It would be helpful if you could send us sample fingerprints. This could lead to long-term improvements in our algorithm which would improve reliability of your application. Short-term, there are a couple of things you can try:
1. Try to artificially stretch the fingerprints. If you have 500dpi sensor, set DPI setting in IDKit (IDKit.Dpi in .NET or CFG_RESOLUTION_DPI in C++) to 300dpi before template extraction. This will enlarge the fingerprints by 66%. Children's fingerprints tend to have denser ridges. Be careful not to overstretch fingerprints of adults, because IDKit is tuned to standard adult ridge frequency.
2. Try a better sensor. If you are using sub-500dpi sensor then definitely replace it with a 500dpi one. If you are already using 500dpi sensor, try 1000dpi. This might help considerably if ridges are hard to discern in fingerprints your current sensor captures.
3. Try to split the two populations (children and adults) during registration and identification. Adjust threshold and DPI separately for each population. Our API allows identification search to be performed on database subsets rather than whole database.