Talk:Time-of-flight camera/Archives/2014
![]() | This is an archive of past discussions. Do not edit the contents of this page. If you wish to start a new discussion or revive an old one, please do so on the current talk page. |
Needs work
This page needs some work. Time-of-flight cameras have been around for a while, in various forms. There are several different technologies. The Swiss Ranger [1] works by modulating the outgoing beam with an RF carrier, then measuring the phase shift of that carrier on the receive side. This is an old trick, with some problems (there's a modular error; at multiples of the RF carrier wavelength). There are devices with a pulsed laser and a fast counter behind every pixel. The devices from Advanced Scientific Concepts [2] work that way. That's the best approach, but currently an expensive one. There are also "range gated imagers", where anything outside a specified distance range can be suppressed. Those are useful for seeing through fog.[3] The Z-cam[4] is also a range-gated system, not a true depth measurer, although you can accumulate depth over multiple frames by changing the gate timing. --John Nagle (talk) 22:33, 12 August 2009 (UTC)
- (On talk pages, add new paragraphs, don't edit previous material. Separated out comment.)
- The PMD = ToF works with a very similar principle! This is an well known approach. But with new solid state standard CMOS devices, the phase detection could be realized in one imager. The uniqueness range ToF or PMD principle can be increased up to 500m with frequency switching or with special codes! ToFExpert. ExpertToF (talk) 14:54, 19 August 2009 (UTC)
- Oh, the PMD thing is a phase-shift detector. That wasn't clear before, and it's still not clear from the explanation. --John Nagle (talk) 16:50, 19 August 2009 (UTC)
- To TofExpert: Thank you for the correction, otherwise I had to do it... :-) Captaindistance (talk) 08:18, 24 August 2009 (UTC)
- Flash-imaging LIDAR
- Most of the devices described in this page are short/medium-range consumer-class to industrial/commercial-class devices, and are pitched in the digital camera/sensor space. Advanced Scientific Concepts focuses on a somewhat of a different class of devices (science/research, government/military) extending from the LIDAR space, typically referred to as "scannerless" or "flash-imaging" LIDAR, or some variant thereof. ASC is not the only entity which has operated in that space, but it does seem to market the most aggressively. Someone from ASC even created a flash LIDAR article focused heavily on the company and its preferred terminology. The underlying concept is the same, and while it should be clear that they occupy distinct markets, the articles have a significant amount of overlap. I think the two articles should be restructured so that there is one main article covering the concept. It doesn't need to be a full merge. Dancter (talk) 00:12, 30 July 2011 (UTC)
- 3D Flash LIDAR vs. other TOF approaches
- While a TOF "measurement" system/approach is not a unique concept, there are significant enough between using a TOF CMOS/CCD sensor (e.g. PMD, Canesta, et al) vs. an TOF avalanche photo diode with CMOS Read-out IC (ASC, Raytheon Vision Systems). The analogy is similar to "any vehicle capable of supporting itself in the air" should be consider an "airplane" when there are enormous differences between a helicopter, a hang-glider, a turbo-prop airplane and the Space Shuttle Orbiter. The approaches are different enough to justify individual pages, though a "superset" page directing to the different approaches would be worthwhile. There is a lot of development and breakthroughs occurring in 3D Flash LIDAR with the next generation iterations (lasers & cameras) beginning to emerge in various applications. — Preceding unsigned comment added by Tlaux (talk • contribs) 20:19, 2 August 2011 (UTC)
The article has at least one serious error. The Canesta technology has never been anywhere close to the 3DV range gate mechanism, and is instead of the phase measuring variety. — Preceding unsigned comment added by 75.147.132.46 (talk) 04:34, 27 September 2011 (UTC)
- I tried to correct the error. The misleading text was introduced by the "Antonio Medina" editor. Frankly, even without those additions, the names and distinctions between the different approaches seem a bit muddled to me. As I understand it, while the approaches used by MESA, PMDTec, and Canesta do not use square-wave pulses and shuttering mechanisms the way 3DV's does, both types perform the same function of measuring a phase shift of a modulated-light signal through some sort of gating or cross-correlation at the sensor end. Dancter (talk) 02:04, 7 October 2011 (UTC)
Target market of mentioned cameras?
Can we get the target market for each of the cameras listed (home computer users, video game players, military, robotics, OEM etc) ? --TiagoTiago (talk) 08:44, 6 July 2012 (UTC)
New Kinect reportedly has a time-of-flight camera.
The new Kinect supposedly is a real per-pixel time-of-flight device.[5]. Not enough info is available on this yet. --John Nagle (talk) 18:52, 21 May 2013 (UTC)
- I'm not sure what you mean by "real per-pixel". I think the Wired article is certainly enough to confirm that the next-generation Kinect sensor uses a time-of-flight camera. It may not work exactly as currently described in this article, but this article isn't written all that well at the moment. Most of the edits to this article have been from industry folk, and tend to be overly technical and/or promotional, and often skewed to one company's particular implementation, rather than the general principle. Dancter (talk) 21:32, 21 May 2013 (UTC)
- "Real per-pixel" means a time counter on every pixel like the Advanced Scientific Concepts devices, as opposed to a gated imager like the ZCam. Where a photonic mixer device like the PMDTechnologies device falls is an interesting question. Watch for sources with more info. This is going to be important. --John Nagle (talk) 03:43, 22 May 2013 (UTC)
- It seems as if you perceive that ASC's counters are simply super-fast clocks, and that the ZCam basically works by taking a bunch of cross-sections. That's not a fair characterization of either technology. If anything, the ZCam system is somewhat of a hybrid of both the "flash-imaging LIDAR" and the phase-demodulation techniques. Regardless, the next-generation Kinect sensor is not going to use ASC technology (note the analogy Mr. Laux uses about hang gliders and Space Shuttle Orbiters). Also keep in mind that the Kinect sensor is designed to be used for people in front of a TV, in a space no larger than a living room. Phase-wrapping is not really a major issue. Even from the currently available information, I think it's pretty clear (though unconfirmed) that the technology extends from the work at Canesta. Dancter (talk) 19:54, 22 May 2013 (UTC)
Pulse width, range and depth resolution
The article in it's current form states that
- The pulse width of the illumination determines the maximum range the camera can handle. With a pulse width of e.g. 50 ns, the range is limited to
This statement applies only to time gated, integrating sensing, i.e. the working principle in which the scene is illuminated with a pulse of specific length and the sensors shutter is opened for only the twice (full round trip time) pulse's duration. The larger the integrated signal, the larger the portion of the pulse to be reflected back to the sensor within the gated interval.
In those systems the range is in fact limited to the pulse's length.
However, gated, integrating sensing is susceptiple to environmental illumination. While it is possible to apodize the signal by background subtraction this method is not used in modern ToF cameras.
The other measurement principle involves the delay (LIDAR principle). In such systems are very short pulse is emitted and the delay until arrival of the reflection measured. In such systems the depth resolution in inversely proportional to the pulse duration: The longer the pulse, the larger the uncertainty in measured depth. Also the depth range is determined by value range the pulse repetition rate and the value range of the delay timing counter, the later however being trivially scalable to extreme length. For example a delay timer running at 10 GHz (3cm resolution) with 64 bit register size would allow for a range of nearly 30 lightyears.
Modern ToF cameras combine pulse duty cycle and delay measurements. Each sensor is gated at a high frequency and the integrated signal within each interval measured. Pulse arrival time is determined with a threshold circuit, at which time the ratio of signals measured and is determined, by which the fraction of the thresholded gating interval at which the pule maximum arrived can be derived from, increasing the depth resolution. For this method to work, the pulse duration must be as short as two gating intervals. For a 10 GHz gating frequency this would be 0.1 ns. Combined with a 4 bit signal processing this allows for depth resolution. — Preceding unsigned comment added by 2001:A60:F078:12:653E:A73A:BCB2:F6C4 (talk) 10:09, 5 June 2013 (UTC)
- Could you revise the article? Dancter (talk) 16:21, 11 June 2013 (UTC)