The signal-to-noise (SNR) of space debris laser ranging (DLR) data is too low to extract the effective echo with fast and effective methods. To solve this problem, a DLR echo extraction method based on randomized Hough transform (RHT) is proposed. Considering the time correlating property of O-C residuals′ adjacent echo, the effective echo extraction problem is transformed to the curve recognition of the image space. N pixels that determine a curve are randomly picked and they are mapped into one point in the parameter space. The parameter score accumulation strategy is used for curve recognition in the parameter space, which extracts the effective echo. The processing of DLR data of Yunnan Observatory shows that this method can extract echo signal on the approximated curve effectively with high speed, small storage and low misdetection rate. And it provides a new thought for fast and automatic DLR echo extraction.