How to correctly recognize robot positioning technology (2)

Let's continue the discussion on how to properly recognize robotics

Second, machine vision positioning

Machine vision positioning includes two types: two-dimensional positioning and three-dimensional positioning. In China, there are relatively few 3D positioning own technologies, and there are very few patent applications for this technology. But abroad, taking companies such as Microsoft, Google, and Apple as examples, they have already laid out areas such as 3D information detection and 3D space reconstruction.

1.Three-dimensional attitude tracking

Microsoft launched the Kinect somatosensory gaming device in April 2010, which is a peripheral device for the Xbox 360 console. The main function is to obtain the body movements of the player's entire body to operate the game on the Xbox.

1441783722958570.png

The left side of the device is an infrared light transmitter, which is an optical coding structured light technology; the center is a color CMOS camera with a resolution of 640 * 480; the right side is an infrared CMOS camera with a resolution of 320 * 240. An important technique for acquiring three-dimensional movements of players is a binocular stereo vision technology. Relying on the calibrated baseline distance, focal length, and parallax data between the dual cameras, and then calculating the depth information based on the infrared bitmaps collected by the infrared camera.

However, the measurement accuracy announced by Kinect is: when measuring a range of 2 meters, the theoretical accuracy can reach 1 cm (the actual measurement by a third party agency indicates that the accuracy cannot be achieved). Because the device is geared towards the entertainment industry, the requirements for accuracy and consistency of test results are low. Of course, Microsoft's software algorithm is robust, mainly because its hardware uses low-cost CMOS sensors. Compared with the binocular stereo vision equipment MV-VS220, which is developed by Wei Shi Imaging Company for industrial applications, Wei Shi Imaging is better in terms of hardware configuration. It uses a 1280 * 960 resolution CCD camera, which can sense visible light and near-infrared light at the same time. When the standard configuration is in the range of 0.8 meters, the repeatable measurement accuracy can reach 0.6mm.

2.Three-dimensional spatial positioning

Having seen Microsoft's Kinect device, let's look at the three-dimensional spatial positioning technology represented by KUKA robots. The recent "Robot vs. Ping-Pong Champion" video released by KUKA Robot is quite popular. For the time being, let's not talk about the speed and accuracy of the robot itself. Here we mainly talk about the three-dimensional spatial positioning technology for positioning the table tennis for the robot. As shown below

:

1441783751363973.png

1441783767504705.png

1441783778604964.png

In fact, the basic technology of this application is not particularly cutting-edge, and belongs to the more conventional binocular 3D positioning technology. Due to the high speed of table tennis during exercise, it can reach 3 meters per second, and the measured object is a high-speed moving target. Therefore, you can use a high-speed camera to capture and then use the frame difference method to obtain the measured target and then filter out the background image (because In the field of view of the camera, the speed of movement of other objects except for table tennis is low. Only between two consecutive frames of high-speed cameras, only the position of the table tennis changes). In this way, the objects to be measured in the left and right cameras have only one point (simplifying the table tennis to one point), and the speed of stereo matching and three-dimensional attitude measurement can be fast. Able to keep up with table tennis.

Of course, since the above products and equipment have been officially put into commercial operation, it is impossible for manufacturers to publish specific technology implementations. We can only guess from the results. However, we can obtain a lot of useful information from the products launched by some non-profit laboratories or college laboratory equipment providers. In the case of the CCAS binocular vision system launched by Wei Shi Images, the official website has a detailed introduction to the principle of the system.