Acoustic signals have been used to sense important contextual information of targets such as location and movement. In this paper, we explore its potential for measuring the geometric information of moving targets, i.e., shape and size. To do so, we propose SONDAR, a novel shape and size measurement system using Inverse Synthetic Aperture Radar (ISAR) imaging on commodity devices. We first propose a Doppler-free echo alignment method to accurately align the target reflections even if there exists a severe Doppler effect. Then we down-convert the received signals reflected from multiple scatters on the target and construct a modulated downchirp signal to generate the image. We further develop a lightweight approach to extract the geometric information from a 2-D frequency image. We implement and evaluate a proof-of-concept system on both a Bela platform and a smartphone. Extensive experiments show that we can correctly estimate the target shape and achieve a millimeter-level size measurement accuracy. Our system can achieve high accuracies even when the target moves along a deviated trajectory, at a relatively high speed, and under obstruction.