"Project Combat" hardware platform selection of image processing project I have always been interested in the embedded system, because the image processing will eventually move to portable, mobile. Here we talk about the image processing project hardware platform selection this piece of ideas to exchange, welcome to exchange, criticism.
Construction of a typical image processing system
① camera and Lens-this part belongs to the imaging device, the common vision system is composed of one or more sets of such imaging system, if there are multiple cameras, may be switched from the image card to obtain image data, it may also be synchronized control to obtain the multi-camera channel data. Depending on the application needs the camera may be the output standard monochrome video (RS-170/CCIR), composite signal (Y/C), RGB signal, may also be non-standard progressive scan signal, line scan signal, high-resolution signal, etc. ② Light Source-as a secondary imaging device, the quality of the image can often play a crucial role, a variety of shapes of LED lights, high-frequency fluorescent lamps, fiber-optic halogen lamps are easy to obtain. ③ sensors-usually in the form of optical switches, proximity switches, etc., to determine the location and status of the object under test, and to inform the image sensor for proper collection.④ Image capture card-usually in the form of inserting cards installed in the PC, the main task of the image Grabber is to transfer the camera output image to the computer host. It transforms the analog or digital signal from the camera into a certain format of image data stream, and it can control some parameters of the camera, such as trigger signal, exposure/integration time, shutter speed, etc. Image capture cartoon often different hardware structure for different types of cameras, but also have different bus forms, such as PCI, PCI64, Compact Pci,pc104,isa and so on. ⑤PC platform-Computer is the core of a PC-based vision system, where the processing of image data and most of the control logic, for the detection of types of applications, usually require a higher frequency of the CPU, which can reduce processing time. At the same time, in order to reduce the industrial field electromagnetic, vibration, dust, temperature and other interference, must choose industrial-grade computer.⑥ Vision processing software-machine vision software is used to complete the processing of the input image data, and then through a certain operation to obtain results, the result of this output may be pass/fail signal, coordinate position, string and so on. Common machine vision software in the form of a C + + image Library, ActiveX controls, graphical programming environment, etc., can be specialized functions (such as only for LCD detection, BGA detection, template alignment, etc.), can also be a common purpose (including positioning, measurement, barcode/character recognition, spot detection, etc.). ⑦ Control Unit (includes I/O, motion control, level conversion unit, etc.)-once the visual software completes the image analysis (unless only for monitoring), then communication with the external unit is required to complete the control of the production process. The simple control can directly take advantage of the I/O of some image acquisition card, the relatively complex logic/motion control must rely on the additional programmable logic control unit/motion control card to achieve the necessary action. Second, the image processing project hardware platform selection
Common image hardware platforms and tool chains include 1, PC+MFC+OPENCV (or similar commercial software), 2, ARM+QT+OPENCV (or similar software);
3, dsp+ special algorithm library, 4, fpga+ special algorithm library;It is worth noting that OPENCV as a relatively well-known open source algorithm library, in the DSP and FPGA (http://ec.eepw.com.cn/center/shownews/userid/38503/id/160095) There are implementations for porting.
At the same time, Android as a branch of arm, with the mobile phone has become an indispensable part of people's daily life, so the mobile phone aspect of the image recognition program/application has a very great competitiveness,iOS is a similar. From the price of x86 industrial computer only need hundreds of yuan, small 1000 of the price can get a complete computer
DSP and FPGA Development Board, if you want to be able to use the price of image processing is generally more expensive, at least thousands of yuan. Third, summary of the final choice, or should be based on the needs of the project to choose the platform. All hardware is ultimately supported by the need to implement. For image processing, the general has 1, the main function is single, that is, do not need rich control resources, 2, the algorithm is highly complex, 3, real-time to a certain degree of requirements. So, I think in general, choose X86 Industrial computer +MFC+OPENCV or QT+OPENCV form should be able to solve many problems. Based on this platform to do the work, choose a stable and easy to expand the platform, can support many future jobs.
From for notes (Wiz)
"Project Combat" image processing project hardware selection
