Road Scene Detection : Docker Vision

Object detection, a computer vision technology designed to locate all instances of the object present in an image, and it has been most widely accomplished by searching all locations in the image, at all possible scales, and comparing a small area at each location with known templates or patterns of the object. Human and vehicle detection and tracking are tasks of computer vision systems for locating and following people and vehicles in video footage.

Cars and person tracking from CCTV camera

Detecting human beings accurately in a visual surveillance system is crucial for diverse application areas including abnormal event detection, human gait characterization, congestion analysis, person identification, gender classification and fall detection for elderly people. Human tracking is the process of temporally associating the human detections within a video sequence to generate persistent paths, or trajectories, of the people.

As an essential task in Intelligent Transportation Sysytem or Autonomous Vehicles, vehicle detection aims to provide information assisting vehicle counting, vehicle speed measurement, identification of traffic accidents, traffic flow prediction, etc.

How detection is done using Computer Vision and Deep Learning?

This task of object detection uses a combination of Computer Vision and Deep Learning technique to detect the objects of interest in the frame. It also tracks the objects as well as detects the direction of the objects like whether it is moving towards or away from the camera, taking left turn etc. The process also gives the total number of vehicles and people that passed the road at a given time.

The deep learning model which detects cars and person in the given videos is YOLO (You Only Look Once) trained on COCO dataset. YOLO model can detect upto 80 classes of objects. It also has a tiny light weight architecture which compromises on accuracy but gives improved fps. The objects in the videos are tracked using Centroid Tracker. It uses euclidean distance to track the objects in the successive frame.

From the above video it is clear that when an object moves out of frame, the ID is retained for some of the successive frames. This is to avoid occlusion scenario where an object is hidden by another object. The below video depicts an example of the occlusion scenario where ID 50 and ID 51 are occluded by other vehicles and the lighting, but retains the same ID.

Docker Vision

Visit: http://dockervision.com