Space-based detection algorithm in American SBV named moving target indicator(MTI) algorithm was improved. The improved algorithm could detect target streak ‘submerged’ in noise and clutter more quickly and more effectively. MTI improved algorithm used a new two-dimensional image detection preprocessing algorithm named windowed detection in probability algorithm. There were more the non-zero points in target detection window than in pure noise detection window with the same size. According to the characteristic, windowed detection in probability algorithm used threshold filter in detection window. The threshold filter extremely suppressed noise along with eliminating portion target points. Then, doubtful target streak was eliminated by three-point collinear streak detection algorithm. The streak detection algorithm reduced false-alarm probability and increased detection probability. By algorithm was performance analysis, false-alarm probability of MTI improved algorithm was reduced fifteen times than MTI algorithm, and needed signal-to-noise ratio (SNR) was reduced from 6 to 3 when detection probability equals to 99%. Furthermore, the calculation amount of MTI improved algorithm was reduced six order of magnitude than MTI algorithm. According to algorithm performance, MTI improved algorithm avails real-time processing in engineering application.