An automotive camera is a camera device installed on a vehicle, used mainly to capture images of the surrounding environment and support driving and safety. In conventional cars the term meant only a reversing camera, but as vehicles have become more intelligent and electrified, cameras have been given many new functional requirements and their applications have widened, from dash cameras to lane change assist, parking assist, intelligent driving and in-cabin monitoring.

Classification of automotive cameras

Scope

For cameras with very different functions, the widely accepted classification is by mounting position and application: front view, rear view, surround view, side view and in-cabin monitoring. Their mounting positions and names are shown in the figure below.

The NTEK optical laboratory

With the rapid development of autonomous driving, automotive cameras have become a key part of the perception system, capturing image or video data of the surrounding environment. That data is processed by the autonomous driving system to recognise and detect road signs, lane markings, pedestrians and vehicles, providing the information the decision system needs. Because cameras are so important, testing their reliability and optical performance to assure product quality has also become essential. NTEK Group has introduced a full set of test capabilities for automotive cameras and exterior mirrors and built an optical laboratory to the requirements of the standards, supporting driving safety.

The layout of the optical laboratory is shown below.

Typical test equipment:

Optical indicators for automotive cameras

According to QC/T 1128, Automotive Cameras, there are 16 optical indicators for automotive cameras:

Frame rate
Effective pixels
Field of view (FOV)
Resolution, MTF (modulation transfer function)
Signal to noise ratio (SNR)
Dynamic range
Maximum illumination
Minimum illumination
Optical axis centre precision
Automatic gain control
White balance
Startup time
System delay
Colour rendition
Flare
Ghost

MTF and the derived SFR value

Because there are so many parameters, they are covered in several parts. This part focuses on one of the most important indicators, the MTF value and the SFR value derived from it.

Before explaining how to read an MTF value or curve, a few basic concepts need to be clear.

1. Resolution, also called resolving power, is the ability of a lens to reproduce the detail of a subject clearly. The higher the resolution, the clearer and finer the image. It is expressed in line pairs per millimetre (lp/mm) or line widths per picture height (LW/PH).

2. Contrast, also called sharpness, is the ability of a lens to reproduce clearly the highlights, mid tones, shadows, texture detail and brightness differences of a subject. A high contrast lens produces crisp outlines, sharp edges and clean tonal separation; a low contrast lens produces softer edges and outlines and poorer rendering of light and dark contrast. Resolution and contrast are the two main indicators of a photographic lens.

3. Radial and tangential. The black and white lines in a resolution target or grating in the image field should be placed in two principal directions, which are the standard directions for testing an optical imaging system: radial, parallel to the radius of the image circle, and tangential, parallel to the tangent of the image circle.

4. Spatial frequency is expressed in line pairs per millimetre (lp/mm) or line widths per picture height (LW/PH). Spatial frequency is the horizontal axis of the MTF curve; lp/mm is the number of black and white line pairs a camera can resolve within one millimetre, and the other units have a similar meaning.

The MTF value is an important parameter for measuring lens resolving power. It is defined as the ratio of the modulation M of the original image to the modulation M' of the image formed by the lens: MTF = M'/M.

Because M' is smaller than M, MTF is less than one. The higher the MTF, the stronger the resolving power and the better the rendering of detail. Modulation is defined as M = (Imax - Imin)/(Imax + Imin), where Imax is maximum brightness and Imin is minimum brightness. For example, if Imax is 100 and Imin is 1, then M = 99/101 = 0.98.

MTF depends on spatial frequency, and the curve of MTF against increasing spatial frequency is the MTF curve. In practice, the closer the MTF curve is to the ideal, that is the higher the MTF stays from the origin out to high spatial frequencies, the better the imaging quality. Conversely, if the MTF falls quickly at high spatial frequencies, resolving power there is poor.

If the image formed by a lens were an exact copy of the original black and white lines in both resolution and sharpness, the lens would have perfect resolving power. Such a perfect lens does not exist, and a normal lens behaves as follows.

There are two losses. First, sharpness falls and the black to white boundary becomes blurred, partly because of light wave diffraction. Second, very dense black and white lines become indistinguishable, a result of reduced sharpness in dense patterns and therefore also related to diffraction. The loss of sharpness or contrast ultimately limits lens resolution.

Even a perfect lens is limited by diffraction. The edge of a shadow in sunlight is never perfectly sharp but always somewhat blurred, because light is an electromagnetic wave and electromagnetic waves diffract, deviating from straight line propagation at an object edge. A lens has a finite aperture, and light passing the edge of the aperture also diffracts, blurring the black to white boundary. The higher the line frequency, the greater the effect of diffraction, until the lines can no longer be resolved.

The lowest contrast the human eye can detect is 0.05, so the spatial frequency at which MTF equals 0.05 is generally taken as the visual resolution. In the example below, the lens shown by the purple curve resolves 70 lp/mm and the lens shown by the red curve resolves 100 lp/mm.

Calculating MTF from the description above would require the response at many different spatial frequencies and therefore many photographs, which is very inconvenient. To simplify testing, the usual approach is to take a single photograph, calculate the SFR value and derive the MTF curve from it. The MTF values in QC/T 1128 are measured this way.

SFR (spatial frequency response) follows from mathematical theory: a single black and white slanted edge is enough to derive an MTF approximately equal to that at all spatial frequencies. MTF is commonly used for optical systems, while SFR is used for imaging systems, which include an optical system.

SFR measures the effect of increasing line frequency on a single image, and the result is influenced by the lens, the image sensor and the processing pipeline.

SFR does not need line pairs at different spatial frequencies. One black and white slanted edge is enough to derive a response roughly equivalent to all spatial frequencies.

The main steps in calculating SFR are differentiation plus a Fourier transform:

(1) The slanted edge image is oversampled into a finer black and white transition profile, the edge spread function (ESF).
(2) That profile is differentiated to obtain the line spread function (LSF).
(3) An FFT or DFT of the LSF gives the MTF value at each frequency.

Finally, MTF50 and MTF50P are commonly used parameters describing the visual sharpness of an imaging device.

As analysed above, if the imaging device were purely optical, its MTF would peak at zero spatial frequency and fall as spatial frequency rises. MTF50 is the spatial frequency at which MTF falls to 50 percent of its value at the origin, which to some extent indicates the ability of the imaging system to capture detail.

Imaging devices in use today, however, sharpen the image in software after optical imaging, so the MTF peak does not occur at the origin. On the curve this appears as an overshoot and a peak away from the origin, and MTF50P is the spatial frequency at which the peak MTF falls to 50 percent. MTF50P is generally a more useful reference than MTF50. The curve below is a measured MTF curve.

QC/T 1128 test items