Is it necessary to do an accuracy assessment for an unsupervised classification done for a Landsat 5 image?
When an accuracy assessment is being performed an error matrix would be generated and kappa. From there you would notice that omission and commision error is there from which the producer and user accuracy can be calculated from the formula given below:
Producer Accuracy = 100 - omission error and User Accuracy = 100 - commission error
From the calculated values above if the user accuracy is between 70-100% then you are good to go otherwise you have to reclassify for good results.