Stereo vision technology, TOF, RGB binocular, structured light contrast
This content comes from the Internet
Current depth cameras can be divided into three types according to their working principles: TOF, RGB binocular, and structured light.
Parameter comparison of three cameras:
Consider it from aspects such as resolution, frame rate, software complexity, power consumption, etc.
(1) Resolution
It is difficult to improve the depth map resolution of the TOF solution, and generally cannot reach VGA (640x480) resolution. For example, the depth map resolution of Kinect2's TOF solution is only 512x424. The rear TOF depth camera resolution of the PHAB2 mobile phone jointly developed by Google and Lenovo is only 224x171. The TOF solution is limited by physical devices. It is difficult to achieve a resolution close to that of VGA. Even if it can be achieved, it will increase exponentially with the power consumption.
The resolution of structured light is within the recent range of use, and the resolution of the structured light solution will be much higher than the TOF solution. For example, the depth map of the current structured light solution can reach a resolution of around 1080p.
In terms of frame rate, the TOF solution can achieve a very high frame rate, almost hundreds of fps. The frame rate of the structured light solution will be lower, typically 30fps, but this is basically enough.
(3) Software complexity
Because structured light needs to decode the encoded structured light, the complexity is higher than that of direct ranging TOF.
(4) Power consumption
TOF is a full-scale laser irradiation, while structured light only irradiates a local area. For example, the pseudo-random speckle pattern of PrimeSense only covers less than one-tenth of the space. In addition, TOF emits high-frequency modulated pulses, while structured light projection patterns do not require high-frequency modulation, so the power consumption of structured light is much lower than that of TOF. Let’s take pseudo-random speckle structured light as an example. The power consumption of the structured light solution is less than one-tenth of TOF.
There are several calculation methods: One is to directly use ASIC (Application Specific Integrated Circuit) for calculation. The cost is slightly higher, but the processing speed is fast, and it supports high frame rate and high-resolution depth cameras. The key is that it consumes less power than general-purpose chips. The second is DSP+ software algorithm. The cost is similar to that of ASIC, but it cannot support high frame rate and high resolution, and the power consumption is slightly higher than ASIC. The third is to directly use the AP (Application Processor) of the mobile phone to perform pure software calculations. This does not require additional hardware costs, but consumes more AP computing resources. It also does not support high frame rates and high resolutions, and consumes relatively large power.
