Structure-Aware Feature Disentanglement With Knowledge Transfer for Appearance-Changing Place Recognition.

Long-term visual place recognition (VPR) is challenging as the environment is subject to drastic appearance changes across different temporal resolutions, such as time of the day, month, and season. A wide variety of existing methods address the problem by means of feature disentangling or image style transfer but ignore the structural information that often remains stable even under environmental condition changes. To overcome this limitation, this article presents a novel structure-aware feature disentanglement network (SFDNet) based on knowledge transfer and adversarial learning. Explicitly, probabilistic knowledge transfer (PKT) is employed to transfer knowledge obtained from the Canny edge detector to the structure encoder. An appearance teacher module is then designed to ensure that the learning of appearance encoder does not only rely on metric learning. The generated content features with structural information are used to measure the similarity of images. We finally evaluate the proposed approach and compare it to state-of-the-art place recognition methods using six datasets with extreme environmental changes. Experimental results demonstrate the effectiveness and improvements achieved using the proposed framework. Source code and some trained models will be available at http://www.tianshu.org.cn.

Bioinspired Approach to Modeling Retinal Ganglion Cells Using System Identification Techniques.

The processing capabilities of biological vision systems are still vastly superior to artificial vision, even though this has been an active area of research for over half a century. Current artificial vision techniques integrate many insights from biology yet they remain far-off the capabilities of animals and humans in terms of speed, power, and performance. A key aspect to modeling the human visual system is the ability to accurately model the behavior and computation within the retina. In particular, we focus on modeling the retinal ganglion cells (RGCs) as they convey the accumulated data of real world images as action potentials onto the visual cortex via the optic nerve. Computational models that approximate the processing that occurs within RGCs can be derived by quantitatively fitting the sets of physiological data using an input-output analysis where the input is a known stimulus and the output is neuronal recordings. Currently, these input-output responses are modeled using computational combinations of linear and nonlinear models that are generally complex and lack any relevance to the underlying biophysics. In this paper, we illustrate how system identification techniques, which take inspiration from biological systems, can accurately model retinal ganglion cell behavior, and are a viable alternative to traditional linear-nonlinear approaches.

Multiscale Edge Detection Using a Finite Element Framework for Hexagonal Pixel-Based Images.

In recent years, the processing of hexagonal pixel-based images has been investigated, and as a result, a number of edge detection algorithms for direct application to such image structures have been developed. We build on this paper by presenting a novel and efficient approach to the design of hexagonal image processing operators using linear basis and test functions within the finite element framework. Development of these scalable first order and Laplacian operators using this approach presents a framework both for obtaining large-scale neighborhood operators in an efficient manner and for obtaining edge maps at different scales by efficient reuse of the seven-point linear operator. We evaluate the accuracy of these proposed operators and compare the algorithmic performance using the efficient linear approach with conventional operator convolution for generating edge maps at different scale levels.

Edge detecting for range data using Laplacian operators.

Feature extraction in image data has been investigated for many years, and more recently the problem of processing images containing irregularly distributed data has become prominent. Range data are now commonly used in the areas of image processing and computer vision. However, due to the data irregularity found in range images that occurs with a variety of image sensors, direct image processing, in particular edge detection, is a non-trivial problem. Typically, irregular range data would require to be interpolated to a regular grid prior to processing. One example of an edge detection technique than can be directly applied to range images is the scan-line approximation, but this does not employ exact data locations. Therefore, we present novel Laplacian operators that can be applied directly to irregularly distributed data, and in particular we focus on application to irregularly distributed 3D range data for the purpose of edge detection. Within the data distribution framework commonly occurring in range data acquisition devices, our results illustrate that the approach works well over a range of levels of irregularity of data distribution. The use of Laplacian operators on range data is also found to be much less susceptible to noise than the traditional use of Laplacian operators on intensity images.