A novel approach to realize broadband microwave spectrum sensing based on photonic RF channelization and compressive sampling (CS) is proposed. The photonic RF channelization system is used to slice the input broadband signal into multiple sub-channel signals with narrow bandwidth in parallel and thus the rate of pseudo-random binary sequence (PRBS) and the bandwidth of the MZM for CS can be largely decreased. It is shown that a spectrally sparse signal within a wide bandwidth can be captured with a sampling rate far lower than the Nyquist rate thanks to both photonic RF channelization and CS. In addition, the influence of the non-ideal filtering of the photonic channelizer is evaluated and a novel approach based on measuring twice is proposed to overcome the problem of frequency aliasing induced by the non-ideal filtering. It is demonstrated that a system with 20 Gbit/s PRBS and 2.5 GS/s digitizer can be used to capture a signal with multiple tones within a 40 GHz bandwidth, which means a sampling rate 32 times lower than the Nyquist rate.