Compressed sensing is based on the principle of signal sparsity,which achieves that high-order matrix signal is compressed to low-level compression matrix.The paper discusseed the compressed sensing theory,and compared it with traditional shannon sampling.Compression perception had a low sampling rate,compression and sampling were the same process,etc.This paper also discussed key technologies of the signal sparse representation,building measurement matrix,signal reconstruction,related research achievements and reviewed problems in detail.Although the theory had achieved good effect in optical imaging,radar detection and speech coding areas,had good prospects for development in the three-dimensional medical imaging,ultrasound image detection,the theory remaind to further research and validation to meet the needs of the various practical application.