Abstract: ‍ ‍Benefiting from the widespread applications of synthetic aperture radar （SAR） imaging technology and multiple input multiple output （MIMO） arrays in millimeter wave radar， high-resolution three-dimensional point clouds images can be obtained using MIMO arrays for direction of arrival （DOA） estimation based on the two-dimensional SAR imaging results of automotive millimeter wave radar. However， in actual millimeter wave radar echo data， the amplitude and phase errors between array elements may cause multiple inaccurate peaks in the angle spectrum， a decrease in the main lobe peak， and an increase in the side peaks， all of which can lead to a degradation in DOA estimation performance. This affects the target focusing effect， and makes it difficult to meet the requirements of target classification and recognition. Due to the time variant of amplitude and phase errors after SAR imaging， correctly compensating amplitude and phase errors using existing amplitude and phase error calibration algorithms is difficult. Moreover， most current algorithms require estimating the covariance matrix to seek amplitude and phase errors， which results in high complexity and cannot achieve the effect of engineering. Therefore， the amplitude and phase error calibration of the MIMO arrays will become a hot topic for automotive millimeter wave radars， which is of great importance for achieving real-time high-resolution imaging and assisting driving. This paper first describes the construction of a three-dimensional echo model of the automotive millimeter wave radar and proposes a high-resolution three-dimensional point clouds imaging method based on the back projection algorithm. Compared to the traditional point clouds imaging method， this method can achieve a higher azimuth and height resolution. To ensure DOA estimation performance， we demonstrate the equivalence between the amplitude and phase error vectors of the array and the eigenvectors corresponding to the maximum eigenvalues of the covariance matrix by utilizing a one-dimensional array signal model that contains amplitude and phase information. Thus， according to this conclusion， we propose a novel amplitude and phase error calibration algorithm based on PAST technology. This algorithm utilizes PAST technology to avoid estimating the covariance matrix and feature decomposition， directly seeking the main feature vector， and enormously improving the error estimation efficiency. With this method， amplitude and phase errors can be corrected effectively and the consistency between each array element can be maintained. Compared to existing algorithms， it has notable advantages in computational efficiency and accuracy. Theoretical analysis， numerical simulation， and practical experiments verified that the proposed algorithm can maintain robust performance in complex imaging scenarios， effectively achieving amplitude and phase error calibration of MIMO arrays. Especially in measured data， this algorithm can produce better high-resolution three-dimensional point clouds images after correcting amplitude and phase errors. This is particularly evident in the point clouds of the steps. The enlarged image of the step indicates that the algorithm proposed in this paper can effectively eliminate the interference in the scene. After calibration， the point cloud image was more focused， highlighting the contour of the target. This point cloud image greatly facilitates subsequent target recognition and feature extraction. In future work， expanding the practical application of the algorithm proposed in this paper， will be necessary.