Roadside Video Data Analysis by Brijesh Verma Ligang Zhang & David Stockwell

Author:Brijesh Verma, Ligang Zhang & David Stockwell
Language: eng
Format: epub
Publisher: Springer Singapore, Singapore

3.6 Majority Voting Based Hybrid Learning

3.6.1 Introduction

Designing a robust classification algorithm is one of the most crucial components in building an automatic system for video data analysis [40]. There are many algorithms proposed for solving the classification problem including KNN [41], AdaBoost [42], ANN [43], SVM [44], Wavelet based techniques [45], etc. However, none of these algorithms can guarantee any form of optimality in terms of proper classification accuracy in realistic environments. In recent years, the hybrid approach of combining multiple classifiers [46, 47] rather than a traditional way of using a single classifier has received increasing attention due to its effectiveness and robustness in classification problems, particularly in real-world scenarios [48]. However, the majority of existing approaches to vegetation segmentation focus on using a single classifier, and few studies have explored roadside vegetation segmentation using video data captured by vehicle-mounted cameras.

This section describes a majority voting based hybrid approach [49] to combine multiple classifiers for dense versus sparse vegetation classification. Three types of classifiers including ANN, SVM and KNN are specifically selected to improve the diversity, which helps achieve better performance. The strengths of the majority voting based approach include the incorporation of multiple classifiers with majority voting and a new feature extraction technique. However, because the approach combines multiple classifiers, it requires a longer time to process all classifiers and thus poses a challenge for real-time processing.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.