Fall is a common daily activity and a leading cause of death among the older adults. It reveals growing demands to use some non-invasive methods to detect the pose of older people and give a timely and efficient alert, especially in some place with high fall-risk. This article presents a radio tomographic imaging (RTI) based approach for fall detection. A wireless network organized by a group of radio-frequency sensors is used for human pose sensing in the vertical direction. The human body would cause the statistical shadowing losses on the passing links between pairs of nodes in the network. Then an attenuation image of body pose can be obtained by using the received signal strength measurements. The non-negative total variation minimization is used to reconstruct the gray image of body. The fall detection is cast as an image recognition problem. This is a new approach based on the use of RTI to enable the building of a fall detection system. Experimental studies are conducted to validate the proposed method.