Lecture 7 Change Detection PDF

CHANGE DETECTION Digital Image Analysis I  INTRODUCTION  In general, change detection involves the application of multitemporal datasets to quantitatively analyze the temporal effects 6/23/2023  Change detection can be defined as...

CHANGE DETECTION Digital Image Analysis I  INTRODUCTION  In general, change detection involves the application of multitemporal datasets to quantitatively analyze the temporal effects 6/23/2023  Change detection can be defined as the process of identifying differences in the state of an object or EGS 2311 DIGITAL IMAGE ANALYSIS I phenomenon by observing it at different times. This process is usually applied to Earth surface changes at two or more times.  Understanding relationships and interactions to better manage and use resources  Change detection is useful in many applications such as: APPLICATIONS OF CHANGE DETECTION TECHNIQUES  Land-use and land-cover (LULC) change  Forest or vegetation change 6/23/2023  Forest mortality, defoliation and damage assessment  Deforestation, regeneration and selective logging EGS 2311 DIGITAL IMAGE ANALYSIS I  Wetland change  Forest fire and fire-affected area detection  Landscape change  Urban change  Environmental change, drought monitoring, flood monitoring, monitoring coastal marine environments, desertification, and detection of landslide areas  Other applications such as crop monitoring, shifting cultivation monitoring, road segments, and change in glacier mass balance and facies. CHANGE DETECTION  Two main categories of land cover changes:  Conversion of land cover from 6/23/2023 one category to a different category. EGS 2311 DIGITAL IMAGE ANALYSIS I  Modification of the condition of the land cover type within the same category (thinning of trees, selective cutting, pasture to cultivation, etc.) CONSIDERATIONS BEFORE IMPLEMENTING CHANGE DETECTION  Before implementing change detection analysis, the following conditions must be satisfied: 6/23/2023  Precise registration of multi-temporal images;  Precise radiometric and atmospheric calibration or EGS 2311 DIGITAL IMAGE ANALYSIS I normalization between multi-temporal images;  Selection of the same spatial and spectral resolution images if possible GOOD CHANGE DETECTION RESEARCH SHOULD PROVIDE THE FOLLOWING INFORMATION:  Area change and change rate  Spatial distribution of changed types 6/23/2023  Change trajectories of land-cover types  Accuracy assessment of change detection results. EGS 2311 DIGITAL IMAGE ANALYSIS I CHANGE DETECTION TECHNIQUES 1. Algebra based approach 6/23/2023  Image differencing,  Image regression EGS 2311 DIGITAL IMAGE ANALYSIS I  Image ratioing  Vegetation index differencing  Change vector analysis (CVA) IMAGE DIFFERENCING  Concept  Date 1 - Date 2  No-change = 0 6/23/2023  Positive and negative values interpretable  Pick a threshold for change EGS 2311 DIGITAL IMAGE ANALYSIS I IMAGE DIFFERENCING  Image differencing: Pros  Simple (some say it’s the most commonly used method) 6/23/2023  Easy to interpret  Robust EGS 2311 DIGITAL IMAGE ANALYSIS I  Cons:  Difference value is absolute, so same value may have different meaning  Requires atmospheric calibration IMAGE REGRESSION  Relationship between pixel values of two dates is established by using a regression function.  The dimension of the residuals is an indicator of where 6/23/2023 change occurred.  Advantage EGS 2311 DIGITAL IMAGE ANALYSIS I  Reduces impact of atmospheric, sensor and environmental differences.  ► Drawback  Requires development of accurate regression functions.  Does not provide change matrix. 6/23/2023 EGS 2311 DIGITAL IMAGE ANALYSIS I IMAGE RATIOING  Concept  Date 1 / Date 2  No-change = 1 6/23/2023  Values less than and greater than 1 are interpretable  Pick a threshold for change Pros EGS 2311 DIGITAL IMAGE ANALYSIS I   Simple  May mitigate problems with viewing conditions, esp. sun angle  Cons  Scales change according to a single date, so same change on the ground may have different score depending on direction of change; I.e. 50/100 =.5, 100/50 = 2.0 CHANGE VECTOR ANALYSIS  In n-dimensional spectral space, determine length and 6/23/2023 direction of vector between Date 1 and Band 4 EGS 2311 DIGITAL IMAGE ANALYSIS I Date2  No-change = 0 length  Change direction may be interpretable  Pick a threshold for Band 3 change CHANGE VECTOR ANALYSIS  Determines in n-dimensional spectral space, the length and direction of the vector between Date 1 and Date 2. 6/23/2023  Produces an intensity image and a direction image of change. The direction image EGS 2311 DIGITAL IMAGE ANALYSIS I can be used to classify change.  Typically used when all changes need to be investigated.  Advantage  Works on multispectral data.  Allows designation of the type of change occurring  ►Drawback  Shares some of the drawbacks of algebra based techniques but less severe  6/23/2023 EGS 2311 DIGITAL IMAGE ANALYSIS I CHANGE VECTOR ANALYSIS 2. Classification based approach  Post classification comparison  Spectral temporal combined analysis 6/23/2023 EGS 2311 DIGITAL IMAGE ANALYSIS I POST CLASSIFICATION COMPARISON  Separately classifies multitemporal images into thematic maps, then implements comparison of the classified images, pixel by pixel 6/23/2023  Pros  Minimizes impacts of atmospheric, sensor and environmental differences between multitemporal images; EGS 2311 DIGITAL IMAGE ANALYSIS I  Provides a complete matrix of change information  Cons  Requires a great amount of time and expertise to create classification products.  The final accuracy depends on the quality of the classified image of each date  Examples  LULC change,  Wetland change and urban expansion  Key factors  Selects sufficient training sample data for classification SPECTRAL– TEMPORAL COMBINED ANALYSIS  Puts multi-temporal data into a single file, then classifies the combined dataset and identifies and labels the changes 6/23/2023  Pros  Simple and timesaving in classification EGS 2311 DIGITAL IMAGE ANALYSIS I  Cons  Difficult to identify and label the change classes;  Cannot provide a complete matrix of change information  Examples  Changes in coastal zone environments and forest change  Key factors  Labels the change classes