Considering the data collection problem in wireless sensor networks

a matrix completion based data collection scheme is proposed. By transforming the data collection problem into two correlated convex sub-optimal problems according to the low rank matrix factorization model

an imporved Kaczmarz iteration based matrix completion algorithm is proposed to solve these two sub-optimal problems alternatively and iteratively. Furthermore

dimensional reduction operations are excecuted before projections in the Kaczmarz iteration according to the Johnson-Lindenstrauss lemma with the goal of accelerating the speed of the algorithm. Experiment results show that the proposed algorithm presents superior reconstruction performance in terms of reconstruction accuracy

the probability of successful reconstruction

and reconstruction time than the existed algorithms.