Using Image Pattern Recognition Algorithms for Processing Video Log Images to Enhance Roadway Infrastructure Data Collection

摘要

In this research project, two innovative, modularized algorithms, sign detection and sign recognition, are developed. They form a solid foundation for developing an intelligent sign inventory and management system. A two-step sign inventory data collection process is proposed to seamlessly incorporate these two algorithms for batch processing millions of video log images, which can save great amounts of time and significant costs. The generalized sign detection algorithm, the first step in the intelligent sign inventory and management system, is developed using the shape, color, location, and other features of a traffic sign defined in the MUTCD standard. Sign shapes are detected using the polygon approximation approach; sign colors are processed with the Statistical Color Model (SCM) by using an Artificial Neural Network (ANN); the Probabilistic Distribution Function (PDF) of sign locations is obtained from the training video log images in which the sign locations are manually tagged. The generalized sign recognition algorithm, the second step in the intelligent sign inventory and management system, is developed based on the multi-feature fusion. The features include Haar features, sign color, sign shape, and sign PDF. Haar features encode the sign texture information using an Adaboost algorithm to generate strong classifiers with a boosting training approach.