A 32×32 Flash LiDAR SPAD Sensor with Up-to-1kfps Motional Target Detection by Threshold-adaptive 2D Dynamic Vision

The detection and tracking of 3D moving targets have attracted enormous attention in fields like unmanned vehicles and augmented reality. Single-photon avalanche detectors (SPAD) perform excellent ability in 3D imaging due to their ultra-high sensitivity and temporal resolution. However, flash light detection and ranging (LiDAR) systems using SPADs often require thousands of frames to measure distance due to background light interference. This leads to low frame rate and high power consumption, posing challenges in capturing fast-moving targets [1]. [2] introduced a Vision-LiDAR hybrid SPAD sensor that saves power by capturing depth frames only during inter-frame intensity changes. [3] used in-pixel histogramming to track the peak and preclude the detection in the non-peak range. However, both of them output full-format 3D images with redundant static scene pixels. [4] utilized stereo vision to identify the region of interest (ROI) for dynamic objects and facilitate data windowing in depth imaging but is limited by large size and short detection range. Additionally, traditional dynamic vision sensors (DVS) often employ fixed or global adjustable thresholds, failing to match different pixel-level intensity changes [5].