Automotive thermal imaging and car electronics manufacturer since 2006 · Shenzhen, China
Buyer guide

Car thermal camera resolution: 256 vs 384 vs 640

At the same 9.1 mm lens, a higher thermal resolution gives a wider field of view: 18° for 256 × 192, 29° for 384 × 288 and 45° for 640 × 512. The narrow 256 camera has the longest listed AI range on straight roads; the 640 camera sees road edges and junctions with more detail.

Resolution and lens decide what the camera sees

A thermal camera’s resolution is the number of pixels on its detector. The lens decides how wide a slice of the road those pixels cover. Keep the lens fixed and raise the resolution, and the camera sees a wider scene. Keep the scene fixed and raise the resolution, and each target gets more pixels. Buyers often compare resolution alone; the useful comparison is resolution and field of view together.

All six standard configurations use a 9.1 mm lens. That makes the trade-off easy to see:

CameraModelsField of viewAI personAI vehicleAI animal
256 × 192L5P-2 / L7P-218° × 13°100 m200 m60 m
384 × 288L5P-3 / L7P-329° × 22°90 m190 m55 m
640 × 512L5P-6 / L7P-645° × 37°80 m180 m50 m

Reference values from the product sheet for targets: person taller than 1.7 m, vehicle wider than 2.3 m, animal taller than 1.5 m. Real-world range depends on weather, thermal contrast, mounting and sensitivity settings. Not a guaranteed warning distance.

Why the lowest resolution has the longest listed range

The 256 × 192 camera spreads 256 horizontal pixels across 18°. The 640 × 512 camera spreads 640 pixels across 45°. Per degree, they are close: about 14 pixels per degree for both. So a pedestrian 80 m ahead covers a similar number of pixels on either camera, and the AI needs enough pixels to recognise a person. The narrow camera concentrates its pixels straight ahead, which is where the longest listed recognition distance comes from.

The wide camera’s advantage is everything off-centre: the road shoulder, a deer stepping out from the verge, a cyclist at a junction, the inside of a bend. On a 256 camera those are outside the frame.

Detection, recognition and identification (DRI)

Thermal camera makers describe range with three distances based on the Johnson criteria: how many pixels must fall across a target before a viewer can detect it, recognise what it is, or identify details. Different makers use slightly different pixel counts, so ask which ones a datasheet uses. A common set is about 2 pixels for detection, 8 for recognition and 13 for identification.

With a 12 μm detector and a 9.1 mm lens, one pixel covers about 1.32 milliradians, or 13 cm at 100 m. That is the same for 256 × 192, 384 × 288 and 640 × 512, so all three give the same DRI distances straight ahead:

LevelPixels across targetWhat the driver can tellPerson, 1.7 m tallVehicle, 2.3 m wide
Detection2Something warm is thereabout 645 mabout 870 m
Recognition8It is a person, an animal or a vehicleabout 160 mabout 220 m
Identification13Details such as posture or direction of travelabout 100 mabout 135 m

Geometric estimates for a 9.1 mm lens in clear conditions, using target height for a person and width for a vehicle. They are not product specifications and not warning distances.

The AI needs more pixels than a human viewer: the listed AI pedestrian range of 80 to 100 m corresponds to roughly 13 to 16 pixels across a 1.7 m person, about the identification level. So a datasheet that quotes a detection distance of several hundred metres is describing a warm dot on the screen, not an alert. When you compare suppliers, compare like with like: AI range with AI range, and DRI with DRI at stated pixel counts.

Compare resolutions at the same field of view

Many comparisons of 384 and 640 cameras use different lenses, which makes the 640 look better or worse than it is. There are two fair comparisons:

  • Same lens (9.1 mm). The 640 × 512 camera sees a wider scene (45° against 29°) with the same detail per degree and the same DRI distances straight ahead. You gain coverage of the road edges, not range.
  • Same field of view. To give a 640 × 512 detector the same 29° view as the 384 camera, the lens must be about 15.2 mm. Each pixel then covers 0.79 milliradians, so every DRI distance grows by about two thirds: person recognition moves from about 160 m to about 270 m.
SetupHorizontal viewPixel footprint at 100 mPerson recognitionVehicle recognition
384 × 288, 9.1 mm29°13 cmabout 160 mabout 220 m
640 × 512, 9.1 mm45°13 cmabout 160 mabout 220 m
640 × 512, about 15.2 mm29°8 cmabout 270 mabout 365 m

Geometric estimates, 8 pixels across the target, clear conditions. Our standard systems use the 9.1 mm lens; other focal lengths go through an engineering review.

The practical rule: with the same lens, choose resolution for how wide you need to see; to see further with the same width, you need more pixels and a longer lens together.

Which configuration to stock

  • 256 × 192 (L5P-2 / L7P-2): long, straight rural roads and highways where the hazard is far ahead. Lowest cost of the three.
  • 384 × 288 (L5P-3 / L7P-3): the balanced middle option, 29° wide. A sensible default when you stock one SKU.
  • 640 × 512 (L5P-6 / L7P-6): winding roads, junctions, trucks that need to see the shoulder, and buyers who want the most image detail on screen.

Many distributors start with two SKUs, one narrow and one wide, and let installers recommend by route.

Display size is a separate choice

The 5-inch and 7-inch hosts share the same 1024 × 600 resolution and the same AI. Pick the screen by where it is mounted and how far the driver sits from it. A 7-inch screen on a truck dashboard is easier to glance at; a 5-inch screen fits compact cars without blocking vents. See the 5-inch kit and 7-inch system.

What ranges on a datasheet do and do not tell you

Ranges are measured on targets of a stated size in good conditions. Fog, heavy rain, a warm night with little temperature contrast, a dirty window or a low mounting position all shorten them. The reliable test is your own vehicles on your own roads, which is why we recommend an evaluation kit before stocking.

Quick answers

Is 640 × 512 always better than 256 × 192 for a car?

Not always. With the same 9.1 mm lens, 640 × 512 covers a much wider view (45° vs 18°), so each target uses a similar number of pixels at distance. That is why its listed AI pedestrian range (80 m) is shorter than the 256 camera’s (100 m). It wins on road edges, bends and junctions.

Does a 7-inch screen improve thermal detail?

No. Both screens are 1024 × 600. Thermal detail comes from the camera resolution and lens.

Which resolution suits trucks?

Trucks often benefit from a wider view of the road shoulder, so 384 × 288 or 640 × 512 on a 7-inch host is a common choice. For long straight highways, 256 × 192 keeps the longest listed range.

What does NETD mean on a thermal camera?

Noise-equivalent temperature difference: the smallest temperature difference the sensor can separate from noise. Lower is better. These cameras list ≤ 40 mK at 25 °C.

Tell us your market and vehicles. We will recommend a configuration.

Share your company, target market, vehicle types and expected quantity. We reply with a configuration, sample terms and pricing.