Excessive noise point cloud in photon-counting lidar data severely restricts the practical application of photon radar. To solve this problem, this paper proposes a denoising method for the detection system of push-broom photon radar and the characteristics of photon point cloud data distribution. First, we set the density threshold based on the spatial distribution of the point cloud dataset for rough denoising. Then, we calculated the slope of the detection laser beam to classify the remaining data into different intervals. Next, we combined the maximum density point to set the distance threshold for further denoising each interval. Finally, we used statistical filtering for the final denoising of some intervals. The experimental results show that the target point cloud recognition rate of the proposed method reaches 98.2% in the test area and the denoising rate reaches 93.8%. Thus, the proposed method can remove the noise point cloud in the photon data effectively and retain the signal point cloud relatively completely.