Calibration-Under_Different-Resolution

授权协议 Apache-2.0 License
开发语言
所属分类 应用工具、 科研计算工具
软件类型 开源软件
地区 不详
投 递 者 松锐藻
操作系统 跨平台
开源组织
适用人群 未知
 软件概览

Stereo Camera Calibration under Different Resolution

In this project we modified existing Matlab toolbox's code to get the stereo calibration run under different resolution cameras. Especially if you want to work with multiple different camera, you should calibrate them. For instance infrared camera and RBG camera, or as how our demo was on you can calibrate up down mounthed high resolution standart USB camera and relatively low resolution Pan-Tilt-Zoom (PTZ) camera.

Unfortunately neither Matlab's nor OpenCV's does not give us an opportunity to calibrate such a both two very different kind of camera. The following figure shows one of our provided 7 pair of calibration board images from USB and PZT camera. The USB camera has 1920x1080 resolution but PZT has just 720x480.

We slightly modified estimateCameraParameters.m and rectifyStereoImages.m functions of Matlab Toolboxes. You can find them on that repository. Also we provided demo script. You can run it with following command.

> Demo

then you can take following outputs as a results. You can see extrinsic parameter, rectified image pair and position of cameras relative to 1st checkerboard points, their ground truth and also estimated positions from triangulation of both camera.

Here is the K, global intrinsic calibration parameters of USB camera and also R, T matrixes for 1st calibration board.

K1 =  

    1157.0         0    931.2
         0    1147.4    376.0
         0         0    1
R1 =
   -0.9972    0.0177   -0.0732
   -0.0328   -0.9770    0.2108
   -0.0678    0.2126    0.9748

T1 =
   58.6597
  -30.2059
  613.0425

And also here is PZT camera parameters

K2 =
876.9470         0  326.4428
       0  776.9896  196.5821
       0         0    1.0000
       
R2 =
 -0.9997    0.0114   -0.0231
 -0.0137   -0.9944    0.1043
 -0.0218    0.1046    0.9943
 
T2 =
 93.4428
101.1229
584.6069  

Beside that you can get the one image into another image plane (superimpose). But without knowing depth information, theoretically it is not possible to superimpose each pixel in the right place. So let's assume all objects are at Z=0 world plane, where the top-left corner of the first calibration checkerboard is our reference point. Here is the superimposed result of PZT image onto the USB camera image plane.

 相关资料
  • 我有一个项目A,它依赖于我的Eclipse工作区中的项目B和C。 有时我希望A运行时使用来自存储库的工件B和C,而不是工作区中的版本。因此,我选择A,然后选择'Maven->Disable Workspace Resolution',这样A将使用存储库中的版本。 附加信息:这些项目的确切版本在存储库中。事实上,如果我关闭工作区中的项目B和C,项目A会很高兴地使用B和C的存储库版本。

  • 无需克隆: https://jitpack.io/#niklasholtmeyer/maven-EndpointsCoverage pom.xml: 如果我编译这个插件,并根据Commentblock中的所有内容更改groupId/artifactId,它就能正常工作。 我还尝试了以下版本: 1.0.0 主 Master-F2D0242DF8-1 主控-快照 -快照 发布 最新 新建pom.xml

  • AMD FidelityFX Super Resolution (FSR) 是一种开源、高质量的解决方案,用于把较低分辨率的图像输入转变成高分辨率图像输出。 它使用了一系列尖端算法,特别强调创建高质量的边缘,与直接以原始分辨率渲染相比,性能有了很大的提升。FSR 为昂贵的渲染操作(例如硬件光线追踪)提供“实用性能”。 构建说明 前提条件:要构建FSR样本,请遵循以下说明: 1. 安装以下工具:  

  • Coreference Resolution PyTorch 0.4.1 | Python 3.6.5 This repository consists of an efficient, annotated PyTorch reimplementation of the EMNLP paper "End-to-end Neural Coreference Resolution" by Lee et