Feature extraction and neural network-based analysis on time-correlated LiDAR histograms

Time correlated single photon counting (TCSPC) is used to obtain the time-of-flight (TOF) information generated by single-photon avalanche diodes. With restricted measurements per histogram and the presence of high background light, it is challenging to obtain the TOF information in the statistical histogram. In order to improve the robustness under these conditions, the concept of machine learning is applied to the statistical histogram. Using the multi-peak extraction method, introduced by us, followed by the neural-network-based multi-peak analysis, the analysis and resources can be focused on a small amount of critical information in the histogram. Multiple possible TOF positions are evaluated and the correlated soft-decisions are assigned. The proposed method has higher robustness in allocating the coarse position (± 5 %) of TOF in harsh conditions than the case using classical digital processing. Thus, it can be applied to improve the system robustness, especially in the case of high background light.