The double visual test platform will include both software and hardware.
Here is just this part of the software.
For both visual visual, there is a camera labeled, find a matching point, and finally the algorithm of the double visual visual. In order to verify the algorithm of the double visual visual, it is of course to build a software platform, which is the real-time capture video stream, and then the real-time use of binocular visual algorithm is handled, and the specific result is obtained (generally used for ranging, " For example, a moving object distance, etc.).
Because of the future, the algorithm in the double visual visual will be ported to another platform, so this design of this framework for binocular visual test procedures should also take this.
This is a dialog-based program. The entire program is roughly divided into three parts. One part is a capture of two USB camera video streams. Part of the algorithm part of the double visual visual. There is also part of the synchronous control for multi-threading. In addition to the automatically generated class except for the dialog, three classes appear in this program. Two classes of which are used for video stream images for the camera (one is the establishment of Filter Graph, one is used for callback settings at the screenshot), one class for implementing binocular visual algorithms.
During the writing of this program, DirectShow is used on the image capture of the camera. I can write this program in a week. I feel that I can't imagine it. Maybe there will be a brave in pressure, maybe there is a previous accumulation. The writing of this program framework is that the two cameras simultaneously capture images, real-time processing, and there is a need to consider convenient to transplant and algorithms, and design a lot of things on the program framework, on the framework of the program, I I think, I don't do it well, although it is now convenient to make the algorithm and transplantation. I hope to take a good look at the books in the design pattern in the future.
When using the DirectShow to write the camera to capture the video stream and intercept the image, refer to a lot of Microsoft's example programs. Now this program has a small problem, or you can't get the frame rate of image processing. At this point, after this week, I will look at it, it should be solved. This week should mainly see the algorithm of image processing, otherwise, maybe complete the task on time.
Also, because the image processing and computer visual algorithm have not been done, but the test of the framework of this program is still needed, so I will let the BMP picture replace the double visual algorithm, the results of the experiment are successful (Can be saved continuously to save a lot of frames BMP pictures, of course, is in me)
