We have developed the following general method for the semi-automatic detection of circular archaeological objects:
- Pre-process the remote sensing data to a form suitable for further processing
- Convolve the image with templates of a range of sizes, corresponding to the range of expected sizes of the archaeological objects
- Threshold the convolution result to find an initial (large) set of candidate detections.
- (Optional) Remove detections outside masks.
- Merge detections that are very close
- Compute various measures of the deviation from an ideal model of the cultural heritage object
- Remove detections that deviate too much from the ideal model
- Compute a confidence level (1-6) for each detection
- Visual inspection by archaeologist
- If necessary: field inspection by archaeologist