@Dear Jennifer

As Ahmed Ghodieh very correctly pointed out you need to collect enough ground truth sample data for each land cover classes create a matrix of ground truth data and the classified image data. Of course, it provides an obvious foundation for accuracy assessment. Then percentage of cases (pixels) correctly allocated may be used an easily interpretable guide to the overall accuracy of the classification. In the same way percentages of correctly classified pixels in each of land cover classes may indicate level of accuracy in demarcating each class individually. The main problem in this accuracy test is that some cases may have been allocated to the correct class purely by chance. Cohen’s kappa coefficient has often been used and is generally adopted as a standard measure of classification accuracy.

Taking clue from an earlier post of yours, below is explained how to perform accuracy assessment in both ERDAS and ENVI. You may yourself learn it from the tutorials/guides etc. of the two Software or manually by downloading a good article on confusion matrix and kappa coefficient.

Accuracy Assessment in ERDAS:

For this exercise you will need an "un-classified" subset image and a (supervised) classified image of the same subset.

1. Open a viewer and display the "un-classified" subsets – display an "un-classified" subset for which you have also applied your signature file and have therefore a corresponding classified image.

2. Click the Classifier icon from the icon panel.

3. Select ACCURACY ASSESSMENT, once the accuracy assessment viewer opens you can close the classifier menu – it won’t be need anymore.

4. From the accuracy assessment viewer, select FILE | OPEN (or click the Open icon). In the resulting dialog navigate to the location of the classified image. This will be the image file that will be used in the accuracy assessment. Click OK to load the file.

5. From the accuracy assessment viewer, select VIEW | SELECT VIEWER (or click on the icon).

6. Now you’ll select the colours which will help visualized the "reference" pixels. In the accuracy assessment viewer, select VIEW | CHANGE COLORS. Set the colour for "Points With No Reference:" to WHITE (these are the random points that HAVE NOT been assigned a reference value), and "Points With Reference:" to Yellow (these are the random point these HAVE been assigned a reference class value). Click OK in the change colour dialog or accept the default, as long as you know which is which.

The utility ADD RANDOM POINTS will generate random points throughout the classified image. After the points are generated, you must enter the class values for the points, which will be the reference points. These reference values will be compared to the class values of the classified image.

1. In the accuracy assessment viewer, select EDIT | CREATE/ADD RANDOMM POINTS. The Add Random points dialog opens

2. Confirm that the Search Count: is set to 1024. This means that a maximum of 1024 points will be analyzed to see if they meet the defined requirements in the "Add Random Points" dialog. You can increase this value and insure that your requested number of points is selected if you have a problem.

3. Enter a value of 10 in the "NUMBER OF POINTS" field. This means that you will generate ten random points. However, to perform a proper accuracy assessment you may need 250 or more.

4. Note that the distribution parameters can be set to either "random", "stratified random", or "equalized random". Select the distribution parameters to RANDOM. Click OK to generate the points. When completed a list of points is shown in the accuracy assessment CellArray.

5. In the Accuracy Assessment viewer, select VIEW | SHOW ALL. All of the random points will be displayed as WHITE in the viewer that you have selected earlier.

6. Evaluate the location of these points and determine their CLASS VALUE (properly, this would be done with ground truth, maps, aerial photos, or other data. In the Accuracy Assessment CellArray REFERENCE column, you'll enter your best guess of a class VALUE for the pixel covered by each reference point. (As you do this the colour of the point in the viewer will change to yellow). But you will most likely need to refer to the signature file to find the numeric value (not the name) you assigned to each class. Click on the Classifier icon from the panel, select signature editor, from the editor use open to load the signature file and refer to the VALUE column to find the assigned numeric code.

The Error Matrix, Accuracy Totals, and Kappa Statistics can be used to understand accuracy of the classification. This information is recorded in a report form.

1. In the Accuracy Assessment viewer, select EDIT | SHOW CLASS VALUE. The class values for the reference points appear in the CLASS column of the CellArray.

2. In the Accuracy Assessment viewer, select REPORT | OPTIONS. The Error Matrix, Accuracy Totals, and Kappa Statistic check boxes should be turned on.

3. Reports are created and displayed in the IMAGINE text editor by selecting both REPORT | ACCURACY REPORT, and REPORT | CELL REPORT in the Accuracy assessment viewer. If you wish you can save these reports to a text file.

4. Close the text editors, and select FILE | SAVE TABLE in the Accuracy Assessment viewer to save the data in the CellArray, CLOSE the Accuracy Assessment viewer when done.

It isn't that difficult to understand most of the report. The error matrix simply compares the reference points to the classified points. The Kappa coefficient expresses the proportionate reduction in error generated by a classification process compared with the error of a completely random classification. For example, a value of .82 would imply that the classification process was avoiding 82% of the errors that a completely random classification would generate.

Accuracy Assessment in ENVI:

ENVI provides analysts with a Confusion Matrix for assessing the accuracy of the information that they have generated. The method is located in the Classification -> Post-Classification menu and it allows the analyst to use either a Ground Truth Image or Ground Truth Regions of Interest (ROI) to perform the assessment.

1. From the ENVI main menu bar, select Classification -> Post-Classification -> Confusion Matrix. Two options appear, one of which is Using Ground Truth ROIs. Select this option and a dialog box will appear.

2. Select your Classified Image output from the Select Input File list and click OK.

3. The Match Classes Parameters dialog box will open. Under Select Ground Truth ROI select Land Cover Class A. Under Select Classification Image select Land Cover Class A. Select the Add Combination button.

4. Repeat step 3 for all of the classes. Once all of the classes have been matched, select OK.

[Note: the step above may be automatically completed for you by ENVI depending on the ROI names and content.]

5. The Confusion Matrix Parameters dialog box opens. Accept the defaults and select OK.

6. The Class Confusion Matrix will appear. Your values will be different depending on how you defined your classes using the ROI tool.

The Report is easy to understand and interpret.

you need to collect enough ground truth sample data for each landcover class, decide their coordinates using GPS. create a matrix of ground truth data and the classified image data.

@Dear Jennifer

As Ahmed Ghodieh very correctly pointed out you need to collect enough ground truth sample data for each land cover classes create a matrix of ground truth data and the classified image data. Of course, it provides an obvious foundation for accuracy assessment. Then percentage of cases (pixels) correctly allocated may be used an easily interpretable guide to the overall accuracy of the classification. In the same way percentages of correctly classified pixels in each of land cover classes may indicate level of accuracy in demarcating each class individually. The main problem in this accuracy test is that some cases may have been allocated to the correct class purely by chance. Cohen’s kappa coefficient has often been used and is generally adopted as a standard measure of classification accuracy.

Taking clue from an earlier post of yours, below is explained how to perform accuracy assessment in both ERDAS and ENVI. You may yourself learn it from the tutorials/guides etc. of the two Software or manually by downloading a good article on confusion matrix and kappa coefficient.

Accuracy Assessment in ERDAS:

For this exercise you will need an "un-classified" subset image and a (supervised) classified image of the same subset.

1. Open a viewer and display the "un-classified" subsets – display an "un-classified" subset for which you have also applied your signature file and have therefore a corresponding classified image.

2. Click the Classifier icon from the icon panel.

3. Select ACCURACY ASSESSMENT, once the accuracy assessment viewer opens you can close the classifier menu – it won’t be need anymore.

4. From the accuracy assessment viewer, select FILE | OPEN (or click the Open icon). In the resulting dialog navigate to the location of the classified image. This will be the image file that will be used in the accuracy assessment. Click OK to load the file.

5. From the accuracy assessment viewer, select VIEW | SELECT VIEWER (or click on the icon).

6. Now you’ll select the colours which will help visualized the "reference" pixels. In the accuracy assessment viewer, select VIEW | CHANGE COLORS. Set the colour for "Points With No Reference:" to WHITE (these are the random points that HAVE NOT been assigned a reference value), and "Points With Reference:" to Yellow (these are the random point these HAVE been assigned a reference class value). Click OK in the change colour dialog or accept the default, as long as you know which is which.

The utility ADD RANDOM POINTS will generate random points throughout the classified image. After the points are generated, you must enter the class values for the points, which will be the reference points. These reference values will be compared to the class values of the classified image.

1. In the accuracy assessment viewer, select EDIT | CREATE/ADD RANDOMM POINTS. The Add Random points dialog opens

2. Confirm that the Search Count: is set to 1024. This means that a maximum of 1024 points will be analyzed to see if they meet the defined requirements in the "Add Random Points" dialog. You can increase this value and insure that your requested number of points is selected if you have a problem.

3. Enter a value of 10 in the "NUMBER OF POINTS" field. This means that you will generate ten random points. However, to perform a proper accuracy assessment you may need 250 or more.

4. Note that the distribution parameters can be set to either "random", "stratified random", or "equalized random". Select the distribution parameters to RANDOM. Click OK to generate the points. When completed a list of points is shown in the accuracy assessment CellArray.

5. In the Accuracy Assessment viewer, select VIEW | SHOW ALL. All of the random points will be displayed as WHITE in the viewer that you have selected earlier.

6. Evaluate the location of these points and determine their CLASS VALUE (properly, this would be done with ground truth, maps, aerial photos, or other data. In the Accuracy Assessment CellArray REFERENCE column, you'll enter your best guess of a class VALUE for the pixel covered by each reference point. (As you do this the colour of the point in the viewer will change to yellow). But you will most likely need to refer to the signature file to find the numeric value (not the name) you assigned to each class. Click on the Classifier icon from the panel, select signature editor, from the editor use open to load the signature file and refer to the VALUE column to find the assigned numeric code.

The Error Matrix, Accuracy Totals, and Kappa Statistics can be used to understand accuracy of the classification. This information is recorded in a report form.

1. In the Accuracy Assessment viewer, select EDIT | SHOW CLASS VALUE. The class values for the reference points appear in the CLASS column of the CellArray.

2. In the Accuracy Assessment viewer, select REPORT | OPTIONS. The Error Matrix, Accuracy Totals, and Kappa Statistic check boxes should be turned on.

3. Reports are created and displayed in the IMAGINE text editor by selecting both REPORT | ACCURACY REPORT, and REPORT | CELL REPORT in the Accuracy assessment viewer. If you wish you can save these reports to a text file.

4. Close the text editors, and select FILE | SAVE TABLE in the Accuracy Assessment viewer to save the data in the CellArray, CLOSE the Accuracy Assessment viewer when done.

It isn't that difficult to understand most of the report. The error matrix simply compares the reference points to the classified points. The Kappa coefficient expresses the proportionate reduction in error generated by a classification process compared with the error of a completely random classification. For example, a value of .82 would imply that the classification process was avoiding 82% of the errors that a completely random classification would generate.

Accuracy Assessment in ENVI:

ENVI provides analysts with a Confusion Matrix for assessing the accuracy of the information that they have generated. The method is located in the Classification -> Post-Classification menu and it allows the analyst to use either a Ground Truth Image or Ground Truth Regions of Interest (ROI) to perform the assessment.

1. From the ENVI main menu bar, select Classification -> Post-Classification -> Confusion Matrix. Two options appear, one of which is Using Ground Truth ROIs. Select this option and a dialog box will appear.

2. Select your Classified Image output from the Select Input File list and click OK.

3. The Match Classes Parameters dialog box will open. Under Select Ground Truth ROI select Land Cover Class A. Under Select Classification Image select Land Cover Class A. Select the Add Combination button.

4. Repeat step 3 for all of the classes. Once all of the classes have been matched, select OK.

[Note: the step above may be automatically completed for you by ENVI depending on the ROI names and content.]

5. The Confusion Matrix Parameters dialog box opens. Accept the defaults and select OK.

6. The Class Confusion Matrix will appear. Your values will be different depending on how you defined your classes using the ROI tool.

The Report is easy to understand and interpret.

Russell G. Congalton and Kass Green, Assessing the Accuracy of Remotely Sensed Data—Principles and Practices (Second edition), CRC Press, Taylor & Francis Group, Boca Raton, FL (2009) ISBN 978-1-4200-5512-2 183 pp.

Dear friends i am doing my research in satellite image processing indicate urban area classification for that which classification method is better

your question is not clear enough. Please provide more details on what you are doing.

Dr Khan, I have GPS points of selected land uses from the field. Can I use it in the Accuracy Assessment?

Divine Appiah, of coz you can

Dear all, I have field data of vegetation cover and soil's mechanical Composition and Analysis of soil chemical data. So far i computed desertification map by remote sensing data. How i accuracy assessment between desertification map and field data? How i checked my result? Please help me. Which kind of program best in this process erdas or envi? Please send me some lesson. Actually i don't know how do this process.