiveness and spatial tuning of frontal eye field (FEF) neurons were determined using a delayed memory saccade task. Neurons with visual responses were then tested for direction selectivity using moving random dot patterns centered in the visual receptive field. The preferred axis of motion showed a significant tendency to be aligned with the receptive-field location so as to favor motion toward or away from the center of gaze. Centrifugal (outward) motion was preferred over centripetal motion. Motion-sensitive neurons in FEF thus appear to have a direction bias at the population level. This bias may facilitate the detection or discrimination of expanding optic flow patterns. The direction bias is similar to that seen in visual area MT and in posterior parietal cortex, from which FEF receives afferent projections. The outward motion bias may explain asymmetries in saccades made to moving targets. A representation of optic flow in FEF might be useful for planning eye movements during navigation.

Neurons sensitive to optic flow patterns have been recorded in the the olivo-vestibulocerebellar pathway and extrastriate visual cortical areas in vertebrates, and in the visual neuropile of invertebrates. The complex spike activity (CSA) of Purkinje cells in the vestibulocerebellum (VbC) responds best to patterns of optic flow that result from either self-rotation or self-translation. Previous studies have suggested that these neurons have a receptive-field (RF) structure that “approximates” the preferred optic flowfield with a “bipartite” organization. Contrasting this, studies in invertebrate species indicate that optic flow sensitive neurons are precisely tuned to their preferred flowfield, such that the local motion sensitivities and local preferred directions within their RFs precisely match the local motion in that region of the preferred flowfield. In this study, CSA in the VbC of pigeons was recorded in response to a set of complex computer-generated optic flow stimuli, similar to those used in previous studies of optic flow neurons in primate extrastriate visual cortex, to test whether the receptive field was of a precise or bipartite organization. We found that these RFs were not precisely tuned to optic flow patterns. Rather, we conclude that these neurons have a bipartite RF structure that approximates the preferred optic flowfield by pooling motion subunits of only a few different direction preferences.

In birds, the nucleus of the basal optic root (nBOR) of the accessory optic system (AOS) and the pretectal nucleus lentiformis mesencephali (LM) are involved in the analysis of optic flow and the generation of the optokinetic response. In several species, it has been shown that the AOS and pretectum receive input from visual areas of the telencephalon. Previous studies in pigeons using anterograde tracers have shown that both nBOR and LM receive input from the visual Wulst, the putative homolog of mammalian primary visual cortex. In the present study, we used retrograde and anterograde tracing techniques to further characterize these projections in pigeons. After injections of the retrograde tracer cholera toxin subunit B (CTB) into either LM or nBOR, retrograde labeling in the telencephalon was restricted to the hyperpallium apicale (HA) of the Wulst. From the LM injections, retrograde labeling appeared as a discrete band of cells restricted to the lateral edge of HA. From the nBOR injections, the retrograde labeling was more distributed in HA, generally dorsal and dorso-medial to the LM-projecting neurons. In the anterograde experiments, biotinylated dextran amine (BDA) was injected into HA and individual axons were reconstructed to terminal fields in the LM and nBOR. Those fibers projecting to the nBOR also innervated the adjacent ventral tegmental area. However, tracing of BDA-labeled axons revealed no evidence that individual neurons project to both LM and nBOR. In summary, our results suggest that the nBOR and LM receive input from different areas of the Wulst. We discuss how these projections may transmit visual and/or somatosensory information to the nBOR and LM.

Using a novel suite of computer-generated visual stimuli that mimicked components of optic flow, the visual responses of the tropical land crab, Cardisoma guanhumi, were investigated. We show that crabs are normally successful in distinguishing the rotational and translational components of the optic flow field, showing strong optokinetic responses to the former but not the latter. This ability was not dependant on the orientation of the crab, occurring both in “forwards-walking” and “sideways-walking” configurations. However, under conditions of low overall light intensity and/or low object/background contrast, the separation mechanism shows partial failure causing the crab to generate compensatory eye movements to translation, particularly in response to low-frequency (low-velocity) stimuli. Using this discovery, we then tested the ability of crabs to separate rotational and translational components in a combined rotation/translation flow field under different conditions. We demonstrate that, while crabs can successfully separate such a combined flow field under normal circumstances, showing compensatory eye movements only to the rotational component, they are unable to make this separation under conditions of low overall light intensity and low object/background contrast. Here, the responses to both flow-field components show summation when they are in phase, but, surprisingly, there is little reduction in the amplitude of responses to rotation when the translational component is in antiphase. Our results demonstrate that the crab's visual system finds separation of flow-field components a harder task than detection of movement, since the former shows partial failure at light intensities and/or object/background contrasts at which movement of the world around the crab is still generating high-gain optokinetic responses.

Although a number of global mechanisms have been proposed over the years that explain how crabs might separate the rotational and translational components of their optic flow field, there has been no evidence to date that local mechanisms such as motion parallax are used in this separation. We describe here a study that takes advantage of a recently developed suite of computer-generated visual stimuli that creates a three-dimensional world surrounding the crab in which we can simulate translational and rotational optic flow. We show that, while motion parallax is not the only mechanism used in flow-field separation, it does play a role in the recognition of translational optic flow fields in that, under conditions of low overall light intensity and low contrast ratio when crabs find the distinction between rotation and translation harder, smaller eye movements occur in response to translation when motion parallax cues are present than when they are absent. Thus, motion parallax is one of many cues that crabs use to separate rotational and translational optic flow by showing compensatory eye movements to only the former.

An observer locomoting along a straight path sees a pattern of optic flow in which images move approximately radially outward from his heading point. If the observer turns, the optic flow field changes markedly because each object's image now has a component of horizontal motion added to its “optic flow” motion. We tested the responses of 326 cells in the cat's extrastriate area LS (lateral suprasylvian visual area) to movies simulating the optic flow seen during locomotion in a straight line, and during various simulated turns to the left and right. About 60% of 326 cells tested responded to optic flow simulating turns. Of most interest was a subset of cells, 15% of the total, that had “turn-selective” responses. They responded significantly better to turns in a particular direction (usually to the contralateral side) than to turns in the opposite direction or to optic flow simulating straight-ahead locomotion. For each cell, we generated a display of fronto-parallel motion with a direction and speed that matched the image motion in the preferred turn movie, as seen at the receptive-field center. Most turn-selective cells responded significantly better to their preferred turn movie than to this fronto-parallel stimulus. We examined the role of cells' selectivity for stimulus direction, speed, and acceleration in determining cell preference for particular turns. Direction preference played some role for most cells, but about a third of the cells preferred turn movies that did not reflect their direction selectivity. Other factors, including the presence of opposing motion, only rarely appeared to determine cell preferences for particular turn movies.

The ventral intraparietal area (VIP) is located at the end of the dorsal stream. Its neurons are known to have receptive-field characteristics similar to those of MT and MST neurons, but little is known about the temporal characteristics of VIP cells' responses. How fast are directionally selective responses evoked in the ventral intraparietal area after viewing optic flow patterns, and what are the temporal properties of these neuronal responses? To examine these questions, we recorded the activity of 37 directionally selective ventral intraparietal area (VIP) neurons in two awake macaque monkeys in response to optic flow stimuli with presentation times ranging from 17 ms to 2000 ms. We found a minimum response latency of 45 ms, and a median latency of 152 ms. Of all neurons, 10% showed early response components only (response latency < 150 ms and no activity in 500–2000 ms interval after stimulus onset), 55% only late response components (response latency >150 ms and sustained activity in 500–2000 ms interval), and 35% both early and late response components. Early responses appeared to very brief stimulus presentations (33-ms duration), while the late responses required longer stimulus durations. The directional selectivity was independent of optic flow duration in all cells. These results suggest that only a subset of neurons in area VIP may contribute to the fast processing of optic flow, while showing that the temporal properties of VIP responses clearly differ from the temporal characteristics of neurons in areas MT and MST.

The distribution of local preferred directions and motion sensitivities within the receptive fields of so-called tangential neurons in the fly visual system was previously found to match optic flow fields as induced by certain self-motions. The complex receptive-field organization of the tangential neurons and the recent evidence showing that the orderly development of the fly's peripheral visual system depends on visual experience led us to investigate whether or not early visual input is required to establish the functional receptive-field properties of such tangential neurons. In electrophysiological investigations of two identified tangential neurons, it turned out that dark-hatched flies which were kept in complete darkness for 2 days develop basically the same receptive-field organization as flies which were raised under seasonal light/dark conditions and were free to move in their cages. We did not find any evidence that the development of the sophisticated receptive-field organization of tangential neurons depends on sensory experience. Instead, the input to the tangential neurons seems to be “hardwired” and the specificity of these cells to optic flow induced during self-motions of the animal may have evolved on a phylogenetical time scale.

Ever since being described by Mountcastle (Mountcastle, 1957), columnar organization of sensory cortical areas has provided key leverage into understanding the functional organization of neocortex. Columnar or clusteredThe term column specifically denotes an organization in which groups of neurons with similar properties are extended perpendicularly to the cortical surface. Clustered organization, as used here, is intended to be similar but more general, where the groups of similar neurons need not have any particular geometry. Because of the limits of recording in the alert monkey, these cannot be distinguished in the present work, and the more inclusive term “clustered” will be used. organization of neurons sharing like properties is now known to be widespread, and probably universal in primary sensory areas. Visual cortex in primates consists of a primary area and a large number of secondary areas, which are organized in a manner both hierarchical and parallel (Felleman & Van Essen, 1991; Young, 1993; Young et al., 1995). One major component in the organization of extrastriate visual cortex appears to be the division into dorsal and ventral “streams” of processing (Ungerleider & Mishkin, 1982), each of which is organized hierarchically. Within each, columnar organization exists at early stages, but becomes less clear at higher levels. Columnar organization has been described at the highest level of the ventral stream, inferotemporal cortex (IT, Saleem et al., 1993; Fujita & Fujita, 1996; Tanaka, 1996), but has not been well characterized at the higher levels of the dorsal stream. Hints of such organization are found in the literature (Saito et al., 1986; Lagae et al., 1994), but systematic measurements are needed. In this paper, I report the existence of clustered organization in the medial superior temporal area (MST) of the dorsal stream, which is arguably the highest dominantly visual area on this pathway. I have measured the selectivity of both single- and multiple-unit activity along oblique electrode penetrations through this area to three different kinds of optic flow stimuli, and find that nearby neurons are more similar in their tuning than are more distant ones. This observation documents the existence of some form of clustered organization and supports the importance of this area in the processing of optic flow information.