Framework for estimating distance and dimension attributes of pedestrians in real-time environments using monocular camera

摘要: Abstract Automatic distance and dimensions estimation of pedestrians are sometimes imperative in real-time scenes. Such estimations needful when contact based measurements unrealistic. It is desirable to have a non-contact measurement framework. This work exhibits method that obliges simple mathematical automatically discover the (height width) moving pedestrian lying at distant locations from camera. The proposed system confines immovable monocular camera environments. foremost step before single-shot environment learning. An L-shape marker used its cornered points detected stored by placing it first minimum then relatively far At two placements, deemed be joined four straight lines. With help line equations, per-pixel-length object's location calculated. mean filter applied for background subtraction extract foreground objects. Pedestrians classified passing objects convolutional neural network classifier. Afterward, calculated with reference smallest known between Thereafter, approach estimates height width pedestrian. Outcomes compared found results existing methods as well real measurements. show vigor framework worthy lapse rates.