In real scenes,the size of objects is often diverse,so it is difficult to detect all objects based on large image.In o rder to detect small size targets,this paper uses saliency map and stable region fusion to establish a small target detection algorithm model.Firstly,the salien cy map algorithm based on the color name space is used to generate saliency map, and the local stable region is extracted by the maximum stable extreme region (M SER) algorithm.The MSER algorithm is the most stable feature detection algorithm for image deformation.Next,pixel multiplicative fusion is used to fuse the stab le region and saliency map to reduce the false alarm probability.Finally,some im age processing procedures are called.It can effectively suppress the background and highlight the salient targets uniformly to infer and optimize the final resu lts.In order to verify the effectiveness and practicability of the algorithm,the performance of several mainstream algorithms,including AZ-NET,FPN and PGAN,is compared with PR curve.Through the test of sky data set and ground data set,it s hows that the algorithm can adapt to the change of target size,and is better tha n the existing small target detection algorithm in the detection rate and detect ion rate,and has good robustness.