Accurate gas leakage source localization plays an important role in production and life. This paper presents a research on gas leakage source localization based on an improved UAMP-BG algorithm. The direction-of-arrival (DOA) estimation problem can be transformed into sparse recovery of signals. Unitary transformation is introduced to the traditional approximate message passing (AMP) algorithm, and then unitary approximate message passing (UAMP) is used for sparse recovery of signals. To enhance sparsity, this paper improves the UAMP algorithm. In the UAMP iteration process, the LASSO regression function is introduced. First, UAMP iteration is used to optimize model parameters. Then, after a fixed number of iterations, the sparsity of the LASSO regression function is used to re-optimize the estimated parameters, and then UAMP iteration continues. Thus, it maintains the efficient iteration of UAMP and at the same time has the sparsity of L1 regularization. The location of non-zero elements is quickly estimated