Aiming at the problems of low signal-to-noise ratio and blurred imaging, a method of terahertz concealed object detection based on multi-feature fusion is proposed. First, the terahertz image preprocessing is designed to realize the noise filtering and image enhancement. After that, by extracting the oriented gradient distribution feature, the grayscale distribution feature, and the pixel spatial distribution feature, then fusing them into a multi-feature vector, the comprehensive representation of the concealed object is accomplished. Finally, the support-vector machine is employed to achieve the classification task, and the location task is completed by calculating the bounding box of the largest contour. The experiment is conducted on the 0.2 THz data set. The results show that the proposed detection method based on multi-feature fusion outperforms better than other methods employing a single feature or pairwise combination of these features. This method has good performance for detecting passive terahertz concealed objects.