A new range-only measurement-based glass line feature extraction method

Abstract

This letter addresses a new range-only measurement-based glass line feature extraction method, which allows a low-cost light detection and ranging (LiDAR) sensor to detect glass-like objects without intensity and multi-echo data. The principle that the emitted LiDAR ray incidence angle with respect to a glass surface is close to 0 degrees is concerned. The geometric relationship between the sensor and glass objects is also considered for successful glass feature extraction. The extracted features are accumulated and reformulated for the generation of glass line features using the RANdom SAmple Consensus algorithm (RANSAC). Two experiments with different robot paths were conducted in a glass environment. The results showed that the proposed method accurately represented glass objects without any false-positive glass feature.