In this paper, we consider the problem of moving target detection, and a novel radar system with multiple moving platforms is proposed. Each moving platform is equipped with multiple colocated antennas and serves as a transmitter or a receiver. Thus, this system possesses the advantages of both distributed and colocated multiple-input multiple-output radars. To exploit the clutter sparsity in the surveillance area, a novel compressed sensing (CS)-based model is proposed. Since the clutter cannot exactly reside on discretized grids often employed by most CS approaches, a novel two-step algorithm which extends the orthogonal matching pursuit is proposed to reconstruct the off-grid clutter. Then, a fusion center combines all received signals by using a generalized likelihood ratio test to detect the moving target. To further improve the detection performance, a novel online waveform optimization algorithm is developed to maximize the signal-to-clutter-and-noise ratios of each transmitter platform. Extensive simulation results are provided to demonstrate the effectiveness of the proposed radar system and algorithms.