Image target search is a key technology in the fields of intelligent security and regional surveillance. In recent years, with the explosive growth of video data, how to quickly search for objects of interest in massive amounts of video data has become a key issue to be solved in the current intelligent video processing field. A fast and intelligent search method based on the hierarchical cascade of bottom-level features and deep learning features is proposed. First, the intelligent target detection technology is used to intelligently detect the types of targets of interest in the video data. Second, for a large number of target detection results, a gating mechanism based on the bottom-level feature correlation measurement is designed to achieve the screening of target frames. Finally, high-dimensional deep learning features are used to complete the accurate recognition of the target. The experimental results show that the proposed method has a mean average precision (mAP) of 88.5% on the Market1501 dataset. At the same time, the calculation amount of the network reasoning process is reduced by 81.75% compared with YOLO and re-identification network direct cascade target search method, which significantly improves the target search speed.