Object detection is one of the major methods of remote sensing. Generally in object detection features are extracted from the image to represent the object in the image numerically with lesser number of elements than the image itself. Out of multiple features, shape feature is one of the widely used features. In this study, new rotation invariant shape features are introduced and comparative studies are made among them by using them to detect a naturally occurring star-shaped object, palm trees, in aerial images. Detection of the palm trees helps in creating inventory which is an initial step towards monitoring. This study proposes shape features, the circular autocorrelation of polar shape matrix (CAPS), Mean of CAPS and Entropy of CAPS to be used with object detection framework. The framework used in the study to evaluate features uses a well-known machine learning algorithm, support vector machine (SVM), to detect palm trees in aerial images. The implementation on aerial images, taken with unmanned air vehicle (UAV) in five plantation regions in Indonesia, Malaysia, and Thailand, shows promising results. The CAPS produced the best results among the features studied, with an average accuracy of 84% over 8 images, chosen considering different challenges from the five plantation region.     «

Object detection is one of the major methods of remote sensing. Generally in object detection features are extracted from the image to represent the object in the image numerically with lesser number of elements than the image itself. Out of multiple features, shape feature is one of the widely used features. In this study, new rotation invariant shape features are introduced and comparative studies are made among them by using them to detect a naturally occurring star-shaped object, palm trees,...     »