PMC

The framework of the proposed nighttime pedestrian detection algorithm and system. (Left): We build a network binocular camera with two calibrated commercial near-infrared cameras; (Middle): The voxel surface model for foreground pedestrian detection in dynamic scenes; (Right): Three kinds of outputs of our system, including the foreground binary image, foreground cluster depth and bounding boxes of the pedestrian.

(a) Foreground image. (b) Color image of foreground subtracts background. (c) Color image of background subtracts foreground. (d) Gray level image of (b). (e) Gray level of (c). (f) Addition of (d) and (e). (g) Foreground object after intensity adjustment. (h) Binary image of foreground object. (i) Foreground object after morphological opening. (j) Extracted silhouette.

Example of image processing using polar coordinate extreme value projection (PCEVP) processing. (A) Determine the center of gravity of the foreground region (B) by calculating all local maximal and minimal value points on the boundary of the foreground region. (C) Individual foreground regions in (B) segmented using PCEVP. (D) The Euclidean distance and extreme value points of the foreground region boundary. (E) The Euclidean distance in bar graph in a 3D coordinate system.

RSSM facilitates robust quantification on the estimated foreground. (A) Neurites images collected using fMOST system and their corresponding tracing results. (B) Neuronal images with a complete single neuron from the public BigNeuron (left) and Diadem (right) dataset and their tracing results. (C) Neuronal images collected using light-sheet microscopy and their corresponding soma localization results. (D–F) Are the estimated foreground images by RSSM and their corresponding quantitative results. The NeuroGPS-Tree software is used for neurite tracing and soma localization. Thre is a key threshold parameter in NeuroGPS-Tree for neurite tracing and soma localization.