It has been reported that Oculus' sensors can be used to take pictures, so some people worry that hackers may start from this and record our every move in the virtual world.
But are we overly worried? Motheboard recently interviewed the discoverer of the Oculus Rift sensor to capture user images - Oliver Kreylos, a researcher at the University of California, Davis. The following is a compilation of interview clips:
Q: Compared with Vive's Lighthouse sensor, is Oculus' location tracking method relatively sloppy, or does it have its own advantages? Is Oculus worth the risk in the product?
Kreylos: This question is very broad, so let me start with some background notes. Cameras and tracking markers (such as Oculus's LEDs) have a long history of tracking in 3 degrees of freedom and 6 degrees of freedom. Almost all high-end dynamic capture systems are based on this approach. This is the basis for Wiimote (Wii Handle) tracking and has been used for PC games in the form of NaturalPoint's TrackIR Head Tracker for many years.
This method is so popular (especially in the amateur community) because it provides high quality results without the need for too much custom hardware. The camera is very common and the price is close to the people. The only other rigid requirement hardware is the tracking marker, but even amateurs can make it, and everything else is done in software.
When Oculus looked for an excellent and inexpensive 6-DOF tracking system for the Rift DK2, the camera was an obvious and most suitable option. However, Oculus did not use the standard approach, but significantly reduced the computational load of the standard algorithm by identifying its own tracking LEDs to the camera. Therefore, the DK2 tracking system exceeds many people's expectations, including myself.
When Oculus moved from DK2 to CV1 (first-generation consumer version), it used a proven tracking system and optimized by using better hardware (high-resolution camera, using global shutter instead of rolling shutter) It is also a reasonable engineering decision. From a purely technical point of view, this is the right choice. Although there are certain problems, Oculus' Constellation is an excellent 6-degree-of-freedom tracking system. Since Constellation is relatively closed, I didn't do a detailed analysis like Valve's Lighthouse system, but I expect the two to be of equal quality.
I think Lighthouse is more elegant, it collects less raw data, has less computational processing, and achieves the same results. The basic input to the positioning calculation algorithm of the two systems is the three-dimensional ray starting from the center point (camera focus or Lighthouse base station center) and pointing to the tracking LED or photodiode in space. Constellation derives these rays by capturing high-resolution images, transmitting them to the host PC, finding spots of bright pixels in those images, and calculating their center (x, y) position. The same is true for Lighthouse: when the scanning laser meets a photodiode (a device that converts light into a current), the time is converted to an angle based on the known angular velocity of the laser, and the resulting angle is sent to the host PC, then the Lighthouse system Calculate accordingly. Constellation needs to send about 60MB of data to the host, which puts a lot of pressure on the host's USB subsystem, and Lighthouse sends very little data (I estimate only a few tens of KB), it can achieve wireless transmission without causing anything. problem.
But this does not mean that Constellation is a sloppy system. It was the most advanced system when it was developed, but Valve later introduced an unexpected, more elegant system, and Oculus had to be replaced too late. This is an innovation, I don't think anyone is expecting the arrival of the Lighthouse system.
The main disadvantage of Constellation is that some users may have problems during use due to higher USB bandwidth requirements. Another non-technical disadvantage is that high resolution images are sent from a number of cameras to the host PC, and these cameras must be placed in an almost ideal location. I personally think that some hackers have little risk of getting these images, but I can't deny that this is theoretically possible.
At this point Oculus can take a different approach to avoid this problem. The Wiimote I mentioned earlier is also tracked by the camera. But unlike Constellation, Wiimote does not send images to the host. The step of converting the image to the (x, y) LED position is done inside the camera chip itself, done by a custom chip. If Oculus adopts this approach and integrates such ASICs (application-specific integrated circuits or custom circuits with specific uses) into the camera itself, they can avoid Constellation problems. It can reduce the bandwidth from the camera to the host by about 1000 times (solving most users' problems), while others can't snoop the transmitted image because the image will never be sent to the host PC.
I don't know if you can call it "slack", to a certain extent Oculus engineers ignore the possible privacy issues, or this is the decision they made after being measured. But sending images to a host has several benefits: designing an ASIC takes time and money, software solutions are more flexible, and it's easier to improve over time. Oculus may be currently researching algorithms that use camera images to track objects other than LEDs, such as Leap Motion or Kinect to bring the user's hand, or even the whole body, into VR.
In addition, Oculus has always referred to CV1 cameras as "sensors" rather than "cameras" and has always insisted that they are not the same as cameras. Initially this made me think they did integrate the image processing ASIC into the camera, but the result was not.
Q: Is it possible to capture "images" with Vive's sensors, even without the traditional camera?
Kreylos: Lighthouse itself does not collect any data that can reconstruct the user's environment or the user's own image. The only data sent from the Lighthouse to the host PC by the tracking device is the timestamp at which each photodiode that can be tracked meets the scanning laser and samples from the integrated inertial measurement unit that can be tracked. Together, the system is allowed to establish a position and direction that can be tracked in 3D space. In theory, by observing the position of all traceables over a long period of time, you can reconstruct a rough 3D model of the user's environment, but that's it.
The Vive head-up display is equipped with a front-facing camera, which is a standard webcam and is connected to the operating system of the host PC. Therefore, the camera is as vulnerable to attack as all other webcams. Unlike Constellation, the Vive system does not require this camera. If users are concerned about privacy issues, they can use tape to cover the camera.
Q: The complexity of the process of obtaining identifiable images (and the relative limitations of hardware availability) is enough to discourage hackers from invading sensors on a large scale?
Kreylos: I can't judge because I am not a computer security expert. I know the real case of a remote attacker accessing a webcam, but I don't know how these attacks are implemented and whether they are suitable for the Constellation camera. The approach I follow is simple, but it may not work for remote attackers because it requires patching the Linux kernel's webcam driver to recognize the Constellation camera.
What I can say is that the widely deployed "standard" webcam attack does not work with Constellation cameras because they don't "proclaim themselves to be standard web cameras" to the operating system. If a common webcam attack runs on an Oculus user's computer, then the attack will not find the webcam. Since the Constellation camera is still a standard webcam, hackers can modify the attack to target them, but I don't know how much it needs to be modified.
Another complication for potential attackers is that Oculus' Runtime software uses a Constellation camera when the heads are active. If another process is already using the camera, the existing webcam vulnerabilities may not be taken over, or the tracing may be turned off, which warns the user of "rogue." In addition, each Constellation camera is equipped with an indicator light. I don't know if these are tied to the camera sensor at the circuit level, but I know that when I start using the software to record images, the camera's indicator light is on.
Closed impellers have a back and front wall around the vanes, to increase strength. Closed impellers are used primarily in larger pumps and can be used in applications that handle suspended-solid service.
These types of impellers are commonly found in clear liquid applications. They don't do well with solids and are difficult to clean if they become clogged.
Still not sure which impeller will be best for your next pumping application? Be sure to check wtih an application engineer who is well versed in pumps and selection best practices.
Closed Impeller,Closed Impeller Design,Closed Impeller Pump,Semi Closed Impeller
Shenyang Zhicheng Heavy Machinery Manufacturing Co., Ltd. , https://www.zhichengmachinery.com