Nowadays, smart communities are developing rapidly. The construction of various public facilities and buildings makes the community environment complex, which affects the safety of vulnerable groups of residents. Thus, the localization of residents in real time is particularly important. As one of the precise intelligent services, intelligent localization mainly realizes the trajectory tracking of the elderly and young people in a community through radio frequency identification (RFID) technology to ensure security. In addition, it can also analyze the trend of residents gathering so that it is capable of putting forward suggestions for the construction of public facilities in communities. In this study, to tackle the traditional problems such as poor localization accuracy and localization result deviation caused by various noises when an RFID reader transmits data, we use a Kalman filter to eliminate the process noise and observation noise during signal transmission and insert the improved density clustering algorithm into the traditional algorithm to eliminate the influence of environmental noise. Further, we design a Kalman filter method based on improved density peak clustering (K-IDPC) for localization. Experimental verification shows that the localization accuracy of K-IDPC is about 0.565 m, which is greatly improved compared with that of the Kalman filter.