A new algorithm based on the principle of light polarization can significantly increase the resolution of commercial inexpensive sensors. Researchers at MIT claim they can increase the resolution of traditional 3D imaging equipment by a factor of 1000 by using light polarization. In addition, they said they could use the technology to integrate 3D cameras on smartphone devices, even take photos and rehabilitate objects on 3D printers, and have proposed technologies that can be applied to driverless cars. Achuta Kadambi, MD, Ph.D., imaging lab at MIT, one of the developers of the system, said: "While the current 3D cameras are small enough to be installed in mobile phones, they compromise 3D imaging, Resulting in a very coarse geometry of the imaged object, making it difficult to recover, but this can easily be accomplished using polarization technology, which is better than many commercially available laser scanners with the addition of polarizers, even with low quality imaging devices Image quality. " Researchers at the International Conference on Computer Vision, to be held in Santiago, Chile, published an article detailing their system and named it Polarization 3D. Kadambi is the lead author of the article, co-funded by Rameshar Rasker and Boxin Shi, and co-author Vage Taamazyan from the Skolkovo Institute of Science and Technology in Russia. The Institute was established by MIT in 2011. Polarization can change the way light reflects on an object's surface. If the light is on the surface of the object, most of the light will be absorbed, but the reflected light will always have the same direction of polarization as the incident light. However, at high-angle reflections, light of a particular angular range is more easily reflected. Therefore, the polarization information of the reflected light carries the surface information of the reflecting surface. MIT's team said that although the above phenomenon of polarization has been discovered centuries ago, it is difficult to actually use it because there is a fundamental limitation of polarized light: light of a particular polarization can be distinguished from light of the opposite polarization. Polarization combined with depth sensing To eliminate this limitation, imaging lab researchers used other methods to perform coarse calculations of depth information, such as the time it takes to use light to return from the reflective surface to the original material. However, even with such additional information, it is still complicated to calculate the surface of the object from the polarization information. However, real-time processing can be realized through an image processing unit, which is a special image chip commonly used in video game machines. The researchers' experimental setup included a Microsoft Kinect system, a structure that places a normal polarized camera lens in the imaging lens to calculate the time it takes for light to travel back and forth. In each experiment, the researchers took three pictures of the experimental object by rotating the polarizing filter and compared the light intensity changes between the images by an algorithm. For its part, within a few meters, the Kinect system reconstructs surface features with centimeter accuracy. But by adding polarization information, researchers can increase the accuracy of system measurements to the order of microns. As a comparison, the researchers compared the test results with a high-precision laser scanner and found that the accuracy of polarization 3D was still higher. MIT's paper also shows the broad application of this technology and can even be used in driverless vehicles. The team said, "Today's experimental driverless vehicles are highly reliable under normal lighting conditions, but their visual algorithms fail in rain, snow, fog, etc. Because particulate water can affect light in the air Unpredictable interference with the transmission, making the results unpredictable. " put one's oar in The researchers said that in some simple tests that have not been possible with traditional computer vision algorithms, the use of their system can capture the information contained in the light scattering in the interference. Kadambi commented: "Getting information by using scatterings is only a small step when it is possible to enlighten more on big problems." Yoav Schechner, an associate professor of electrical engineering at the Technion-Israel Institute of Technology, commented: "This work combines two principles of 3D sensing that take advantage of their strengths and shorten one. One principle provides the imaging range per pixel, namely The current state of the art for most 3D imaging technologies. The other principle, which does not cover the imaging range, derives the target slope, in other words tells us the flatness of the surface of the object. " "This work addresses a single issue by interpenetrating the principles, which almost completely overcomes the limitations of polarization-based shape measurement methods and makes polarization more attractive in machine vision."