Analysis of hybrid classification approach to differentiate dense and non-dense grass regions

Vegetation classification from satellite and aerial images is a common research area for fire risk assessment and environmental surveys for decades. Recently classification from video data obtained by vehicle mounted video in outdoor environments is receiving considerable attention due to the large number of real-world applications. However this is a very challenging task and requires novel research techniques. This paper presents an analysis of hybrid classification approach to distinguish vegetation in particularly the type of roadside grasses from videos recorded by the Queensland transport and main roads. The proposed framework can distinguish dense and non-dense grass regions from roadside video data. While most of the recent works focuses on infrared images, proposed approach uses image texture feature for vegetation region classification. Analysis of hybrid approach using texture feature and multiple classifiers is the main contribution of this research work. The classifiers include: Support Vector Machine (SVM), Neural Network (NN), k-Nearest Neighbor (k-NN), AdaBoost and Naïve Bayes. The different images were created from video data containing roadside vegetation in various conditions for training and testing purposes. The hybrid classification approach has been analysed on roadside data obtained and results are discussed.