Nowadays, there are a wide variety of machine vision cameras. In addition, there are camera manufacturers that support designs with user-specific functions such as OEM, and they are becoming more and more distant from the TV standard.
In the future, there is no doubt that it will become more intelligent than just outputting captured images, such as incorporating part of image processing into camera functions. How to integrate the functions of the camera, computer, and image processing software? That may be the key in the future.
Machine vision camera functions
Since machine vision cameras are not bound by the standard TV frame, they are equipped with functions devised by each camera manufacturer. Among them, the common functions required for machine vision cameras will be explained.
Progressive scanning and number of pixels
Since the standard camera performs interlaced scanning, the vertical resolution of one frame drops to half of all pixels, but the machine vision camera performs progressive scanning that sequentially scans all pixels of the image pickup element. .. All the pixels in one frame are effective, and a high-resolution image can be obtained. The number of pixels of a machine vision camera is generally as follows.
For CCD cameras, VGA (330,000 pixels), XGA (800,000 pixels), SXGA (1.45 million pixels), UXGA (2 million pixels), 5M (5 million pixels), 11M (11 million pixels). In addition to the above, CMOS cameras include 4M (4 million pixels) and 12M (12 million pixels).
For CCD cameras, up to 5M (16fps) is a general-purpose practical range, and for 11M, the frame rate is slow at 5fps and the price is high, so it is not widely used. On the other hand, CMOS cameras have a high frame rate of 180fps at 4M and 60fps at 12M, and are reasonably priced, so it can be said that they are promising cameras in the future.
Electronic shutter function
The electronic shutter function is also installed in consumer cameras, but in the case of machine vision cameras, the shutter time can be controlled from the outside. This is the difference from the consumer product. There is a way to select the shutter time preset inside the camera, and a way to decide the shutter time from the outside as needed.
External trigger function
The machine vision camera requires an “asynchronous reset” that inputs a pulse signal from the outside and starts exposure at that timing. If you want to capture a fast-moving subject in the center of the image sensor, a trigger pulse is applied from the outside even during the V (vertical) synchronization period, but at this time, the camera temporarily deletes the image being scanned. Imaging is started at a new timing.
This operation is an asynchronous reset. The camera starts exposure for the set shutter time at the timing when it accepts the external trigger. In addition, there are settings such as exposure only for the pulse width period of the external trigger, two pulse interval times, and exposure settings.
You can select “Gradation”, which indicates the level of contrast in the output image. Usually 8bit is often used, but if a neutral color level is required, 10bit, 12bit, etc. are used.
In the case of a CCD camera, only a part of continuous horizontal scan lines can be partially read out. Since unnecessary scanning lines are thrown away, the time required for one frame is shortened and the frame rate is increased.
In the case of a CMOS camera, only a part of the vertical and horizontal areas (windows) can be read, so the scanning time is further shortened. In a high-pixel CMOS camera, a method is adopted in which only the necessary area is cut out in a wide imaging range and the image processing is completed quickly.
The “gamma characteristics” that represent the ratio of brightness to output voltage can be changed. The gamma value is usually set to 1 (brightness and output voltage are directly proportional), but this value is changed to brighten and emphasize dark areas.
FPGA (Field Programmable Gate Array)
There is also a camera equipped with an FPGA that is open to the user. When you want to build a function other than the above on the user side, you can programmatically upgrade the function. For example, by constructing a frame memory or performing preprocessing for machine vision, the burden on the computer can be reduced.
Mechanical structure of machine vision camera
From the conventional flow, many machine vision cameras are provided with tripod screw holes. Since the tripod screw is an inch screw, it is often inconvenient because there is no screw nearby when attaching it to a stand. Therefore, many millimeter-screw taps such as M4 are provided in the camera housing.
Furthermore, in machine vision applications, the camera itself may be rotated 90 ° or 180 ° and mounted, and there are cameras designed to maintain the center of the image sensor regardless of the posture in which it is mounted.