Sports tracking camera selection: resolution, speed and reach

A camera that shows the whole playing area may leave the ball only a few pixels wide. Increasing resolution helps, but it also raises data throughput and does nothing to freeze motion. For buyers planning player or ball tracking, the useful starting point is measurable detail at the hardest viewing position. Work through coverage, sampling, exposure and timing before choosing an interface or lens.

Start with the tracking decision

Define what the system must report: player positions, ball locations, trajectories or technique measurements. These tasks need different image detail and timing. Detecting a moving object is less demanding than estimating its center accurately through partial occlusion.

Write down the smallest target dimension, the widest scene coverage, the fastest transverse motion and the acceptable measurement error. Also record the nearest and farthest useful positions. A camera looking obliquely across a field has different sampling at those positions; a calculation at the scene center is not sufficient.

For an initial design, specify a target width in pixels rather than a megapixel count. A hypothetical requirement might be at least 12 pixels across a ball in the most distant tracking zone. That is a design assumption to validate with your detector, not a universal detection threshold.

Use the area scan cameras category to identify shutter and interface options, then narrow candidates in the camera finder. Keep the image-processing requirement separate from the available product list until the arithmetic is complete.

Calculate resolution from the smallest target

For a plane approximately perpendicular to the viewing direction, horizontal object-space sampling follows directly from scene width and image width.

Suppose a hypothetical tracking zone is 20 m wide, and the ball diameter is 70 mm. Convert the field width to 20,000 mm before calculating. For a 12-pixel target:

An assumed image width of 4,096 pixels would provide:

This leaves some sampling margin, but not necessarily enough for every tracking algorithm. Defocus, noise, motion blur and partial occlusion can consume that margin. A color filter array also affects the detail available in individual color channels; see color camera spatial resolution.

Do not confuse sampling with measurement accuracy. A detector may estimate a center between pixels under favorable conditions, but that does not guarantee the same accuracy on blurred or partly hidden targets. The related guide to camera resolution for dimensional inspection explains why pixel allocation needs an error budget.

If the distant ball fails the sampling requirement, divide the playing area into overlapping zones before assuming a higher-resolution sensor is the only solution.

Set frame rate and exposure independently

Frame rate controls the time between observations. Exposure controls the motion accumulated inside each observation. Raising frame rate without shortening exposure can produce more blurred frames rather than better trajectories.

Choose an allowable target displacement between frames based on association, occlusion and the motion model. For transverse speed of 40 m/s and an assumed maximum displacement of 0.25 m:

That requirement comes from tracking continuity, not motion sharpness. At 200 frames/s, the same target travels 0.20 m between frames. The software must still associate observations across that gap.

Now calculate exposure from the earlier sampling estimate. If you allow 0.5 pixel of motion blur:

A 200 frames/s sequence has a 5 ms frame interval, but this example needs an exposure near 61 microseconds. Most of each interval is therefore not spent collecting light. Whether the images remain usable depends on illumination, aperture, sensor response and acceptable noise.

For the distinction between these settings, use exposure time versus shutter speed for moving parts. The underlying motion arithmetic also applies to sports tracking.

Compare shutter behavior before choosing a camera

A global shutter exposes the image simultaneously. A rolling shutter exposes different rows at different times, potentially distorting fast-moving targets or complicating multi-view geometry. A short exposure reduces blur within each row but does not remove the time difference between rows.

Shutter type alone does not establish sensitivity, noise, dynamic range or achievable frame rate. Verify those separately at the intended resolution, pixel format and exposure.

The MV-CU020-90GC is a global-shutter color camera with a GigE interface. The MV-CH120-11UC is a global-shutter color camera with a USB3 interface. Their catalog differences provide useful comparison points, but neither should be treated as meeting the worked example without checking its full operating specifications.

All 2 share: Sensor type CMOS, global shutter; Exposure mode Off/Once/Continuous exposure mode; Binning Supports 1 × 1, 2 × 2, 4 × 4; Decimation Supports 1 × 1, 2 × 2, 4 × 4; Lens mount C-mount; Weight Approx. 100 g (0.2 lb.); Certification CE, RoHS, KC; Mono/color Color. Values as printed in each product's data sheet; open a part for its drawing and full specifications.

The calculated requirement is a horizontal pixel count, not merely a total pixel count. Confirm the actual image dimensions and the frame rate available at those dimensions. Also verify whether a proposed region of interest changes readout speed, coverage or both.

Choose the lens from coverage and usable detail

Focal length follows sensor dimensions and field coverage, not the camera's megapixel label. Use the sensor's physical active width rather than treating its nominal optical format as a millimeter measurement.

For a first estimate with a conventional lens:

Assume a hypothetical active sensor width of 12.8 mm, a working distance of 25,000 mm and a field width of 20,000 mm:

This thin-lens estimate is a starting point. Actual coverage depends on lens construction, focus setting and distortion. Distance references also differ between optical calculations and mechanical drawings.

The LS-FFL-0567 has a 16 mm focal length, a 1.1" sensor format and a C-mount. It is a focal-length candidate to investigate, not a confirmed match to the hypothetical sensor used above. Verify active sensor coverage, image quality and mechanical fit with the chosen camera.

Browse machine vision lenses and the lens finder using sensor coverage and mount requirements alongside focal length. The machine vision lens selection guide provides the broader selection workflow.

Check target sharpness at the near edge, center and far edge of the tracking volume. Stopping down can increase depth of field, but it also reduces light and eventually increases diffraction blur. If depth variation is substantial, use the depth-of-field calculator below rather than assuming a small aperture solves every focus problem.

Synchronize exposures, not just frame arrivals

Multi-camera tracking needs observations that represent a sufficiently common instant. Matching frame numbers or software arrival times does not demonstrate simultaneous exposure.

Translate timing error into spatial error before choosing the synchronization method:

An offset of 100 microseconds therefore matters when your error budget is only a few millimeters. A larger position tolerance may permit looser synchronization.

Check trigger support, exposure-start delay, timestamp meaning and timing variation in the camera documentation. Network clock alignment, where supported, does not by itself prove that cameras expose together. For rolling shutters, the relevant row also affects the observation time.

The multi-camera exposure synchronization guide explains shared-trigger planning. The guide to camera hardware trigger frame rate helps distinguish requested triggers from accepted exposures.

Verify timing with a common visible event or suitable measurement equipment. Repeat the test at full acquisition rate, because a camera may reject triggers when its acquisition pipeline is busy.

Budget bandwidth and choose the cable path

Calculate uncompressed image payload before adding protocol overhead, processing copies or storage. For the hypothetical image stream below, assume one byte per pixel and no compression:

Four such streams require about 7.08 GB/s of image payload. This is a system requirement, not a claim that any featured camera produces that stream. Increasing precision or expanding packed data into larger host-side containers raises the memory and storage burden.

Check every shared resource: camera link, host adapter, expansion bus, memory transfers, processing and recording. The guide to camera-to-GPU data transfer addresses transfer-path constraints, while camera frame-rate bottlenecks covers acquisition checks.

Select camera cables by interface, connector, retention and intended routing. Use the accessories finder after documenting the complete cable route, not just the straight-line distance.

The LS-CBL-0219 is a 3 m USB3.0 camera cable specified for high-flex drag-chain use. Confirm connector compatibility and the camera's transport requirements before selecting it.

For longer routes, evaluate documented transport and extension options as a complete system. Do not assume a cable from another interface family is interchangeable, or that a passive cable can be extended indefinitely.

Validate changing illumination and difficult backgrounds

Short exposures make available light a central constraint. Evaluate the scene under the dimmest expected illumination, not just the brightest commissioning conditions. Include bright backgrounds, shadows and rapidly changing light in the image set.

Set exposure limits first, then determine whether aperture and gain provide enough signal. Increasing gain can make the image brighter without increasing the photons collected, so inspect noise and tracking consistency rather than brightness alone.

Highlight clipping can erase target edges, while dark backgrounds can hide them in noise. Histogram inspection helps, but also examine the actual target region. Multi-exposure dynamic-range modes can introduce temporal differences around moving objects; verify their behavior rather than assuming they improve tracking.

The guide to automatic iris, exposure and gain control explains how these controls interact. Any remote focus or iris requirement must be checked against the actual lens, camera and control interface; do not infer it from a standard mount.

Collect representative sequences before committing to hardware. Record dropped frames, target pixel width, blur, saturation and tracking failures together. A sharp isolated frame does not prove that a continuous sequence meets the requirement.

Check the design before ordering

Use this checklist to turn the calculations into an acceptance specification:

• Coverage: Confirm the smallest target still has enough pixels at the farthest useful position.
• Motion: Calculate exposure from transverse speed and allowable blur, separately from frame rate.
• Continuity: Check that inter-frame displacement and occlusion gaps suit the tracking algorithm.
• Optics: Verify sensor coverage, mount, focus range and edge sharpness with the selected lens.
• Timing: Measure exposure alignment rather than relying only on host timestamps.
• Throughput: Test acquisition, processing and recording together at the required pixel format.
• Lighting: Validate the dimmest and highest-contrast scenes within the exposure ceiling.
• Calibration: Include overlap and usable image margins; the calibration coverage guide explains why filling the image edge-to-edge can be misleading.

The usual answer—more pixels and more frames—is wrong when occlusion prevents observation, perspective makes the far zone too small, or available light cannot support the exposure. Additional viewpoints, narrower tracking zones or a revised measurement requirement may help more than a larger sensor.

Start a candidate build in the camera configurator. For a requirements review, send the field dimensions, smallest target, maximum speed, illumination conditions and timing tolerance through application support. Those inputs make the next decision specific enough to verify.