An Edge-Less Approach to Horizon Line Detection

Abstract

Abstract-Horizon line is a promising visual cue which can be exploited for robot localization or visual geo-localization. Prominent approaches to horizon line detection rely on edge detection as a pre-processing step which is inherently a non-stable approach due to parameter choices and underlying assumptions. We present a novel horizon line detection approach which uses machine learning and Dynamic Programming (DP) to extract the horizon line from a classification map instead of an edge map. The key idea is assigning a classification score to each pixel, which can be interpreted as the likelihood of the pixel belonging to the horizon line, and representing the classification map as a multi-stage graph. Using DP, the horizon line can be extracted by finding the path that maximizes the sum of classification scores. In contrast to edge maps which are typically binary (edge vs no-edge) and contain gaps, classification maps are continuous and contain no gaps, yielding significantly better solutions. Using classification maps instead of edge maps allows for removing certain assumptions such as the horizon is close to the top of the image or that the horizon forms a straight line. The purpose of these assumptions is to bias the DP solution but they fail to produce good results when they are not valid. We demonstrate our approach on three different data sets and provide comparisons with a traditional approach based on edge maps. Although our training set is comprised of a very small number of images from the same location, our results illustrate that our method generalizes well to images acquired under different conditions and geographical locations.

Fusion of edge-less and edge-based approaches for horizon line detection

“…Both machine learning based edge-less [4] and edge-based [3] approaches outperform the classical edge-based approach [5] by a high margin (see Table I). A quick look of table I reveals that edge-based method [3] outperforms edge-less approach [4]. To better understand the strengths and weaknesses of each method, we present below several reasons.…”

“…Although the full DCSI can be used for horizon line detection, it was found that keeping only the m highest classification scores in each column does not compromise accuracy while reduces computations. The reduced DCSI is referred to as mDCSI [4]. The multi-stage graph corresponding to the mDCSI contains less vertices; as a result, fewer paths need to be considered when searching for the shortest path which results in considerable speedups.…”

“…In [4] Ahamd et al proposed a machine learning approach for horizon line detection. Specifically, they apply DP on a classification map that associates with each pixel a classification score which can be interpreted as the confidence of the pixel being part of the horizon line.…”

“…In section 2, we review the edge based approaches of Lie et al [5] and Ahmad et al [2], [3], [18]. Section 3 reviews the edge-less approach of Ahmad et al [4]. Section 4 presents the proposed fusion strategy while section 5 presents our experimental results.…”

“…More recently, Ahmad et al [4] have proposed an edgeless approach for horizon line detection. Their approach is motivated by the fact that conventional edge based methods [ [2], [1], [5]] suffer from gaps and provide no confidence about the likelihood of an edge point belonging to the horizon.…”