消费级深度相机详细参数对比

本文对比了多种3D深度摄像头的技术规格,包括KinectV1/V2、RealSense系列、LeapMotion等产品,在价格、测量原理、有效范围、视场角、精度等方面进行了详细比较。

摘要生成于 C知道 ,由 DeepSeek-R1 满血版支持, 前往体验 >

 Kinect V1Kinect V2RealSense F200RealSense R200RealSense SR300RealSense D415RealSense D435Leap MotionOrbbec AstraOrbeec PerseeOccipital StructureZED Stereo CameraDUO MLXPercipio.xyz
Companyhttp://www.k4w.cn/news/8.html http://www.k4w.cn/news/1.htmlhttps://software.intel.com/zh-cn/realsense/previoushttps://software.intel.com/zh-cn/realsense/previoushttps://software.intel.com/zh-cn/realsense/previoushttps://ark.intel.com/zh-cn/compare/92256,92329,128256,128255https://ark.intel.com/zh-cn/compare/92256,92329,128256,128255https://www.leapmotion.com/product/desktop#108https://orbbec3d.com/https://orbbec3d.com/https://structure.io/https://www.stereolabs.com/zed/https://duo3d.com/http://www.percipio.xyz/
ReleasedJun-11Jul-14Jan-15Sep-15Mar-16Jan-18Jan-18Oct-12Sep-15Dec-16Feb-14May-15May-13Jul-16
Price $100.00$99.00$99.00$150.00$149.00$179.00$100.00$150.00$240.00$499.00$449.00$695.00$245.00
Depth MeasureIR+structured
light
TOFIR+structured
 light
Dual IR+
structured light
IR+structured
 light
Active IR StereoActive IR StereoDual IRIR+structured
 light
IR+structured
 light
IR+structured
 light
Stereo RGB
cameras
Dual IRDual IR+
structured light
Range0.4m-4m0.4m-4.5m0.2m-1.2m0.5m-3.5m0.3m-2m0.16-10m0.11-10m0.025m-0.6m0.6m-8m0.6m-8m0.4m-3.5m0.5m-20m0.3m-2.4m0.25m-2m
or 0.5m-10m
Field of
 Vision
57°H×43°V70°H×60°V73°H×59°V59°H×46°V73°H×59°V69.4×42.5(+/- 3°)69.4×42.5 (+/- 3°)135°60°H×49.5°V60°H×49.5°V58°H×45°V110°170°
Distortion<3%
60°H×46°V
Precision          0.5mm at 40cm(0.15%)
30mm at 3m(1%)
120mm baseline,
6 axis Pose Accuracy
Posion:+/-1mm
Orientation:0.1°
baseline 30.0mm<1%
z:5mm@1m
x,y:10mm
RGB Image640×480,30 FPS1920×1080,30 FPS1920×1080,30 FPS1920×1080,30 FPS1920×1080,30 FPSUp to 1920×1080Up to 1920×1080 1280×960,7 FPS
640×480,30 FPS
320×240,30 FPS
1280×720,30 FPS
640×480,30 FPS
320×240,30 FPS
IOS Camera resolution1344×376,100 FPS
2560×720,60 FPS
3840×1080,30 FPS
4416×1242,15 FPS
 2592×1944
1120×920 
560×460
Depth Image320×240,30 FPS512×424,30 FPS640×480,30 FPS640×480,60 FPS640×480,30 FPSUp to 1280×720
Up to 90 FPS
Up to 1280×720
Up to 90 FPS
20 to 200+ FPS640×480,30 FPS
320×240,30 FPS
160×240,30 FPS
640×480,30 FPS
320×240,30 FPS
640×480,30 FPS
320×240,60 FPS
1344×376,100 FPS
2560×720,60 FPS
3840×1080,30 FPS
4416×1242,15 FPS
configurable between
752×480,56 FPS and
320×120,320 FPS
1120×920 
560×460
ConnectivityUSB 2.0USB 3.0USB 3.0USB 3.0USB 3.0USB 3 Type-CUSB 3 Type-CUSB 2.0USB 2.0USB 2.0Lighting on IOS,
USB elsewhere
USB 3.0USB 2.0USB 2.0
Phsical
 Dimensions
280×64×38mm250×66×67mm110×12.5×3.75mm101.6×9.6×3.8mm110×12.5×3.75mm99×20×23mm90×25×25mm76×30×17mm165×30×40mm172×63×56mm119.2×28×29mm175×30×33mm52×25×13mm124×24×64mm
Works
Outdoors?
××××××××
Skeleton
 tracking?
√(two skeletons)√(two skeletons)×××  ×√(only hand
positions)
    
Facial
tracking?
  ×××    
3D scaning?  ×××3rd party
Gesture
Tracking?
×(only via
 thirdparttools)
√(Visual Gesture Builder)×××  ×××××3rd party
Gesture
Detection?
√(hand grip,
release,press,
scroll)
√(hand open,closed,lasso)×  ××××3rd party
Toolkits?OpenNIDense3D,OpenCV,Qt5WPF,Cinder,OpenFrameworks,JavaScript,vvvv,
Processing,Unity3D,more
Java,JavaScript,
Processing,Unity3D,
Cinder
Java,JavaScript,
Processing,Unity3D,Cinder
  WPF,Cinder,OpenFrameworks,
JavaScript,vvvv,
Processing,
Unity3D,more
OpenNIC++,Java,OpenNI,ROSIOS,Unity3D,OpenNIJava,JavaScript,Processing,Unity3D,CinderJavaScript,OculusRift,Unity3D,UnrealOpenNI and native SDK for Win,Linux,Android
Project
Examples
Many examples of
skeleton tracking,
face traching,and
speech detection
on a variety of
different platforms
and frameworks
Many examples of
skeleton tracking,
face traching,and
speech detection
on a variety of
different platforms
and frameworks
Many examples of
face traching,gesture traching,
speech detection
on a variety of
different platforms
and frameworks
Various face tracking
examples,mostly with C++,
only one Unity3D sample
Various face tracking
examples,mostly with C++,
only one Unity3D sample
  Various examples available for each of the lanugaes and platforms supportedHandViewer,Depth Date viewer,
RGB Date Viewer
Depth Date viewer,
RGB Date Viewer
Lots of examples on
3D scanning,mixed reality and indoor navigation
Background subtraction,
right image disparity,
depth map
Very few samples
in each of the
supported languages,
mostly to get raw image and depth data
Moving Robotics,
Industrial 3D measurement,3D scaning
### 深度相机与双目相机在三维坐标测量中的对比 #### 测量原理差异 深度相机通过发射并接收红外光或其他形式的结构化光线来计算物体距离,能够直接获取场景中各点的距离信息。而双目立体视觉则模仿人类双眼的工作方式,利用两个摄像头拍摄同一景物的不同视角图像,并基于视差原理推算出空间位置数据[^1]。 #### 准确性和分辨率 对于精度需求较高的应用场合而言,高质量的深度传感器可以在近距离内提供非常精确的距离读数;然而随着目标离设备越来越远,其误差也会相应增大。相比之下,在理想条件下工作的双目系统可以保持较为稳定的测距准确性,尤其是在处理较大范围内的对象时表现良好。不过这依赖于良好的特征匹配算法以及足够的纹理信息支持。 #### 计算复杂度和实时性能 由于深度摄像机通常内置专门硬件用于快速解析深度图,因此能够在较低功耗下实现高效的帧率输出。相反地,构建有效的双目匹配模型往往涉及到复杂的数学运算过程,这对处理器提出了更高要求,尽管如此现代GPU加速技术也在一定程度上缓解了这一挑战。 #### 成本因素考量 一般来说,工业高精度深度感应模块价格昂贵,而且可能受到专利保护限制某些特定功能的应用开发。与此同时,普通的RGB-D消费类产品虽然成本低廉但是性能参差不齐。另一方面,由一对普通CMOS/CCD组成的双目装置相对容易获得,并且可以根据具体项目定制参数设置以满足不同应用场景的需求。 ```python import cv2 import numpy as np # 假设这是来自深度相机的数据 depth_image = ... # 加载深度图像 (单位: 米) # 对应像素处的真实世界Z轴坐标可以直接从深度图像中读取 z_depth_camera = depth_image[y, x] # 而对于双目相机来说,则需要先找到左右图片之间的对应关系(即视差) left_img = ... right_img = ... stereo_bm = cv2.StereoBM_create(numDisparities=16, blockSize=15) disparity_map = stereo_bm.compute(left_img, right_img).astype(np.float32)/16. focal_length_in_pixels = ... # 已知焦距转换成像素值 baseline_distance_meters = ... # 双目基线长度(两镜头中心间距) # 使用三角公式计算实际距离 z_stereo_camera = baseline_distance_meters * focal_length_in_pixels / disparity_map[y,x] ```
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值