Background knowledge of machine vision

What is camera

We all have seen varieties of cameras，such as:

So what is camera?In the final reckoning,camera is an equipment which can turn optical signal into electrical signal.In machine vision,the simplest camera model is pinhole camera model.

Pinhole camera model is an ideal camera model without thinking about field curvature and aberration existing in camera.However,these problems can be solved by importing aberration parameter in actual use so that pinhole camera model is still the most widely used camera model.

Through camera,image is taken in a sensor,photosensitive chip can turn information like wavelength and intensity in to digital signal to be distinguished by computer(digital circuit), sensor looks like this:

（Quadrate element in the middle is sensor）

What’s pixel and resolving capability?

Sensor makes up of a great many of points.Take 640*480 as an example,every point is a pixel,collect every point’s pixel and arrange them making a picture.In that way,the resolving capability of this is 640*480.

What is frame per second(FPS)

Frame per second(FPS) is the picture quantity solved per second.If more than 20 frame,human eyes can barely distinguish unsmooth. As it should be,the higher frame rate,the better.

The highest frame rate of Openmv compares:
Note:image without label is not transmission image to IDE because this process is waste of time.

What is color?

Phycially ,color is electromagnetic wave with different wavelength.


But according to human eye’s visual effect,you can describe color of visible light due to color gamut RGB,CMYK,HSB,LAB.

RGB tricolor

The theory of tricolor is not physically but thanks to human’s physiology.There are a few varieties tapered photosensory cells,respectively most sensitive to yellowish green,green and royal purple(or call it violet)(wavelength is 564,534 and 420 nm relatively)

Thus RGB is usually used on display screen to display pictures.

LAB Luminosity-contrast

In Lab color space,L refers to luminosity,a’s positive number refers to red,negative number refers to green, the positive number of B stands for yellow, and the negative end represents blue.Unlike color space RGB and CMYK,Lab color is designed to be close to human vision.

So a measure of L can adjust luminosity.Amend the color gradation to export of the measure of a and b to balance color exactly.

Note: in search of color blobs of OpenMV’S arithmetic,use this model LAB!

Choose to light source

If your machine is industrial,or equipment running for twenty four hours,keeping a stable light source is vital,especially on color arithmetic.when the luminosity changes,the value of the whole color will change a lot.

waiting to add

Focus of lens

Because image is optical refraction through lens to sensor.Thus lens decides the whole image’s size and distance,One important parameter is focus.

Focus of lens refers to the distance from pricipal point to focus,which is a vital performance index.Focus of lens’ length decided the image’s size ,size of viewing angle,size of depth of field and perspectivity of image.While shooting the same target from the same distance,the bigger image produced by the long focus of lens,the smaller image would be produced by the short focus of lens.Pay attention to that the bigger focus,the smaller visual angle.

Another point is aberration of lens.Because of optical theory,in the different position of sensor,different distance from lens,simply it is when anear is big while when distant is small.Therefore there will be fish-eye effect (barrel distortion) on the brink.To solve this problem,use arithmetic to remedy aberration in the code.
Note:In OpenMV,use image.lens_corr(1.8) to remedy lens whose focus is 2.8mm.Furthermore,you can use non-aberration lens directly.Non-aberration camera add extra part to remedy lens,gradually the price will be high.

In the following is comparative image when OpenMV is approximately 20cm far away from the desk produced by different focus lens.

Filter of lens

Normally,there is a filter on the lens.

What’s this filter for?

As is known to all of us,different color has different wavelength.In normal circumstance,besides visible light,there are a deal of infrared light which is exactly the kind of light used in night-viewing.

However,in normal color application,no need for infrared light.Because infrared light also makes sensor response then whiten the whole frame.In that case,we place an optical filter inside 650 nm stopping infrared light.