In recent years, cognitive radar techniques have been developed quickly. How to improve radar detection performance basing on the obtained sensing information has become hot topic. A cognitive target detection method during tracking stage is presented based on tracking information, in order to improve target detection and tracking performance. For the target during tracking stage, the target existence has been proved, therefore the tracking information can be sent to target detector to expect an improved detection performance. As to the proposed cognitive detection method, firstly, target motion characteristics are extracted from tracking information, and a predicted region is established for the next instant; secondly, detection thresholds in the predicted region are adjusted by using target prediction probability information, under the Bayesian criterion and constant false track rate; finally, cognitive target detection is accomplished. Numerical results indicate that the cognitive target detection method can significantly improve target detection and tracking performance.