Practical Motion Capture in Everyday Surroundings

2:34 am 立體追蹤 (3D Motion Tracker), 電腦圖像(Computer Graphic)

mit.jpg

By Daniel Vlasic, Rolf Adelsberger, Giovanni Vannucci, John Barnwell, Markus Gross, Wojciech Matusik, Jovan Popovi´c

Commercial motion-capture systems produce excellent in-studio reconstructions, but offer no comparable solution for acquisition in everyday environments. We present a system for acquiring motions almost anywhere. This wearable system gathers ultrasonic time-of-flight and inertial measurements with a set of inexpensive miniature sensors worn on the garment. After recording, the information is combined using an Extended Kalman Filter to reconstruct joint configurations of a body. Experimental results show that even motions that are traditionally difficult to acquire are recorded with ease within their natural settings. Although our prototype does not reliably recover the global transformation, we show that the resulting motions are visually similar to the original ones, and that the combined acoustic and inertial system reduces the drift commonly observed in purely inertial systems. Our final results suggest that this system could become a versatile input device for a variety of augmented-reality applications.

商業化的動作擷取系統多數可在工作室內重建出來,但提供無法提供解決方法於每日週遭環境中。因次我們提供一套可以在任何地方使用的動作擷取系統。這個 穿戴式系統利用低廉的微型傳感器miniature sensors收集超音波time-of-flight與慣性量測資料。經過紀錄後,將資訊結合外部Kalman Filter建置身體關節點結構。實驗結果顯示,即使動作一般屬於難以透過自然設定被擷取與紀錄。隨然我們的Prototype並不是很穩定的去完成全部3D人體的變形,但我們顯示的結果是Motion與視覺化的人物非常相近,並結合音響和慣性的系統,降低drift,研究的最終結果是建議這套系統可以成為多功能輸入裝置給予擴增實境方面的應用

相關連結 Extenal Link

Practical Motion Capture in Everyday Surroundings

Jovan Popović

One Response

  1. aloi Says:

    Huge Motion Capture Space
    Each Visualeyez tracker can capture over a huge 190m3 rectangular parachute-shaped space. This is due to its widest angle operation of all optical motion capture systems in the world (see Specifications). A 4-tracker system can be easily arranged to yield a 6m x 6m full-body motion capture area if desired. A round camera-based system would require more than 9 cameras to capture the same space as a single Visualeyez tracker does!

    Intrinsic Marker ID Tracking

    Each marker of a Visualeyez system is automatically assigned a distinct ID number during operation. The captured 3D data is automatically associated with this ID by the system. No data cleaning, no special marker arrangement, no special actor motions, nor any post-processing is ever required to help the system correctly identify a marker. Absolutely no marker swapping problem will ever occur!

    Arbitrary Sampling Frequency
    The marker sampling frequency of a Visualeyez system can be set to any value to suit an application. For tracking a slow motion the sampling rate can be set to be extremely low, for days or even weeks of continuous capture. Since there is no marker identification problem, high frequency sampling is not necessary except for tracking very high speed motions. For applications in which very high-speed and low-speed motions occur simultaneously, ‘multi-rate sampling’ (see below) can be applied to make the most effective use of the system capabilities.

    Multi-Rate Sampling
    Sampling slow motions at high frequency would result in large amount of useless data, wasting storage space and any data processing efforts. Visualeyez system allows different markers to be sampled at different frequencies (see Mini TCM or TCM8) to capture motions of very different speeds at about the same spatial resolution. This is highly useful for many scientific applications. For example, the golf club head moves much faster than the golfer’s body in a golf-swing motion. To capture at approximately the same spatial resolution, the clubhead markers can be captured four times faster than the body markers. (This is done in this example application.)
    Microsecond Temporal Sampling Precision
    The marker positions captured by a Visualeyez system are sampled at microsecond precision and the exact time of capture is output with the data. This is highly beneficial for scientific applications which require precise uniform motion data feedback. This feature also allows temporal marker capture problems to be resolved by mathematical interpolation.

    Note that all optical motion capture systems using digital sensors (e.g., CCD or CMOS image sensors) have temporal marker capture problems. Systems based on image processing technologies have particularly worse problems because the exact time of sampling a marker depends on where that marker appears on an image, which varies as the marker moves. Moreover images from several cameras are required to triangulate the position of a marker. Hence the temporal problems of these systems cannot be practically resolved at all! (See the picture on the right. The bent bat is due to the bottom portion of the image being sampled earlier than the top portion.)

    Zero Wrong Data Probability
    Wrong data is worse than no data. An optical marker can be partially occluded and yield inaccurate position data as a result. The larger a marker is the more likely it can be partially occluded, and the larger the inaccuracy can be. Visualeyez system uses tiny LEDs as markers which are nearly impossible to occlude partially. Should partial occlusion ever occur, the introduced inaccuracy will still be very small. Hence there is practically no probability of generating wrong data due to partial occlusion by a Visualeyez system.

    Fast & Reliable Real-Time Operation
    Visualeyez system outputs the captured data after two sampling periods. For a VZ4000 tracker, this means the data can be output within less than 0.5 millisecond (

Leave a Comment

Your comment

You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <code> <em> <i> <strike> <strong>

Please note: Comment moderation is enabled and may delay your comment. There is no need to resubmit your comment.

C.M.U.