3D human modeling using virtual multi-view stereopsis and object-camera motion estimation

Abstract

Abstract-This paper presents a method for multi-view 3D modeling of human bodies using virtual stereopsis. The algorithm expands and improves the method used in [5], but unlike that method, our approach does not require multiple calibrated cameras and/or carefully-positioned turn tables. Instead, an algorithm using SIFT feature extraction is employed and an accurate motion estimation is performed to calculate the position of virtual cameras around the object. That is, by employing a single pair of cameras mounted on a same tripod, our algorithm computes the relative pose between camera and object and creates virtual cameras from the consecutive images in the video sequence. Besides not requiring any special setup, another advantage of our method is in the simplicity to obtain denser models if necessary: by only increasing the number of sampled images during the object-camera motion. As the quantitative results presented here demonstrate, our method compares to the PMVS method, while it makes it much simpler and cost-effective to implement.

Experimenting With Autonomous Calibration of a Camera Rig on a Vision Sensor Network

“…Since the main application for our camera rig is for 3D object modeling, we tested the accuracy of the calibration by reconstructing a sphere based on the idea of visual hull (Laurentini, 1994) and the human body using an algorithm for multi-view 3D modeling presented in (Lam et al, 2009). As before, the tests were conducted for both synthetic data and real data.…”

Comparing limb-volume measurement techniques: 3D models from an infrared depth sensor versus water displacement

“…Its high-end, 6-figure cost makes it definitely not appropriate for self monitoring. Technologies for 3D reconstruction for human bodies and other objects have been around for many years [14], [15], [16], [17], [18], [19], [20], [21]. They provide very accurate and reliable models, which can be used for measuring limb volumes with equal or better accuracy than the methods used today.…”

“…Despite being a topic currently in high demand, 3D modeling of the human body poses a great challenge and while earlier approaches employing multi-view stereopsis [17], [18], [19], [20] could build reasonable 3D models from simple 2D images, they were also not as reliable as newer methods based on new IR sensing technologies. Besides, they required well calibrated cameras [17], [18] and some times, a setup of multiple cameras [19], [20].…”

“…Besides, they required well calibrated cameras [17], [18] and some times, a setup of multiple cameras [19], [20]. In the work presented in [15], human bodies were very accurately modeled using laser scanners at the price of expensive hardware for range imaging (over US$20,000).…”

Stereo Vision Auto-Alignment And The Unsupervised Search For Objects Of Interest With Depth Estimation

“…It was found that having an affine transform matrix was shown to outperform features matching in recognizing object orientations. Another method employing the SIFT feature extraction was presented by Lam et al (2009). The authors implemented virtual stereopsis in 3D modeling of human bodies without the use of multiple calibrated cameras.…”