Frédéric Labrosse

The visual compass: performance and limitations of an appearance-based method

Frederic Labrosse. The visual compass: performance and limitations of an appearance-based method. Journal of Field Robotics, 23(10), pages 913-941, 2006

Short and long-range visual navigation using warped panoramic images

In this paper, we present a method that uses panoramic images to perform long-range navigation as a succession of short-range homing steps along a route specified by appearances of the environment of the robot along the route. Our method is different from others in that it does not extract any features from the images and only performs simple image processing operations. The method does only make weak assumptions about the surroundings of the robot, assumptions that are discussed. Furthermore, the method uses a technique borrowed from computer graphics to simulate the effect in the images of short translations of the robot to compute local motion parameters. Finally, the proposed method shows that it is possible to perform navigation without explicitly knowing where the destination is nor where the robot currently is. Results in our Lab are presented that show the performance of the proposed system.

A multi-resolution area-based technique for automatic multi-modal image registration

To allow remotely sensed datasets to be used for data fusion, either to gain additional insight into the scene or for change detection, reliable spatial referencing is required. With modern remote sensing systems, reliable registration can be gained by applying an orbital model for spaceborne data or through the use of global positioning (GPS) and inertial navigation (INS) systems in the case of airborne data. Whilst, individually, these datasets appear well registered when compared to a second dataset from another source (e.g., optical to LiDAR or optical to radar) the resulting images may still be several pixels out of alignment. Manual registration techniques are often slow and labour intensive and although an improvement in registration is gained, there can still be some misalignment of the datasets. This paper outlines an approach for automatic image-to-image registration where a topologically regular grid of tie points was imposed within the overlapping region of the images. To ensure topological consistency, tie points were stored within a network structure inspired from Kohonens self-organising networks [24]. The network was used to constrain the motion of the tie points in a manner similar to Kohonens original method. Using multiple resolutions, through an image pyramid, the network structure was formed at each resolution level where connections between the resolution levels allowed tie point movements to be propagated within and to all levels. Experiments were carried out using a range of manually registered multi-modal remotely sensed datasets where known linear and non-linear transformations were introduced against which our algorithms performance was tested. For single modality tests with no introduced transformation a mean error of 0.011 pixels was identified increasing to 3.46 pixels using multi-modal image data. Following the introduction of a series of translations a mean error of 4.98 pixels was achieve across all image pairs while a mean error of 7.12 pixels was identified for a series of non-linear transformations. Experiments using optical reflectance and height data were also conducted to compare the manually and automatically produced results where it was found the automatic results out performed the manual results. Some limitations of the network data structure were identified when dealing with very large errors but overall the algorithm produced results similar to, and in some cases an improvement over, that of a manual operator. We have also positively compared our method to methods from two other software packages: ITK and ITT ENVI.

An Area based Technique for Image-to-Image Registration of Multi-Modal Remote Sensing Data

This paper outlines further tests of the automatic image-to-image registration technique of Bunting et al with the inclusion of joint histogram image similarity metrics. These metrics were expected to increase the accuracy of the algorithms results given that previous literature has highlighted these metrics as optimal for multi-modal registration. But, this study has demonstrated the opposite with the correlation coefficient producing better matching results than the mutual information, kolmogorov distance and distance to independence metrics.

Development and Deployment of an Intelligent Kite Aerial Photography Platform (iKAPP) for Site Surveying and Image Acquisition

Aerial photographs and images are used by a variety of industries, including farming, landscaping, surveying, and agriculture, as well as academic researchers including archaeologists and geologists. Aerial imagery can provide a valuable resource for analyzing sites of interest and gaining information about the structure, layout, and composition of large areas of land that would be unavailable otherwise. Current methods of acquiring aerial images rely on techniques such as satellite imagery,manned aircraft, or more recently unmanned aerial vehicles (UAVs) and micro-UAV technologies. These solutions, while accurate and reliable, have several drawbacks. Using satellite imagery or UAVs can prove to be very expensive, costing tens of thousands for images. They can also prove to be time-consuming and in some cases have constraints on use, such as no-fly zones. In this paper, we present an alternative low-cost, versatile solution to these methods, an intelligent kite aerial photography platform (iKAPP), for the purpose of acquiring aerial images and monitoring sites of interest.We show how this system provides flexibility in application, and we detail the system’s design, mechanical operation, and initial flight experiments for a low-cost, lightweight, intelligent platform capable of acquiring high-resolution images. Finally, we demonstrate the system by acquiring images of a local site, showing how the system functions and the quality of images it can capture. The application of the system and its capabilities in terms of capture rates, image quality, and limitations are also presented. The system offers several improvements over traditional KAP systems, including onboard “intelligent” processing and communications. The intelligent aspect of this system stems from the use of self-image stabilization of the camera, the advantage being that one is able to configure the system to capture large areas of a site automatically, and one can see the site acquisition in real time, all of which are not possible with previous methods of AP.

Rotation-invariant appearance based maps for robot navigation using an artificial immune network algorithm

The treatment of image data for robotic applications such as navigation, path planning and localization has always been problematic when working in image space (using the appearance of the environment) rather than in Cartesian space (using the geometry of the environment). This is due to both computational overhead introduced by the large amount of data that needs to be manipulated and the high-dimensionality of the image space. We present results from an approach using an artificial immune network construction algorithm which dramatically reduces the dimensionality of the image space and generates network structures useful for navigation and localization. The technique uses the artificial immune network mechanism to link images with similar properties, thus corresponding to similar poses of the robot, into a network which can be displayed in two dimensions. This generates an intuitive representation of the environment which the robot has experienced in a way which can also be traversed in order to perform path-planning in the space of visual experiences. A network generated as a mobile robot moves around in its environment is presented, and related topologically to the movements made by the robot. Properties of the networks produced are discussed with relation to the visual complexity of the environment experienced by the robot. In general, regions of the environment which appear homogeneous produce fewer nodes and edges in the network, and regions of a more heterogeneous appearance produce denser, more highly connected network structures.

A Vector-based Representation for Image Warping

A method for image analysis, representation and re‐synthesis is introduced. Unlike other schemes it is not pixel based but rather represents a picture as vector data, from which an altered version of the original image can be rendered. Representing an image as vector data allows performing operations such as zooming, retouching or colourising, avoiding common problems associated with pixel image manipulation. This paper brings together methods from the areas of computer vision, image compositing and image based rendering to prove that this type of image representation is a step towards accurate and efficient image manipulation.

On the Use of Ray-tracing for Viewpoint Interpolation in Panoramic Imagery

To acquire seamless visualization of environments fromdifferent viewing positions and orientations, it is desirableto generate virtual images for an arbitrary position givena set of reference views. In this paper, a simple interpolationmethod based on ray-tracing is proposed for viewpointsynthesis from panoramas taken with multi-sensor cameras.Instead of attempting to recover a dense 3D reconstructionof the scene, the method estimates the pose between eachpanorama and then backward projects the point along theray that exhibits the best colour consistency. We show thatby limiting the search space using sparse depth information,both the speed and the accuracy of the interpolationare improved.

Automatic Driving on Ill-defined Roads: An Adaptive, Shape-constrained, Color-based Method

Autonomous following of ill-defined roads is an important part of visual navigation systems. This paper presents an adaptive method that uses a statistical model of the color of the road surface within a trapezoidal shape that approximately corresponds to the projection of the road on the image plane. The method does not perform an explicit segmentation of the images but instead expands the shape sideways until the match between shape and road worsens, simultaneously computing the color statistics. Results show that the method is capable of reactively following roads, at driving speeds typical of the robots used, in a variety of situations while coping with variable conditions of the road such as surface type, puddles, and shadows. We extensively evaluate the proposed method using a large number of datasets with ground truth available from http://www.aber.ac.uk/en/cs/research/ir/dss/. We moreover evaluate many color spaces in the context of road following, and we find that the color spaces that separate luminance from color information perform best, especially if the luminance information is discarded.

Picture This! Community-Led Production of Alternative Views of the Heritage of Gwynedd

Abstract The digital camera has become ubiquitous. Every mobile phone has one built in, almost everyone has a mobile phone, and people use them constantly for all kinds of things, including taking pictures. In a new collaborative project, funded by the Arts and Humanities Research Council (AHRC), Bangor, Aberystwyth and Manchester Metropolitan Universities have teamed up with Gwynedd Archaeological Trust to develop tools to allow communities to picture their heritage and upload the images to an automated photogrammetry server to create metrical 3D models of the sites and objects they are recording. The data created will then feed into the local Historic Environment Record, providing a valuable tool for monitoring changes to heritage sites, while providing communities with added information and alternative views of their heritage. This paper is not intended to provide a formal research design or a fully developed prototype. Rather, it is intended to outline an experimental and collaborative approach that is situated as both practice and research, with neither enterprise being privileged over the other. The activities outlined here will be developed and evaluated over the next year and a half, after which we will report on whether or how the contingent aims and outcomes expressed were realized.