Data-Driven Bridge Detection in Compressed Domain from Panchromatic Satellite Imagery

摘要

Bridge detection in panchromatic imagery is of great importance in civilian and military applications. Popular algorithms for bridge detection are often based on a priori knowledge to bridge structure or location features, where manually-introduced decision rules are incorporated into a complex algorithm in spatial domain. Instead of knowledge-based approach in spatial domain, in this paper, we proposed a fast data-driven algorithm in compressed domain for panchromatic satellite imagery. Our algorithm consists of two main steps: firstly, bridge region candidates detection with hierarchical saliency model in compressed domain; and secondly, bridge region candidates validation with Local Binary Patterns (LBP) and Extreme Learning Machine (ELM). Experiments are conduced, and detection results demonstrate the effectiveness and efficiency of our proposed algorithm. The main contributions of our work are twofold: 1) to the best of our knowledge, we are among the first to introduce the concept of compressed domain techniques for bridge detection; and 2) compared with other knowledge-based algorithms, no assumptions are made beforehand for our algorithm, which makes it applicable for bridges of various cases.