Indoor Mobile Target Perception Model Construction Based on Fusion-EK Algorithm in Smart Cities

In order to break through the current precision and sensitivity constraints of indoor mobile target perception methods, this paper introduces a model based on a multi-source data fusion algorithm through the extended Kalman filter to sense dynamic targets indoors. The model first combines multi-source information and performs nonlinear processing, then performs singular value decomposition for perception process enhancement, and lastly uses a gated recurrent neural network to implement mobile target tracking and localization. Experimental findings suggest that the model iterates over the York campus benchmark video dataset in 2.45 s and mean absolute error of perception accuracy is 2.39%. The optimized model achieves the perception accuracy of 0.952 and the average distance error of 0.281 m. The GSFE model overall has better feature recognition in perceiving dynamic targets and high precision and sensitivity, overcoming the current societal inadequacies of dynamic target perception.