In view of the characteristics of flicker and jitter in the moving target imaging of the through-wall-radar, the moving target tracking algorithm based on improved Camshift for through-wall-radar imaging was proposed. With respect to the continuous multi-frame images formed by through-wall-radar and the corresponding distribution maps of color probability, the target prediction process was firstly introduced to determine the wave gate for target searching in the image, removing clutter interference outside the wave gate. Then in order to effectively extract target location, by utilizing the distribution map of color probability, the iterative adjusting strategy for the scale of target searching window was designed to adaptively match the target image with changing shape and size. Finally, a contiguous and smoothing trajectory of target moving was obtained by performing α-β filtering on the extracted multi-frame target locations. Experimental results of Multi-Input Multi-Output (MIMO) through-wall-radar show that compared with the traditional Camshift and Meanshift algorithms, the tracking error of the improved algorithm is reduced by 40.99% and 43.09% respectively, achieving a more accurate and smooth moving target trajectory.