Discriminating Underwater LiDAR Target Signatures Using Sparse Multi-Spectral Depth Codes

The analysis and discrimination of underwater multi-spectral full-waveform LiDAR signatures acquired using a single-photon counting sensor is presented. We use a realistic scaled exemplar of a marine environment, with known and unknown targets, and show how we can both discriminate different materials and detect and locate mines. Each waveform is a temporal photon histogram whose inherent nature changes with the laser wavelength, target geometry and environment. Discriminatory dictionaries for target materials and mine types are learnt by making multi-spectral measurements. An accuracy of 97.8% and 98.7% was achieved for material and mine type discrimination, respectively.