Abstract:Mobile mapping is a new technology for capturing georeferenced data. It is, however, still not practical to extract spatial and attribute information of objects such as infrastructure elements filly automatically. In this article, a new framework for 30 object recognition by hypothesis-and-test techniques is proposed and developed. An example of traffic-light recognition from mobile mapping images is given in detail. The hypothesis is generated according to the viewpoint dependent theory. We formulate the hypothesis test problem based on Bayesian inference and, in particular, the (Maximize A Posteriori Probability). This approach functions in two major steps: (1) generation of hot-spot maps by vanishing point detection and template matching, and (2) estimation of the parameters of 3D objects (traffic lights) by Markov Chain Monte Carlo (MCMC). The developed hot-spot map generation method is, in general, faster than general color image segmentation algorithms. For example, it can handle the recognition problem with a color image of 720 by 400 pixels within a couple of minutes rather than tens of minutes to even hours when using the segmentation algorithms. The parameter estimation method uses MCMC to simulate an ergodic stochastic process so that a robust and global optimal solution can be found. The approach shows great potential for automatic object recognition in image sequences acquired by mobile mapping systems. lntroductlon Automatic recognition of 3D objects from color images is a challenging, yet unsolved, problem. Furthermore, recognition of spatial features from images acquired in outdoor scenes, outside of a controlled laboratory environment, by a mobile mapping system (Li, 1997) poses an even more difficult research topic. The ways in which the data are acquired, for example using active or passive sensors, may affect the methods of object recognition. In this paper, we mainly discuss object recognition from color mobile mapping images and show how traffic lights, in particular, are recognized by the proposed system. The human stereo vision system is an extremely comprehensive and effective system that functions very fast and accurately to support human decision-making processing in an ever-changing environment. "How are 3D objects represented in the human visual system?" becomes the initial question we ask if we want to produce a similar visual system (Bulthoff et al., 1994). Different answers to this question yield different model representations, and thus lead to different approaches. ?tvo common answers to this question are viewpoint independent and viewpoint dependent approaches that are further

A framework of traffic lights detection, tracking and recognition based on motion models

A Robust Video Based Traffic Light Detection Algorithm for Intelligent Vehicles

&Lt;title>automatic Traffic Real-Time Analysis System Based on Video</title>

A Fusion Method Aiming at Environmental Perception of Autonomous Vehicle Based on Visual Scheme

Traffic Light Detection and Recognition for Autonomous Vehicles

Robust and Real-Time Traffic Lights Recognition in Complex Urban Environments

Simultaneous detection and tracking using deep learning and integrated channel feature for ambint traffic light recognition

Spatial-Temporal-Rule-based Traffic Movement Status Interpretation at Intersections *

Traffic Lights Detection Based on Deep Learning Feature.

Traffic Light Recognition for Complex Scene with Fusion Detections.

An algorithm for moving vehicle detection and tracking based on adaptive background modeling

The recognition and tracking of traffic lights based on color segmentation and CAMSHIFT for intelligent vehicles

Real-Time Vehicle Detection Framework Based on the Fusion of LiDAR and Camera

Real-time recognition system of traffic light in urban environment

Traffic lights detection and tracking for HD map creation

A Framework for Automatic Recognition of Spatial Features from Mobile Mapping Imagery

An Incremental Framework for Video-Based Traffic Sign Detection, Tracking, and Recognition

Traffic lights recognition based on concatenated filtering method

Detection and classification of traffic lights for automated setup of road surveillance systems

Research on Road Environmental Sense Method of Intelligent Vehicle Based on Tracking Check