Computational Intelligence in Automotive Applications

Abstract

AbstractIn this chapter, we discuss one of the most popular machine vision applications in the automotive industry: lane detection and tracking. Model-based lane detection algorithms can be separated into lane modeling, feature extraction and model parameter estimation. Each of these steps is discussed in detail with examples and results. A recently proposed lane feature extraction approach, which is called the Global Lane Feature Refinement Algorithm (GLFRA), is also introduced. It provides a generalized framework to significantly improve various types of gradient-based lane feature maps by utilizing the global shape information and subsequently improves the parameter estimation and the tracking performance. Another important aspect of this application lies in the tracking stage.We compare the performances of three different types of particle filters (the sampling importance resampling particle filter, the Gaussian particle filter and the Gaussian sum particles filter) quantitatively and provide insightful result analysis and suggestions. Furthermore, the influence of featuremaps on the tracking performance is also investigated.KeywordsFeature PointAutomotive ApplicationIntelligent Transportation SystemPosterior Probability DensityLane DetectionThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.