LiDAR Saturated Waveform Compensation Based Real-time Ranging Method for Traffic Sign Detection

Saturated waveform reduces the accuracy of the distance and intensity measured by LiDAR, hampering the object detection and recognition in autonomous driving. To accurately extract the distance and intensity from the saturated waveform by the Gauss-Newton method, a saturated waveform compensation based real-time ranging method is proposed for enhancing the imaging and recognition of traffic signs with high reflectivity. The key steps of saturated waveform compensation are the calculation of the scale factor relying on the saturation length correction and the estimation of the delay factor. To fulfill the requirements of extracting waveform parameters in real time, the implementation on FPGA (ZYNQ-7000) is accelerated by the pipeline structure, which enables the ranging method to be executed at a laser repetition rate of up to 341.53 kHz. The experimental results demonstrated that, as the saturation ranged from 0 to 1631%, the ranging method consistently achieved a ranging standard deviation (RStD) below 0.86 cm and an absolute mean ranging error (MRE) less than 0.99 cm. After saturation compensation, the plane fitting error of the three-dimensional (3D) images of traffic signs was less than 0.44 cm, and the separability of intensity was obviously enhanced. This approach provides an effective solution for the detection and recognition of near-field high-reflectivity objects based on 3D images.