Roadside vegetation classification is an essential task for roadside fire risk assessment and environmental surveys. The vegetation such as type of grasses and their biomasses are used to identify the fire risk, however it is very difficult to distinguish vegetation, in particular, the type of roadside grasses. The purpose of this study is to develop a technique which can distinguish vegetation structure and automatically identify fire risk. This paper presents a novel hybrid learning technique for the classification of roadside vegetation with a new feature extraction strategy. The hybrid technique is based on texture feature and fusion of three classifiers: Support Vector Machine (SVM), Neural Network (NN) and k-Nearest Neighbor (k-NN). The segmented image regions are created from image data and texture features are extracted. The three diverse classifiers are trained with extracted features and decisions are fused using the majority vote. The proposed hybrid learning technique has been evaluated on roadside data obtained from Queensland Transport and Main Roads and results are discussed.
Funding
Category 1 - Australian Competitive Grants (this includes ARC, NHMRC)
History
Volume
14
Issue
1
Start Page
1
End Page
6
Number of Pages
6
ISSN
1321-2133
Location
Canberra, Australia
Publisher
Australian National University
Language
en-aus
Peer Reviewed
Yes
Open Access
No
External Author Affiliations
Department of Transport and Main Roads; Not affiliated to a Research Institute; School of Engineering and Technology (2013- );
Era Eligible
Yes
Journal
Australian journal of intelligent information processing systems.