To assess the accuracy of LULC classified images, a confusion matrix is created and overall accuracy, Kappa coefficient and user's and producer's accuracy are to be calculated. In a confusion matrix, which among user's and producer's accuracy is significant? Also, if we multiply user's and producer's accuracy (for eg. user's accuracy is 80% and the producer's accuracy is 78%, on multiplying 80% and 78% we get 62.4%), can we draw any discussion on the classified image?
Hello..
You have to state overall accuracy and kappa coefficient first, then
Confusion matrix, producer and user accuracy in a table form..
If the overall accuracy is more than 85% which is USGS standard limit.Then the LULC is acceptable. If your overall accuracy is more than 85% then you can also get good kappa coefficient value (expected).
Comming to producer and User accuracy
Producer accuracy is based on your classification point of view and user accuracy shows the reality on ground.
Higher user accuracy reflect the accuracy of your LULC..
Go through any standard paper of LULC for writing result the format...This format may varry journal to journal..
Check reference
Cohen J. 1960. A coefficient of agreement for nominal scales. Educ Psychol Meas. 20(1):37-46.
Manserud RA, Leemans R. 1992. Comparing global vegetation maps with the kappa statistics. Ecol Model. 62:275–279.
Lillesand T, Kiefer RW, Chipman J. 2014. Remote sensing and image interpretation. New Jersey (US): John Wiley & Sons.
Gounaridis D. Apostolou A. Koukoulas S. 2016. Land cover of Greece, 2010: a semi-automated classification using random forests. J Maps. 12(5):1055-1062.
Imam E. 2017. Habitat Suitability Modelling for Sambar (Rusa unicolor): A Remote Sensing and GIS Approach. In Environment and Earth Observation Cham: Springer; p. 231-246.
First of all, Kappa and overall accuracy is very important than user and producer accuracy
Thank you @ Souvik Barik. I have done this before. I just want clarity on if we can compare producer's and user's accuracy? The individual inference is understood. But, "can we make a common conclusion using both producer's and user's accuracy?" is my question. If yes, which one is to be given importance? user's or producer's accuracy?
Yes you can..If the accuracy values of producer and user are belonging to an acceptable range...Then only you can make any strong statement based upon these.
However much importance were given to user accuracy most of the time..As it reflects the ground reality...
Although generally overall accuracy and kappa coefficient value these two are mainly stated in the result and conclusion section.
Hope this will help..
Overall accuracy of the remote sensing product can be calculated by adding all the diagonal elements divided by the overall sample units used for accuracy in the error matrix. If overall accuracy is more than 85% then we can say that it is meaningful but it depends upon your application. User accuracy refers how actually classified map is real on the ground. For example your user accuracy is 80% means your classified item is 80% of mapped area in actually that items other may not referred to that item. Producer accuracy refer to the classification scheme. There must more users and producer accuracy. If your user, overall, producer accuracy are high then kappa coefficient will also high.
Remote Sensing and Image Interpretation by Lisllesand and Kiefer, Page no 572
- Badges
- Science topic
- Similar topics
- Correction