The multi-pulsed laser ranging technology is prominent on improving the maximum measuring range of laser active detection,laser range finder and other long-distance measurement. For all laser echo detection techniques, the weak signal detection is an important step, which aims to increase the detection range. Most algorithms are based on the priori knowledge of laser echo or the improvement of laser power. However, we cannot know or estimate the waveform accurately in many applications. Moreover, these means are difficult to satisfy the real-time needs. The present paper proposes an improved algorithm which extended the signal accumulation algorithm for the high power burst laser. This method is mainly based on signal accumulation and tri-cumulant algorithm which can improve the signal to noise SNR of the weak laser echo; moreover it does not need more prior knowledge of echo. In order to reduce the detection time, the algorithm is realized based on FPGA using signal retiming and parallel pipeline structure. The simulations and experiments results demonstrate that the minimum detecting SNR is -5dB and the maximum detecting time is only less than 1us.