You are here

MultiSensor Homogenization and Change Detection

Project researchers

For decades, multispectral sensor images have been a stable and popular data source in Remote Sensing Change Detection (RS CD) application fields. This is mainly due to the availability of long time series databases and in the last decade to the free distribution of this information, at least for medium resolution sensors (e.g., MODIS, Landsat). With the development of sensor technology, the new generation satellite sensors have been capable to acquire images with higher spatial and spectral resolution and with pass of time, to offer a VHR time series database. For acquiring VHR data over any specific area, a previous order to the owners of the satellite sensors must be done. This implies that a continuous mapping over the same area is not always possible from a single satellite sensor, thus VHR time series with good time resolution can be constructed by using images from different satellite sensors. Therefore, for a proper time series CD analysis, development of novel techniques are required to address problems such as: i) analysis of high (or very high) spatial resolution multitemporal images and; ii) analysis of multi-sensor multitemporal images. VHR images contain richer information for detecting changes than standard multispectral images, but also pose great challenges in the CD process. In this research, both topics are analyzed with the goal to define effective CD and homogenization techniques for multi-sensor images.

In the literature several works can be found where methodologies for information extraction from single date acquisitions are presented  and only few deal with bi-temporal images. However, most of the works have been focused on change detection (mainly for low and medium spatial resolution), but not in dealing with time series of VHR images and not even with multi-sensor. Due to the data complexity and the lack of previous research orientated in the same direction, the problem of CD in multi-sensor multitemporal VHR images needs to be further investigated.

Objectives: 

We aim at developing an approach to the detection of changes in multisensor multitemporal VHR optical images. The main steps of the proposed method are: i) multisensor data homogenization; and ii) change detection in multisensor multitemporal VHR optical images.

Fig.1. General block scheme followed for the MS integration and change detection process

Research topics: