Change detection from remotely sensed images: From pixel-based to object-based approaches

Abstract

The appetite for up-to-date information about earth’s surface is ever increasing, as such information provides a base for a large number of applications, including local, regional and global resources monitoring, land-cover and land-use change monitoring, and environmental studies. The data from remote sensing satellites provide opportunities to acquire information about land at varying resolutions and has been widely used for change detection studies. A large number of change detection methodologies and techniques, utilizing remotely sensed data, have been developed, and newer techniques are still emerging. This paper begins with a discussion of the traditionally pixel-based and (mostly) statistics-oriented change detection techniques which focus mainly on the spectral values and mostly ignore the spatial context. This is succeeded by a review of object-based change detection techniques. Finally there is a brief discussion of spatial data mining techniques in image processing and change detection from remote sensing data. The merits and issues of different techniques are compared. The importance of the exponential increase in the image data volume and multiple sensors and associated challenges on the development of change detection techniques are highlighted. With the wide use of very-high-resolution (VHR) remotely sensed images, object-based methods and data mining techniques may have more potential in change detection.

Introduction

In remote sensing (RS) applications, changes are considered as surface component alterations with varying rates. Land-cover (LC) and land-use (LU) change information is important because of its practical uses in various applications, including deforestation, damage assessment, disasters monitoring, urban expansion, planning, and land management. Singh (1989) defined change detection (CD) as “the process of identifying differences in the state of an object or phenomenon by observing it at different times”. The CD frameworks use multi-temporal datasets to qualitatively analyze the temporal effects of phenomena and quantify the changes. The RS data has become a major source for CD studies because of its high temporal frequency, digital format suitable for computation, synoptic view, and wider selection of spatial and spectral resolutions (Chen et al., 2012a, Coops et al., 2006, Lunetta et al., 2004). The general objectives of CD in RS include identifying the geographical location and type of changes, quantifying the changes, and assessing the accuracy of CD results (Coppin et al., 2004, Im and Jensen, 2005, Macleod and Congalon, 1998).

Developing CD methods in RS is an ongoing research agenda. The principle behind using RS data for CD is that changes in the object of interest will alter the spectral behavior (reflectance value or local texture) that is separable from changes caused by other factors (e.g. atmospheric conditions, illumination and viewing angles, and soil moistures) (Deer, 1995, Green et al., 1994, Jensen, 1983, Singh, 1989). The CD from RS data is affected by various elements including spatial, spectral, thematic and temporal constraints, radiometric resolution, atmospheric conditions, and soil moisture conditions (Jensen, 2005). Different CD techniques have been developed in the past, depending on the requirements and conditions. However, the selection of the most suitable method or algorithm for change detection is not easy in practice (Lu et al., 2004). Researchers have made enormous efforts in developing various change detection methodologies including both the traditional pixel-based (Mas, 1999) and more recently, the object-based (Araya and Hergarten, 2008).

Various CD reviews based on pixel-based analysis of RS data have been published (see e.g. Coppin et al., 2004, Deer, 1995, Jianyaa et al., 2008, Lu et al., 2004, Mouat et al., 1993, Singh, 1989) which have summarized and categorized CD techniques based on different viewpoints. A common one is grouping them into pre-classification and post-classification CD techniques (Chen et al., 2012c). Chan et al. (2001) categorized them as change enhancement techniques and nature-of-change techniques. Lu et al. (2004) presented a comprehensive review and grouped CD techniques into seven categories. They categorized an exhaustive list of CD studies with respect to the change domain or applications aspects. Most of these reviews cover CD techniques for coarse and relatively fine spatial resolution data such as MODIS, Landsat (MSS, TM), SPOT, and radar. However, these reviews do not extensively examine techniques and methods suitable for data from very-high-resolution (VHR) optical satellites such as IKONOS, QuickBird, GeoEye, RapidEye, EROS A and B. The object-based imaged analysis techniques are considered more suitable for VHR image data and some discussion can be found in (Blaschke, 2010, Chen et al., 2012a, Jianyaa et al., 2008, Lang, 2008).

In this paper we place the CD methodologies into two discrete groups based on the unit of image analysis. The first is the traditional/classical pixel-based, employing an image pixel as fundamental unit of analysis. The second group is the object-based method, emphasizing, first, creating image objects and then using them for further analysis. This paper is broadly organized into three parts. First pixel-based change detection (PBCD) methods are discussed followed by object-based change detection (OBCD) techniques. Also, the spatial data mining techniques are discussed for their potential for analyzing changes from RS data. The focus of this paper is to provide a review of commonly used CD methods and techniques in RS, their applications, and related issues.