Abstract: Ultra-wideband multiple-input multiple-output (MIMO) radar can obtain multi-dimensional information of the target, and has great advantages in target detection and human motion classification. However, in practical application, the human target imaging results obtained by the ultra-wideband MIMO radar are usually low-resolution, and the farther the target distance is, the lower the resolution of radar image is, which is hard to understand. In response to the above problems, this paper proposes a distance-assisted ultra-wideband MIMO radar image human body pose reconstruction method. First, the convolutional neural network is used to extract the signal strength and spatial position characteristics of the human target imaging, and then the deconvolution modules are used to reconstruct the position of each joint point of the human target. At the same time, considering that the result of radar imaging deteriorates as the distance increases, the distance of the target is used as auxiliary information to select the appropriate network model parameters to improve the accuracy of pose reconstruction. Experimental results show that this method can convert abstract human target radar images into easy-to-understand human joint poses, and has excellent pose reconstruction performance, which greatly enhances the visualization performance of traditional radar images. Meanwhile, the introduction of distance information improves the accuracy of pose reconstruction and effectively overcomes the influence of distance increase.