The vehicle-mounted forward-looking ground penetrating imaging radar (VFGPIR) is a feasible facility to detect shallow buried objects, but it faces a high false alarm rate. The combined detection, by using the sequence features, is proposed as the key to solve this problem. Firstly, we use the Fisher's discriminating ratio (FDR) to quantificationally evaluate the discriminating power of the individual feature component extracted from the single or serial images. Since the optimal individual sequence feature cannot satisfy the needs, a multi-features joint detection method based on the decision-level weighting fusion is presented in this paper. Finally, the receiver operating characteristic (ROC) curve is exploited to demonstrate the validity of the proposed method. Experimental results of the data show that the features extracted from the sequence images have better discriminating power, and the suggested method has a better detection performance than that of the optimal individual sequence feature, feature vector, and majority voting fusion rule. The approach is expected to satisfy the requirements of landmine detection in the practical application.