With the rapid development of computer vision, pedestrian detection technology has become one of the hot topics. SSD algorithm in pedestrian detection algorithm is more stable than other pedestrian detection algorithms. Aiming at the problem of low detection accuracy when SSD algorithm is used in pedestrian detection, this paper proposes a pedestrian detection algorithm which uses RESNET residual network as the feed-forward network of SSD to extract pedestrian features and generate the target box through the eigenvalues. The original SSD algorithm and the improved SSD algorithm are compared through the training set and the test set of INRIA pedestrian detection data set. The experimental results show that, compared with SSD algorithm, the improved pedestrian detection algorithm is more robust, and the average accuracy of the algorithm is improved by nearly 5%.