Noninvasive detection of body composition plays a significant role in the improvement of life quality and reduction in complications of the patients, and the near-infrared (NIR) spectroscopy, with the advantages of painlessness and convenience, is considered as the most promising tool for the online noninvasive monitoring of body composition. However, quite different from other fields of online detection using NIR spectroscopy, such as food safety and environment monitoring, noninvasive detection of body composition demands higher precision of the instruments as well as more rigorousness of measurement conditions. Therefore, new challenges emerge when NIR spectroscopy is applied to the noninvasive detection of body composition, which, in this paper, are first concluded from the aspects of measurement methods, measurement conditions, instrument precision, multicomponent influence, individual difference and novel weak-signal extraction method based on our previous research in the cutting-edge field of NIR noninvasive blood glucose detection. Moreover, novel ideas and approaches of our group to solve these problems are introduced, which may provide evidence for the future development of noninvasive blood glucose detection, and further contribute to the noninvasive detection of other body compositions using NIR spectroscopy.