An ensemble approach for the detection and classification of mixed pixels of remotely sensed images

Accuracy has been necessity condition to render fine spatial resolution in the mapping of land patches. Every remotely sensed image can be characterized as objects whose accuracy varies as a function of the spatial resolution. The objects can be assigned a spectral signature which demonstrates the reflection factor of the pixels. This further prompts the confusion of pixels occupying more than one class. Such pixels can be termed as mixed pixels, whereas pixels of a homogeneous class designated as pure pixels. This paper sights the mixed pixels of the satellite images and focuses on their classification employing the texture feature. The primary reason of focus on texture is that it is based on the spatial arrangement of intensities of an image. The texture features such as colour intensity, energy and local binary pattern are used to analyze the mixed pixel problem. Supervised learning techniques such as artificial neural network and ensemble bagged learning has been reviewed and compared to handle the mixed pixel issue, thus imparting better resolution.