人形机器人相机选择
人形机器人的一般功能需求是什么
基于这些需求该怎么选型深度相机
什么情况下需要使用激光雷达（在我的认知下，结构光相机和激光雷达的作用是类似的，可以代替激光雷达）
-
Translation:
Humanoid Camera Selection
What are the general functional requirements of a humanoid robot
How to choose a depth camera based on these requirements
Under what circumstances do you need to use lidar (in my understanding, the role of structured light camera and lidar is similar and can replace lidar)
Both are active depth cameras, what is the difference between the applications of TOF cameras and structured light cameras**********************
Time of Flight (TOF) cameras cast a laser onto the scene and measure how long it takes for light to return to the camera. A structured light camera projects a pattern such as striped bars onto objects / surfaces in the scene and calculates distance from that pattern.
The RealSense L515 uses principles similar to a TOF camera, whilst the RealSense SR300 and SR305 use the 'coded light' technique which is similar to structured light. Both of these RealSense camera models are now retired though, so I would recommend choosing a model from the RealSense 400 Series range of stereo depth cameras. They are called stereo because of depth image is constructed from a pair of left and right sensors.
The specifications of the camera models in the 400 Series range are compared at the link below.
https://www.intelrealsense.com/compare-depth-cameras/
In regard to the general functional requirements of a humanoid robot, it should be able to detect obstacles, avoid them and navigate a path around them or over them. An example of this is a humanoid robot called Digit that uses 1 lidar camera and 4 RealSense cameras that "provide complete ground coverage from approximately +2 meters to -1 meter ... there are 2 RealSense cameras on the front, 1 between the legs, and 1 on the rear".
https://twitter.com/agilityrobotics/status/1100477143241216000?lang=en
The camera between the legs helps the robot to walk on stairs.
If you anticipate that your robot will need to be able to sense distance at close range to the camera then the D435, D435i, D435f or D435if model will be a good choice. The differences between D435 and the other listed models are:
D435i is equipped with an IMU component for tracking camera rotation and acceleration.
D435f utilizes an IR pass filter to enhance depth noise quality and performance range in many robotic environments.
D435if has both the filter of the D435f and the IMU of the D435i.
If the robot does not have to sense surfaces that are closer to the camera than 0.5 meters then the D415, D455 and D457 models could be considered.
D415 has better image quality and less depth noise than the D435 models.
D455 is capable of depth-sensing at longer range and has 2x better accuracy over distance of the D435 models.
D457 supports fast GMSL type connections instead of USB and has IP65 protection against water and dust. It has a switch for using the camera in USB mode too, where it is detected as a D455 and performs as one but has the added environmental protection.
********************
飞行时间 (TOF) 相机将激光投射到场景上，并测量光线返回相机所需的时间。 结构光相机将条纹等图案投射到场景中的物体/表面上，并计算与该图案的距离。
RealSense L515 使用类似于 TOF 相机的原理，而 RealSense SR300 和 SR305 使用类似于结构光的“编码光”技术。 不过，这两种 RealSense 相机型号现在都已退役，因此我建议您从 RealSense 400 系列立体深度相机中选择一种型号。 它们被称为立体图像是因为深度图像是由一对左右传感器构成的。
以下链接比较了 400 系列相机型号的规格。https://www.intelrealsense.com/compare-depth-cameras/
关于类人机器人的一般功能要求，它应该能够检测障碍物、避开障碍物并在障碍物周围或上方导航。 这方面的一个例子是名为 Digit 的人形机器人，它使用 1 个激光雷达摄像头和 4 个实感摄像头，“提供从大约 +2 米到 -1 米的完整地面覆盖……前面有 2 个实感摄像头，腿之间有 1 个 , 和 1 个在后面”。https://twitter.com/agilityrobotics/status/1100477143241216000?lang=en
两腿之间的摄像头帮助机器人在楼梯上行走。
如果您预计您的机器人需要能够在近距离内感知相机的距离，那么 D435、D435i、D435f 或 D435if 型号将是一个不错的选择。 D435与其他列出的型号之间的区别是：D435i 配备了用于跟踪相机旋转和加速度的 IMU 组件。
D435f 利用 IR 通滤波器来增强许多机器人环境中的深度噪声质量和性能范围。
D435if既有D435f的滤波器，又有D435i的IMU。
如果机器人不需要感应距离相机小于 0.5 米的表面，则可以考虑使用 D415、D455 和 D457 型号。与 D435 型号相比，D415 具有更好的图像质量和更少的深度噪点。
D455 能够在更远的范围内进行深度感应，并且在距离上的精度是 D435 型号的 2 倍。
D457 支持快速 GMSL 类型连接而不是 USB，并具有 IP65 防水防尘保护。 它也有一个用于在 USB 模式下使用相机的开关，它被检测为 D455 并像一个一样运行，但增加了环境保护。
Please sign in to leave a comment.
Comments
2 comments