Abstract: Aiming at the noise source localization problem in the power industry, this paper proposes a frequencysliding generalized cross-correlation and a low-rank approximation alternative to process the sub-band generalized cross-correlation matrix and thereby obtain better time difference of arrival estimation. The steering response power localization algorithm is then used to locate the noise signal. The efficacy of this algorithm is demonstrated through simulations, showing superior precision in TDOA estimation and noise source localization compared to the traditional generalized cross-correlation methods. Based on this, an acoustic camera with a 72-channel microphone array is designed, which demonstrates high positioning accuracy and strong robustness in actual environments. At the same time, compared with the acoustic cameras using traditional generalized cross-correlation, the experimental results show that the positioning results of this method are better than those of the conventional generalized cross-correlation delay calculation methods.