Objects detection is not only an important research direction in computer vision, but also the basis of object tracking and behavior detection. In this paper, we analyze the characteristics and principles of Binarized normed gradients (BING) algorithm, and propose an objects detection algorithm based on multi-BING feature model. At first, the proposed algorithm uses K-means clustering algorithm to cluster the training data, and then trains each category of images to establish the corresponding BING feature model. At the detection stage, multiple BING feature models are respectively used for testing. We collected all the detected results from all models as the final detection results. Experimental results demonstrate that the proposed algorithm effectively improves the object detection rate (DR) under various overlaps at the expense of a small amount of time by generating a small set of high-quality proposals.