Optimized recognition method for familiar target in familiar scene

The invention discloses an optimized recognition method for a familiar target in a familiar scene. The method comprises the following steps: 1, acquiring a familiar target image and constructing a familiar target library; 2, performing salient region detection on all familiar target images in the familiar target library, judging whether a detected target is a familiar target, and if yes, outputting a recognition success result; and if not, establishing an intelligent optimized space-time sequence fusion model according to the familiar target image which is not successfully identified, then identifying the familiar target image which is not successfully identified based on the intelligent optimized space-time sequence fusion model, and outputting an identification result. According to the method, an irregular target detection technology based on a saliency pixel cluster and a target pre-judgment modeling and heuristic optimization technology based on full feature fusion are adopted, and based on combination of