Do you want to combine OpenCV and MoveIt! to make your robot detect and pick up an object? ROS has you covered, providing seamless integration with these and other popular open source projects, along with a message passing system that allows you to easily swap out different data sources, from live sensors to log files.
Gazebo is a 3D indoor and outdoor multi-robot simulator, complete with dynamic and kinematic physics, and a pluggable physics engine. Integration between ROS and Gazebo is provided by a set of Gazebo plugins that support many existing robots and sensors. Because the plugins present the same message interface as the rest of the ROS ecosystem, you can write ROS nodes that are compatible with simulation, logged data, and hardware. You can develop your application in simulation and then deploy to the physical robot with little or no changes in your code.
OpenCV is the premier computer vision library, used in academia and in products around the world. OpenCV provides many common computer vision algorithms and utilities that you can use and build upon. ROS provides tight integration with OpenCV, allowing users to easily feed data published by cameras of various types into OpenCV algorithms, such as segmentation and tracking. ROS builds on OpenCV to provide libraries such as image_pipeline, which can be used for camera calibration, monocular image processing, stereo image processing, and depth image processing. If your robot has cameras connected through USB, Firewire, or Ethernet, ROS and OpenCV will make your life easier.
PCL, the Point Cloud Library, is a perception library focused on the manipulation and processing of three-dimensional data and depth images. PCL provides many point cloud algorithms, including filtering, feature detection, registration, kd-trees, octrees, sample consensus, and more. If you are working with a three-dimensional sensor like the Microsoft Kinect or a scanning laser, then PCL and ROS will help you collect, transform, process, visualize, and act upon that rich 3D data.
MoveIt! is a motion planning library that offers efficient, well-tested implementations of state of the art planning algorithms that have been used on a wide variety of robots, from simple wheeled platforms to walking humanoids. Whether you want to apply an existing algorithm or develop your own approach, MoveIt! has what you need for motion planning. Through its integration with ROS, MoveIt! can be used with any ROS-supported robot. Planning data can be visualized with rviz and rqt plugins, and plans can be executed via the ROS control system.