Who controls position?
DJI’s FocusTrack family distinguishes Spotlight, ActiveTrack and Point of Interest. Broadly, these support keeping a subject framed, tracking it or moving around it, but the official explanation explicitly notes model differences. Identical names do not guarantee identical supported subjects, recording modes, speeds, sensing or manual control. For the exact aircraft and firmware, document three things: who determines aircraft position, who determines camera aim, and how the operator stops the automated movement. That is more useful than memorizing a marketing label. A tracking box is not proof that the intended person or object was selected. Recheck in scenes with similar subjects and coordinate the start before the subject begins moving. Just as you settle who sets position inside the aircraft, settle who decides what among the crew. Learning the roles of pilot, observer and operations lead, and the briefing order (Tobira, in Japanese) is a primer. A review of vision-based control is the Inspection guide’s paper card Vision and Control for UAVs.
Plan for losing the subject
Consider occlusion, backlight, similar-looking subjects, frame boundaries and abrupt subject motion as test conditions. Check the documented behavior after target loss for the exact model and mode; do not generalize one example into a universal hover guarantee. Obstacle sensing has directional and environmental limits, so do not build a route around the assumption that wires or branches will be detected. Agree on a subject route and stop signal, and maintain an operating arrangement that allows aircraft and surroundings to be monitored. Decide intervention criteria before enabling automation. Define interruption criteria by visibility, subject route, bystanders and communication rather than only “if it looks risky.” Unexpected entry into the route should trigger the agreed procedure. Not relying on automatic avoidance is the same rule on inspection sites. Reading what each obstacle-sensing method can and cannot detect (Airffic Inspection) also helps predict when tracking will drop the subject.
Begin with a bounded test
Our original test procedure starts with a stationary subject in an open area where the intended operation is permitted. Observe a short, bounded movement and verify the stop action before introducing the planned subject motion. A subject’s consent does not resolve conditions involving other people or the location. Review framing stability, changes in subject scale and behavior after any recognition failure. If automation does not serve the shot, use a more suitable combination of manual movement and camera assistance. Automation is a means of achieving an image, not an achievement in itself. Record distance, light, background, subject motion and recording mode after a failed test. Those details define useful operating boundaries and inform retesting at another location or on another aircraft. Tracking tests belong in a wide space with nobody underneath. Airffic Training Hyperspace sketches how such test fields could be found and booked (sample listings, in Japanese).
YOUR NEXT STEP
Put it into practice.
- Do not generalize one model’s avoidance behavior to every similarly named mode.
- Do not let target framing replace monitoring of the actual flight path.
Sources and editorial scope
Explanations and exercises are original editorial material. Verify product-specific conditions against the linked documentation version.
Research date: 21 October 2026. Check the linked version, supported hardware and applicable conditions. Evidence includes directly reviewed text and official search-index records.
Read the methodology and open questions →