6_Object_Oriented_Classification

6_Object_Oriented_Classification - Single-pixel...

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Single-pixel Classification versus Single-pixel Classification versus Object-oriented Image Segmentation Object-oriented Image Segmentation Classification algorithms based on single-pixel analysis often are not Classification algorithms based on single-pixel analysis often are not capable of extracting the information we desire from high-spatialcapable of extracting the information we desire from high-spatialresolution remote sensor data (e.g., QuickBird 61 61 cm). For resolution remote sensor data (e.g., QuickBird 61 61 cm). For example, the spectral complexity of urban land-cover materials results example, the spectral complexity of urban land-cover materials results in specific limitations using per-pixel analysis for the separation of in specific limitations using per-pixel analysis for the separation of human-made materials such as roads and roofs and natural materials human-made materials such as roads and roofs and natural materials such as vegetation, soil, and water. Furthermore, aasignificant but such as vegetation, soil, and water. Furthermore, significant but usually ignored problem with per-pixel characterization of land cover is usually ignored problem with per-pixel characterization of land cover is that aasubstantial proportion of the signal apparently coming from the that substantial proportion of the signal apparently coming from the land area represented by aapixel comes from the surrounding terrain. land area represented by pixel comes from the surrounding terrain. Improved algorithms are needed that take into account not only the Improved algorithms are needed that take into account not only the spectral characteristics of aasingle pixel but those of the surrounding spectral characteristics of single pixel but those of the surrounding (contextual) pixels. We need information about the spatial (contextual) pixels. We need information about the spatial characteristics of the surrounding pixels so that we can identify areas characteristics of the surrounding pixels so that we can identify areas (or segments) of pixels that are homogeneous. (or segments) of pixels that are homogeneous. Object-oriented Image Segmentation Object-oriented Image Segmentation This need has given rise to the creation of image classification This need has given rise to the creation of image classification algorithms based on object-oriented image segmentation. The algorithms based on object-oriented image segmentation. The algorithms incorporate both spectral and spatial information in the algorithms incorporate both spectral and spatial information in the image segmentation phase. The result is the creation of image objects image segmentation phase. The result is the creation of image objects defined as individual areas with shape and spectral homogeneity which defined as individual areas with shape and spectral homogeneity which one may recognize as segments or patches in the landscape. In many one may recognize as segments or patches in the landscape. In many instances, carefully extracted image objects can provide aagreater instances, carefully extracted image objects can provide greater number of meaningful features for image classification. In addition, number of meaningful features for image classification. In addition, objects don't have to be derived from just image data but can also be objects don't have to be derived from just image data but can also be developed from any spatially distributed variable (e.g., elevation, slope, developed from any spatially distributed variable (e.g., elevation, slope, aspect, population density). Homogeneous image objects are then aspect, population density). Homogeneous image objects are then analyzed using traditional classification algorithms (e.g., nearestanalyzed using traditional classification algorithms (e.g., nearestneighbor, minimum distance, maximum likelihood) or knowledge-based neighbor, minimum distance, maximum likelihood) or knowledge-based approaches and fuzzy classification logic. approaches and fuzzy classification logic. Object-oriented Image Segmentation Object-oriented Image Segmentation There are many algorithms that can be used to segment an image into There are many algorithms that can be used to segment an image into relatively homogeneous image objects. Most can be grouped into two relatively homogeneous image objects. Most can be grouped into two classes: classes: edge-based algorithms, and edge-based algorithms, and area-based algorithms. area-based algorithms. Unfortunately, the majority do not incorporate both spectral and spatial Unfortunately, the majority do not incorporate both spectral and spatial information, and very few have been used for remote sensing digital information, and very few have been used for remote sensing digital image classification. image classification. Object-oriented Image Segmentation Object-oriented Image Segmentation One of the most promising approaches to remote sensing image One of the most promising approaches to remote sensing image segmentation was developed by Baatz and Schape (2000). The image segmentation was developed by Baatz and Schape (2000). The image segmentation involves looking at individual pixel values and their segmentation involves looking at individual pixel values and their neighbors to compute aa(Baatz et. al., 2001): neighbors to compute (Baatz et. al., 2001): color criterion ((h ), and color criterion hcolor ), and color aashape or spatial criterion ((h ). shape or spatial criterion hshape ). shape Object-oriented Image Segmentation Object-oriented Image Segmentation These two criteria are then used to create image objects (patches) of These two criteria are then used to create image objects (patches) of relatively homogeneous pixels in the remote sensing dataset using relatively homogeneous pixels in the remote sensing dataset using the general segmentation function ((S): the general segmentation function Sf f): S f wcolor hcolor 1 wcolor hshape where the user-defined weight for spectral color versus shape is where the user-defined weight for spectral color versus shape is 00< wcolor < 1. If the user wants to place greater emphasis on the < wcolor < 1. If the user wants to place greater emphasis on the spectral (color) characteristics in the creation of homogeneous spectral (color) characteristics in the creation of homogeneous objects (patches) in the dataset, then wcolor is weighted more heavily objects (patches) in the dataset, then wcolor is weighted more heavily (e.g., wcolor = 0.8). Conversely, if the spatial characteristics of the (e.g., wcolor = 0.8). Conversely, if the spatial characteristics of the dataset are believed to be more important in the creation of the dataset are believed to be more important in the creation of the homogeneous patches, then shape should be weighted more homogeneous patches, then shape should be weighted more heavily. heavily. Object-oriented Image Segmentation Object-oriented Image Segmentation Spectral (i.e., color) heterogeneity ((h)of an image object is Spectral (i.e., color) heterogeneity h) of an image object is computed as the sum of the standard deviations of spectral values computed as the sum of the standard deviations of spectral values of each layer (())(i.e., band) multiplied by the weights for each of each layer kk (i.e., band) multiplied by the weights for each layer ((w): layer wkk): h wk k k 1 m Object-oriented Image Segmentation Object-oriented Image Segmentation The color criterion is computed as the weighted mean of all The color criterion is computed as the weighted mean of all changes in standard deviation for each channel kkof the m band changes in standard deviation for each channel of the m band remote sensing dataset. The standard deviation k are weighted by remote sensing dataset. The standard deviation k are weighted by the object sizes nn (Definiens, 2003): the object sizes ob (Definiens, 2003): ob h wk nmg k k 1 m mg nob1 k ob1 nob 2 k ob 2 where mg means merge. where mg means merge. Object-oriented Image Segmentation Object-oriented Image Segmentation The shape criterion is computed using two landscape ecology The shape criterion is computed using two landscape ecology metrics: compactness and smoothness. Heterogeneity as deviation metrics: compactness and smoothness. Heterogeneity as deviation from aacompact shape ((cpt)is described by the ratio of the pixel from compact shape cpt) is described by the ratio of the pixel perimeter length lland the square root of the number of pixels nn perimeter length and the square root of the number of pixels forming an image object (i.e., aapatch): forming an image object (i.e., patch): 1 cpt n Object-oriented Image Segmentation Object-oriented Image Segmentation Shape heterogeneity may also be described as smoothness, which is Shape heterogeneity may also be described as smoothness, which is the ratio of the pixel perimeter length ll and the shortest possible the ratio of the pixel perimeter length and the shortest possible border length bb of aa box bounding the image object (i.e., aa patch) border length of box bounding the image object (i.e., patch) parallel to the raster: parallel to the raster: l smooth b Object-oriented Image Segmentation Object-oriented Image Segmentation The shape criterion incorporates these two measurements using the The shape criterion incorporates these two measurements using the equation (Definiens, 2003): equation (Definiens, 2003): hshape wcpt hcpt 1 wcpt hsmooth where 0 < wcpt <1 is the user-defined weight for the compactness where 0 < wcpt <1 is the user-defined weight for the compactness criterion. criterion. Object-oriented Image Segmentation Object-oriented Image Segmentation The change in shape heterogeneity caused by each merge is The change in shape heterogeneity caused by each merge is evaluated by calculating the difference between the situation after evaluated by calculating the difference between the situation after and before image objects ((ob)are merged. This results in the and before image objects ob) are merged. This results in the following algorithms for computing roughness and smoothness following algorithms for computing roughness and smoothness (Definiens, 2003): (Definiens, 2003): hcpt nmg lmg nmg nob1 lob1 nob 2 lob 2 nob1 nob 2 hsmooth nmg lmg bmg lob 2 lob1 nob1 nob 2 bob 2 bob1 Where nnis the object size in pixels. Where is the object size in pixels. Object-oriented Image Segmentation Object-oriented Image Segmentation One criteria used to segment aaremotely sensed One criteria used to segment remotely sensed image into image objects is aapixel neighborhood image into image objects is pixel neighborhood function, which compares an image object being function, which compares an image object being grown with adjacent pixels. The information is grown with adjacent pixels. The information is used to determine if the adjacent pixel should be used to determine if the adjacent pixel should be merged with the existing image object or be part merged with the existing image object or be part of aanew image object. a) If aaplane 44neighborof new image object. a) If plane neighborhood function is selected, then two image objects hood function is selected, then two image objects would be created because the pixels under would be created because the pixels under investigation are not connected at their plane investigation are not connected at their plane borders. b) Pixels and objects are defined as borders. b) Pixels and objects are defined as neighbors in aadiagonal 88neighborhood if they neighbors in diagonal neighborhood if they are connected at aaplane border or aacorner point. are connected at plane border or corner point. In this example, image object 11can be expanded In this example, image object can be expanded because it connects at aadiagonal corner point. because it connects at diagonal corner point. This results in aalarger image object 1. Other This results in larger image object 1. Other types of neighborhood functions could be used. types of neighborhood functions could be used. Classification based Classification based on on Image Image Segmentation Logic Segmentation Logic takes into account takes into account spatial and spectral spatial and spectral characteristics characteristics Object-oriented Image Segmentation Object-oriented Image Segmentation The object-oriented classification of aasegmented image is The object-oriented classification of segmented image is substantially different from performing aaper-pixel classification. substantially different from performing per-pixel classification. First, the analyst is not constrained to using just spectral First, the analyst is not constrained to using just spectral information. He or she may choose to use a) the mean spectral information. He or she may choose to use a) the mean spectral information in conjunction with b) various shape measures information in conjunction with b) various shape measures associated with each image object (polygon) in the dataset. This associated with each image object (polygon) in the dataset. This introduces flexibility and robustness. Once selected, the spectral introduces flexibility and robustness. Once selected, the spectral and spatial attributes of each polygon can be input to aavariety of and spatial attributes of each polygon can be input to variety of classification algorithms for analysis (e.g., nearest-neighbor, classification algorithms for analysis (e.g., nearest-neighbor, minimum distance, maximum likelihood). minimum distance, maximum likelihood). Classification based Classification based on on Image Image Segmentation Logic Segmentation Logic takes into account takes into account spatial and spectral spatial and spectral characteristics characteristics ...
View Full Document

This note was uploaded on 09/08/2010 for the course GEOG 182 at San Jose State University .

Ask a homework question - tutors are online