Abstract: The current roadside fusion perception system generally uses millimeter wave radar as the main sensor, but its detection results are easily affected by detection algorithms and environmental factors. Traditional target tracking methods rely on information such as position and velocity for matching, making it difficult to distinguish between real targets and noise, which can easily lead to target loss or false detection. This article focuses on the performance optimization methods of millimeter wave radar in roadside fusion perception systems. The proposed method has improved the performance of roadside millimeter wave radar in multiple aspects such as target detection rate, target tracking effect, and response time.