We investigate the design and performance of an integrated waveform-agile (WAS) multi-modal track-before-detect (TBD) sensing system using an experimental platform. The system jointly operates on radio frequency and electro-optical sensors, and it uses a particle filtering based recursive TBD algorithm for target tracking under low signal-to-noise environments. An adaptive waveform configuration scheme in integrated with the TBD in order to increase tracking performance and algorithm robustness. As simultaneous measurements statistically dependent, the exponential embedded family estimation method to approximate the unknown measurement noise distribution. We designed an experimental platform and used experimental data to evaluate the performance of the WAS-TBD system performance.