Image features and seasons revisited

Abstract

Abstract-We present an evaluation of standard image features in the context of long-term visual teach-and-repeat mobile robot navigation, where the environment exhibits significant changes in appearance caused by seasonal weather variations and daily illumination changes. We argue that in the given longterm scenario, the viewpoint, scale and rotation invariance of the standard feature extractors is less important than their robustness to the mid-and long-term environment appearance changes. Therefore, we focus our evaluation on the robustness of image registration to variable lighting and naturally-occuring seasonal changes. We evaluate the image feature extractors on three datasets collected by mobile robots in two different outdoor environments over the course of one year. Based on this analysis, we propose a novel feature descriptor based on a combination of evolutionary algorithms and Binary Robust Independent Elementary Features, that we call GRIEF (Generated BRIEF). In terms of robustness to seasonal changes, the GRIEF feature descriptor outperforms the other ones while being computationally more efficient.

Content-Based Image Retrieval Based on Late Fusion of Binary and Local Descriptors

“…The spectral texture techniques are dependant on shape and have shown good performance in-case of square regions [5]. The sensitivity to noise and computational good performance in various applications of image retrieval [7], [13]. Local feature descriptors are robust to translations, scaling, rotation, and small distortions and are applied in various image matching applications [7].…”

On Fpga Based Acceleration of Image Processing in Mobile Robotics

“…These comparisons form a set of unique binary tests which are subsequently stored into a q-dimensional bit vector. The pairwise comparisons can be chosen either randomly or evolutionary, e.g., they can be trained by methods of reinforcement learning [22], which optimize the descriptor for the particular environment.…”

Image features for visual teach-and-repeat navigation in changing environments

“…The work presented here broadens our previouslypublished analysis [19] by including new datasets ('Nordland' [18]), image features (SpG/CNN) and feature training schemes. In particular, we separate the influence of the detector and descriptor phases on the robustness of the feature extractors to appearance changes and demonstrate that combination of detection and description phases of different features can result in feature extractors that are more robust to seasonal variations.…”

“…Moreover, we evaluate a feature based on a Convolutional Neural Network (CNN) descriptor and a Superpixel Grid detector (SpG) [18]. We also propose a trainable feature descriptor based on evolutionary methods and binary comparison tests and show that this algorithm, called GRIEF (Generated BRIEF), and the SpG/CNN feature outperform the engineered image feature extractors in their ability to deal with naturally-occurring seasonal changes and lighting variations [19]. This adaptive approach allows to automatically generate visual feature descriptors that are more robust to environment changes than standard hand-designed features.…”