In order to realize the early warning of forest fires, a laser radar system with high double frequency polarization is designed and built based on the light scattering and polarization effect of smoke particles generated in the early stage of fire. In order to meet the real-time and high-resolution requirements of forest fire detection, the high repetition frequency laser is used as the detection light source, and the high Angle resolution PTZ is used to achieve fine scanning detection. In order to eliminate the background noise during detection, time integration is required when the echo signal is collected. However, when the repetition rate of LiDAR rises to the Khz level, there is no mature method to be used. Aiming at this technical bottleneck, a novel and special "reading-cumulation-storage" architecture is proposed, which effectively realizes the acquisition and time integration of echo signals. The repetition rate of the built system can reach 5KHz, the Angle resolution is 1 degree, and the forest with a radius of 10 km can be detected by scanning within 45 min. In the simulation experiment, the system can quickly and accurately identify the fire point smoke at 600 m and 4 km, which verifies that the system can effectively detect forest fires in the early stage.