Infrared dim and small target detection is a hot research topic in the field of target recognition. Considering the low signal-to-noise ratio of targets in infrared dim and small images and the large scale variation of imaging targets, constructs an infrared dim and small target detection framework considering both local salient features and global salient features. Constructs a saliency detection algorithm based on multi-scale convolution kernel, and calculates the saliency map of the algorithm and the spectral residual algorithm respectively; after obtaining the local and global saliency map, this paper uses the morphological method to fuse the saliency map, and then uses the adaptive threshold method to perform binary segmentation. Experimental results on a given public data set show that the proposed method has obvious advantages over the benchmark saliency algorithm in terms of target detection accuracy and false alarm rate.