This paper proposes an improved least squares acoustic source localization method using a sensor array.Firstly,according to the effect of wind velocity,direction,and temperature on sound velocity,the correction model of the observation is established.The weight matrix is correct by the law of error propagation,after that,the least square method is used to give out localization calculation.Considering that the existence of the gross error is possible,IGG3 robust model is integrated into the data processing to improve the anti-interference ability of the system.Additionally,the wrong observations are judged by observing the residuals vector,and then the time series chaos question is solved by adjusting the series.Finally,the effectiveness and feasibility of the proposed method are verified by the real experiments.The results show that the improved least square localization method achieves the localization accuracy within 4 m.The proposed method can not only effectively reduce the influence of gross errors,but also deal with the time series chaos problem perfectly,and the reliability of the system can be improved.This research work may provide some references for the design and application of acoustic source localization equipment.