I've been looking for a lot of detection-based tracking. Most of the current multi-target tracking is detection-based tracking. Is there any non-detection-based tracking?
There are a number of track-before-detect technologies. The best known is multiple hypothesis tracking (MHT). Others include warping the range-Doppler map between scans to build up signal strength in all range-Doppler map cells, and there are others. Do a web search on "track before detect".
In general, there are track before detect (TBD) algorithms that are deployed in certain cases like low SNR and dim targets present in the given surveillance etc. The H-PMHT, Particle Filter (PF), and ML-PDA are some of the track filters that are used with all the un thresholded measurements with out detection. You can find good resources by searching these key words.