Selecting the correct machine vision lens is a science. There are many factors to consider. Some key lens criteria include:

Contrast: Delivering light contrast to the sensor

Resolution: Delivering object detail to the sensor

Depth of Field: Depth range in which the object will be in focus

Field of View: Area imaged by a lens

Working Distance: Distance between the lens face and the object

Distortion: The many aberrations that modify your otherwise perfect image

Mount: Most cameras are compatible with C-mount lenses, but be sure to confirm

Size: Most cameras are compatible with 2/3 size (11 mm diagonal) lenses, but some require a lens having a larger 'optical circle'.

Cost (No point in spending more than necessary to get the required performance)

Most of these criteria are intrinsically connected, and there is no value in considering one outside the context of the rest. Lens selection is often a compromise between many competing factors. And it is all too easy to limit your camera's performance by using the wrong lens.

So, how do you select the appropriate lens for your application? First consider the general type of lens you need. You may use the table below as a rough guide.

Object Width Lens Type Calculator

> 12 mm General Purpose Lens Entocentric Lens Calculator

1 - 25 mm Macro Lens Macro Lens Calculator

< 1 mm Microscopic Lens

Precision Gauging Telecentric Lens Telecentric Lens Calculator

1 - 240 mm

The basic considerations are the size of the camera chip and the field of view required. These two will give you the lens required.

As for the MTF: once sensor (size) and lens are defined, the final component defining the MTF will be the resolution (pixel count) of the sensor. Higher resolution sensor chips will give a "better" MTF, but will require way more processing power to use it. Thus sensor resolution will always be some sort of compromise.

Considering the contrast: increased contrast usually helps the algorithms. But contrast is not only a function of lens and sensor: an important role is the one of illumination. But illumination is very application-specific, And you may consider an appropriate illumination the first stage of image processing.

An example? I once saw an application reading shallow lasered barcodes from titanium parts: while the barcode was hardly discernible by the naked eye, it showed up as nearly black-and-white in the sensor image - - - due to an illumination only grazing the surface, illuminating the surface while leaving the grooves in the dark.

U. Dreher thank you very much for your answer.

Sun is the illuminator of my project (outdoor application), so I can't control the illumination problem.

For the contrast problem, I think that I must select a reliable value for MTF in the appropriate spatial frequency and determine the pixel size of the sensor from the selected spatial frequency (according to Nyquist sampling, spatial frequency = 1/(2* pixel size)).

Am I correct?

What is reliable value for MTF?

The sun is the most unreliable illuminator, so skip any contrast discussion!

As about MFT/spatial frequency: Nyquist gives a minimum bandwidth, which is not really practical. Make the smallest features to be discerned at least 10 pixels wide/high and you should succeed. Anything smaller than that might give unreliable/unstable results.

U. Dreher, You mean the minimum dimension of the target occupies 10 pixels of the sensor?

If I am correct, Is pixel size important?

How to determine the required pixel size?

Mahdi Mansouri Maybe you misunderstood me, maybe we're talking about different things: the "minimum feature" to be discerned should not be smaller than 10 pixels h/w - the target itself might require more pixels. Depends.

To clarify it for me: what would be your scene and what would be the targeet ?

Pixel size itself is not important - provided lens and sensor are matched: for a given pixel number, a larger sensor (aka larger pixel area) requires a lens with a larger focal length than a smaller sensor of the same number of pixels.