Slew-to-cue: turning a radar track into a positive visual ID

Slew-to-cue is the automatic hand-off of a target position from one sensor to a pan-tilt camera, so that the camera turns to the target without an operator steering it. In counter-UAS work the cue normally comes from a radar: the radar reports that something small is at a given bearing, elevation and range, and radar PTZ integration points an EO/IR turret at that spot so a person can see what it is. Detection is not identification, and slew-to-cue is what makes the second step fast enough to be useful.

Why radar PTZ integration is needed at all

A radar and a long-range camera have opposite strengths. The radar watches the whole sky around the site continuously but produces a symbol on a map, not a picture. The camera produces a picture that a person can judge, but at long focal length it looks at a very small part of the sky at any moment.

An operator searching for a small drone with a joystick at high zoom is unlikely to find it before it has moved on. The radar already knows where the drone is. Passing that position to the turret automatically replaces a manual search with a slew of about a second.

The visual step matters because a response usually needs more than a radar classification. Before a site is closed, a jammer is used or police are sent to a location, someone has to confirm that the track is a drone and not a bird or a balloon, and ideally see whether it is carrying anything.

The hand-off, step by step

1. The radar forms a track. After several consistent detections it publishes a track with position, altitude, course, speed and a classification. The Ga-5000 drone detection radar outputs tracks in ASTERIX or JSON.
2. A rule selects the track. The integration software or the operator decides which track the camera should look at: the nearest, the one classified as a drone, the one entering a defined zone, or one picked by hand.
3. Coordinates are converted. The radar reports the target relative to its own position. The camera is somewhere else, perhaps tens or hundreds of metres away. The software converts the target position into pan and tilt angles as seen from the camera, and predicts ahead along the track to allow for the age of the data.
4. The turret slews. The pan-tilt unit drives to the commanded angles and sets zoom and focus to suit the reported range.
5. The video tracker takes over. Once the target is in the picture, an image-based tracker locks on and keeps the turret centred on it frame by frame, at a much higher rate than the radar updates.
6. The operator identifies. With a stable image, the operator narrows the zoom, confirms type and payload, and can take a laser range for a precise target coordinate.

The pointing budget

For the target to appear in the picture on the first slew, the total pointing error has to be smaller than half the camera’s field of view at the zoom used for acquisition. Several separate errors add up to that total.

Radar measurement error

The radar’s angular accuracy is usually the largest term. The Ga-5000 specification lists angular accuracy of 0.5° or better in both azimuth and elevation and range accuracy of 10 m RMS or better. An angular error of 0.5° corresponds to about 9 m across the line of sight at 1 km and about 44 m at 5 km. The acquisition field of view has to be comfortably wider than this, which is why the camera should arrive at a moderate zoom and narrow down after the video tracker has locked.

Staleness of the cue

A rotating radar sees the target once per rotation. The Ga-5000 update cycle is 3 or 6 seconds, selectable. A drone flying at 20 m/s travels 60 m in 3 seconds and 120 m in 6. At a range of 1 km, 60 m across the line of sight is more than 3° of angle, which is larger than the radar’s measurement error. Two things reduce this. The software extrapolates the track forward using its reported course and speed, and the radar can refresh a priority target more often: the Ga-5000 terminal provides a priority track lock that refreshes the selected target’s data at 2 Hz.

Alignment and parallax

The camera’s zero direction must agree with the radar’s, and both must agree with true north. Any residual offset in azimuth or tilt adds directly to the pointing error for every target. Parallax is the second geometric term: if radar and camera are at different positions and the conversion ignores this, the error is small for distant targets and large for close ones. Both are fixed at installation by surveying the sensor positions and calibrating against a target of known position.

Turret accuracy and speed

The pan-tilt unit has to go where it is told and get there quickly. On a modern turret this term is small. The Orbit UAV PTZ tracking camera lists positioning accuracy of 0.005°, two orders of magnitude finer than the radar’s angular accuracy, and slew rates of 180°/s in pan and 120°/s in tilt with continuous 360° rotation. At that pan rate the worst case, a target directly behind the current pointing direction, is about a second of travel before acceleration and settling are counted.

Summary of the budget

Error source	Typical size	Published Orbit UAV figure	How it is controlled
Radar angular accuracy	Dominant at long range	≤0.5° azimuth and elevation (Ga-5000)	Wider acquisition field of view
Radar range accuracy	Minor for pointing; affects focus and zoom preset	≤10 m RMS (Ga-5000)	Autofocus after acquisition
Cue staleness	Dominant for fast, close targets	3 or 6 s update cycle; 2 Hz on priority track (Ga-5000)	Track prediction, priority track
Alignment and parallax	Constant bias; grows at short range	Site dependent	Survey and calibration at installation
Turret positioning	Negligible	0.005° (PTZ camera)	Closed-loop brushless drive
Slew time	About a second or less	180°/s pan, 120°/s tilt (PTZ camera)	Fast drives, short settling

Failure modes to design out

Stale cue. The turret arrives where the drone was. The causes are an unpredicted cue, a long radar update interval, or delay in the network and software path. The remedies are prediction, a faster update on the selected track and timestamps on every message so that the integration software knows how old the data is.

Wrong target in the frame. Two objects are in the field of view and the video tracker locks on the wrong one, often a bird or a light in the background. Using the radar range to set expectations for apparent size, and re-checking that the camera’s line of sight still agrees with the radar track, guards against this.

Coordinate mismatch. Radar and camera use different north references, different height datums or positions that were entered by hand and are slightly wrong. The symptom is a consistent offset in one direction. It is an installation fault and is solved by calibration, not by software tuning.

Loss of video lock. A small drone against cluttered ground or bright cloud can drop out of the tracker. The system should fall back to the radar track and re-cue automatically, not leave the turret staring at the last known position.

Platform motion. On a vehicle or vessel, the camera base moves between the cue and the slew. Gyro stabilisation is needed to hold the line of sight; it is listed as an option on the Orbit UAV turret for mobile installations.

Day, night and range

The visible channel gives the most detail in daylight. A thermal channel finds the target at night and often picks out a warm drone against cold sky more easily than the eye can in visible light. The Orbit UAV turret carries an uncooled LWIR imager as standard, with an optional cooled MWIR imager with 18 to 275 mm continuous zoom for long-range work, and a laser rangefinder that measures to 6 km. Its published working range is up to 20 km, target dependent: a small drone is identifiable at a far shorter distance than a large aircraft, and atmospheric conditions matter as much as optics.

Other sources of cue

Radar is not the only sensor that can cue a camera. A passive RF detection sensor provides a bearing to the drone’s transmitter, listed at 3° RMS or better for the Orbit UAV unit, and for recognised protocols a position as well. A bearing alone gives the camera a line to search along and no elevation, so it is a coarser cue than a radar track, but it is available before take-off and it tells the operator which model to expect in the picture. How the two sensors complement each other is covered in Doppler radar vs RF drone detection.

FAQ

What is slew-to-cue?

Slew-to-cue is the automatic pointing of a pan-tilt camera at a position supplied by another sensor, such as a radar, an RF detector or a map click. The camera slews to the cue and a video tracker or the operator takes over from there.

Why does a counter-UAS system need a camera if it has radar?

Radar reports that an object is present and classifies it, but it does not show what the object looks like or what it carries. A camera lets a person confirm the detection and provides recorded evidence. Most response procedures require that confirmation.

How accurate does the radar need to be for slew-to-cue?

Accurate enough that the combined error of radar measurement, cue age and alignment is less than half the camera’s acquisition field of view at the target’s range. Radar angular accuracy of a fraction of a degree is workable when the camera acquires at moderate zoom and narrows after lock.

What protocols are used for radar PTZ integration?

Radar tracks are commonly published in ASTERIX or in a JSON format over the network. Cameras are commanded over network or serial interfaces; the Orbit UAV turret lists RJ45 network, RS422/485 control and ONVIF. Integration software or the command system sits between the two and performs the coordinate conversion.

Can the camera keep tracking if the radar loses the target?

Yes, for as long as the video tracker holds lock. The two trackers are independent once the camera has acquired the target. If the video tracker also loses it, the system waits for the radar to re-acquire and cues again.