Abstract: The existing solutions to prevent road traffic accidents and control road traffic noise pollution are to monitor the key areas by visual judgment and determine the event trigger type and location by auditory judgment. In order to realize the real time monitoring of highway abnormal sound sources, a direction of arrival (DOA) estimation of signal parameters via rotational invariance techniques of low frequency broadband sound source based on dual scale estimation of signal parameters via rotational (ESPRIT) is proposed, which is suitable for the distributed array composed of three rectangular sub-arrays with a triangular distribution. The algorithm uses the spatial translation invariance corresponding to the two scales of the distance between adjacent elements in the sub-array and the distance between adjacent sub-arrays in the distributed array to estimate the direction cosine respectively, and uses the defuzzification strategy based on the array distribution to achieve high-precision azimuth estimation. The simulation results verify the effectiveness of the algorithm, show that the DOA estimation accuracy of the distributed array based on the algorithm is better than that of a single uniform rectangular array with the same number of array elements and same array-element spacing, and analyze the relationship between the estimation accuracy and the length of the distributed baseline, which reflects the practical engineering application value of the algorithm.