Comparison of pixel-based and object-based image classification techniques in extracting information from UAV imagery data

Publication Type:
Conference Proceeding
IOP Conference Series: Earth and Environmental Science, 2018, 169 (1)
Issue Date:
Full metadata record
© Published under licence by IOP Publishing Ltd. As the rapid development is being focused in the urban area, there is a need for the utilisation of a rapid system for updating this profile immediately. One of the current technologies being applied in recent years is the use of unmanned aerial vehicle (UAV) for mapping purposes. The use of UAV is widespread in various fields because it is low cost, has high resolution and is able to fly at low altitude without the constraints of cloudy weather. Typically, the method of data extraction for UAV in Malaysia is still very limited and the traditional methods are still being implemented by some industries. The features from aerial photo orthomosaic are manually detected and digitised from visual interpretation for the mapping purposes. Unfortunately, these methods are tedious, expensive, consume much time, and may involve much fieldwork, to acquire only a limited information. Pixel-based technique is often used to extract low level features where the image is classified according to the spectral information where the pixels in the overlapping region will be misclassified due to the confusion among the classes. The supervised object-based image analysis (OBIA) classification technique is widely used nowadays for automatic data extraction. Therefore, the general objective of this study is to assess the capability of UAV with high resolution data for image classifications. The pixel-based and OBIA classifications were compared using the Support Vector Machine (SVM) classifier. The classifications were assessed using different numbers of sample size. The result shows that OBIA gives a better result of Overall Accuracy (OA) than pixel-based. The consequences of this study accommodate further understanding and additional insight of utilising OBIA technique with different classifiers for the extended study.
Please use this identifier to cite or link to this item: