A Summary of Change Detection Technology of Remotely-Sensed Image

Zhou Shilun

ABSRACT：This paper will describe three aspects of change detection technology of remotely-sensed images. At first, the process of change detection is presented. Then, the author makes a summary of several common change detection methods and a brief review of the advantages and disadvantages of them. At the end of this paper, the applications and difficulty of current change detection techniques are discussed.

KEY WORDS：remote sensing;change detection;image registration

1 INTRODUCTION

Change detection technology is an important research field of remote sensing information science, as one of the main development directions of currently remote sensing data processing technology. Change detection is the process of identifying differences in the state of an object or phenomenon by observing it at different times. And the essence of change detection using remotely-sensed images is to detect the changes of pixel spectral response in images from different dates, which are caused by changes of surface features in field of view. Different change detection techniques have been developed in the past, depending on the requirements and conditions. Details on the process of most change detection techniques and principle of several methods are discussed in later sections.

2 THE PROCESS OF CHANGE DETECTION

A complete process of change detection which uses remotely-sensed image typically includes: the selection of remotely-sensed image, image preprocessing, change detection, change information post-processing and result output. This will be followed by a detailed presentation of important parts of the process.

2.1 Image Preprocessing

The remotely-sensed image preprocessing includes image enhancement and filtering, geometric correction, radiometric correction, image cropping and image mosaic, etc. The purpose of that is to highlight the changed objects, and improve the ability to interpret. Among those steps, geometric correction and radiometric correction have the greatest impact to the correction of change detection results.

Change detection can be seen as an operation that compares the multitemporal data of the same area, and obviously the operation needs the consistent of position reference and grayscale range of the same object in different images. Because of the effect of the gesture of aircraft platforms and sensors, atmospheric transmission, terrain, sensor imaging performance and other factors, the geometry of image always has linear and nonlinear image distortion, making corresponding feature in the same coordinate position in multitemporal images inconsistent. Besides, sensor calibration, solar elevation angle, the ground moisture degrees and seasonal variations can cause the changes in grayscale display of the same features. Therefore, most of common change detection techniques, especially the pixel-based, need the geometric correction and radiometric correction process before change detection. The accuracy of geometric correction and radiometric correction effect the accuracy of change detection, and low-precision geometric and radiometric calibration results can cause a lot of pseudo change detection results.

2.2 Change Detection

Change detection is to choose the right method depending on the application, extract and analyze change information, and finally output detection results. Change detection is the core of the whole process, which determines the data preparation and preprocessing, and restricts the scope of change detection method selection. In Section 3, we describe several common change detection methods.

2.3 Change Information Post-processing

Due to the influence of noise, after making discriminant, detection results will contain many false alarms. The function of post-processing part is to filter out some false alarms, remove the changes which are not interested in, and to do the mathematical morphology process of the binary image after classification.

3 CHANGE DETECTION TECHNIQUES

Change detection can be classified into two categories according to image registration and data sources. The first category is change detection after image registration and the second is change detection simultaneous with image registration. In the first category, using single new images and old images for change detection is the most popular method in the change detection application. In this section, several common methods of change detection between new images and old images are described.

3.1 Image Differencing

Image differencing method is the simplest and most commonly method used to detect change. The basic principle is to get two images at different time after image registration, and then make the gray value of the corresponding pixels in two images subtract to obtain a difference image to indicate the changes of target area at the time which is selected. Theoretically, in the obtained image, the area where the difference is zero or close to zero is considered as the constant region. Image differencing method is simple and easy to interpret results, but using this method we cant get complete matrices of change information. In addition, optimal threshold selection of practical applications depends on the grayscale range of the image and cant be easily determined. The difference value is absolute. Therefore same value may have different meaning depending on the starting class.

3.2 Image Ratioing

Ratioing is considered to be a relatively rapid means of identifying areas of change. In ratioing two registered images from different dates with one or more bands in an image are ratioed, band by band. The data are compared on a pixel by pixel basis. By doing relative radiometric correction of images from different dates, change information of ratioing image is enhanced. The area where the ratio of pixels is one or approximately one is considered as the constant region.

Ratioing method bases on the assumption that the ratio of the image showed a Gaussian distribution, but for many practical problems that assumption is not always true. Under these circumstances, the choice of the threshold becomes a valid key. Image ratioing method and image differencing method are intuitive and easy to grasp, and the detection speeds of them are fast. But because these algorithms are too simple, it is difficult taking all factors into account, which is likely to cause the loss of large amounts of information. Besides, the method demands high accuracy of image registration.

3.3 Principal Components Analysis (PCA)

PCA, mathematically based on “Principal Axis Transformation”, is a transformation of the multivariate data to a new set of components, reducing data redundancy. PCA uses either the covariance matrix or the correlation matrix to transfer data to an uncorrelated set. The eigenvectors of the resulting matrices are sorted in decreasing order where first principal component (PC) expresses most of the data variation. After the PCA transformation of the multi-band multitemporal data, in the new images the correlation coefficient between each component is zero or close to zero. And several components contain most of the information in remote sensing images.

The main advantage of PCA transformation is to reduce the correlation of each band, highlight the different objects and retain most of the feature and texture of SPOT images. The method does not limit the number of participating band. The disadvantage is that the determining of the number of components after transformation has no clear method which mainly relies on human experience.

3.4 Vegetation Index Differencing

Vegetation Index is the index that helps us understand the global vegetation distribution from the image data from Earth observation satellites. It calculates common ratios and index that reflect vegetation by different bands of remotely-sensed images. These indices enhance the spectral differences on the basis of strong vegetation absorbance in the red and strong reflectance in the near-infrared band. For change detection, generally, the vegetation indices are produced separately for two images and then standard pixel based change detection (e.g. differencing or ratioing) are applied. The advantage of this method is good results of ground vegetation change detection, but the disadvantage is that the applicability of ground changes of other types is not strong.

3.5 Post-Classification Comparison

Post-classification comparison method is a widely used method of change detection in remote sensing. The principle is to classify the images from different dates separately, and then in the area which has been classified from two images compare by pixel to determine the change information. This method was less affected by atmospheric radiation, sensor differences and image registration accuracy, but the choice of classification methods is difficult and classification of high precision is necessary. In order to improve change detection results, the classification of individual images has to be as accurate as possible.

4 THE APPLICATION OF CHANGE DETECTION TECHNIQUE

Change detection has played an important role in the research of land use/land cover, earthquake, and military, etc. We will describe the application in land use and earthquake in the following part.

Land use/land cover with its changes, is a critical factor in the study of the environment, forests, hydrology, agriculture, geography, ecology and other aspects. Land use/land cover changes as well as the relative changes they brought, has a direct impact on the human environment and ecological processes. For decades, scientists have developed several remote sensing land use change detection methods. These methods have their own advantages, as well as shortcomings.

Earthquake change detection using remotely-sensed images is to compare the same area in one remotely-sensed image before the earthquake and another remotely-sensed image after the earthquake which are both after the image registration, obtain to the earthquake damage data of earthquake area, and according to earthquake damage data give an analysis and assessment. Currently, the number of change detection techniques using remotely-sensed image is large. However, because of uncertainty of the earthquake-affected point, we cant determine which change detection method is more efficient and accurate for earthquake change detection.

5 CONCLUSSION

Remote sensing as an advanced science and technology has become one of the powerful means in the research of the earth resources and the environment. Using the change detection techniques, our country have achieved good results in land use, urban green space research, environmental monitoring and forest pest control area. But there are still a lot of difficulties in the study of change detection. For instance, image preprocessing cant meet all the requirements in the practical applications, and the effective algorithms and techniques of it affect the accuracy of the detection results. In addition, since the difference of change detection algorithms, the abilities of all the change detection algorithms are restricted by spatial, spectral, time domain and professional content. The method used, to some extent, affects the change detection accuracy. Therefore, when using remotely-sensed images to detect changes, selecting the appropriate method has a great significance in extraction and detection of change information.

[REFERENCES]

[1]Bruzzone L,Prieto D F.Automatic analysis of the difference image for unsupervised change detection[J].Geoscience and Remote Sensing,IEEE Transactions on,2000,38(3):1171-1182.

[2]Coppin P,Jonckheere I,Nackaerts K,et al.Review ArticleDigital change detection methods in ecosystem monitoring:a review[J].International journal of remote sensing,2004,25(9):1565-1596.

[3]Deren L,Haigang S,Ping X.Automatic change detection of geo-spatial data from imagery[J].Geo-Spatial Information Science,2003,6(3):1-7.

[4]Hussain M,Chen D,Cheng A,et al.Change detection from remotely sensed images:From pixel-based to object-based approaches[J].ISPRS Journal of Photogrammetry and Remote Sensing,2013,80:91-106.

[5]Singh A.Review Article Digital change detection techniques using remotely-sensed data[J].International journal of remote sensing,1989,10(6):989-1003.

[6]李德仁.利用遙感影像進(jìn)行變化檢測[J].武漢大學(xué)學(xué)報(信息科學(xué)版),2003,28(3):8312.

[7]馬建文,田國良,王長耀,等.遙感變化檢測技術(shù)發(fā)展綜述[J].地球科學(xué)進(jìn)展,2004,19(2):192-196.

[8]張曉朋.遙感圖像地震變化的檢測技術(shù)研究[D].廣西師范大學(xué), 2011.