A Successful PTZ Auto Tracking Case

Auto Tracking functions in PTZs are oversold, but in certain cases can have real value. In this note, we examine a successful case study using Auto Tracking, describe the problem it solved, and provide details on how common pitfalls were avoided.

The Problem

As an integrator, I worked on a project that involved a novel application of autotracking at an airport facility. The customer asked for a camera to capture the tail numbers of aircraft entering a flightline and cross reference them with time/date stamps as an 'operational log' confirming when particular aircraft were active.

The goal of this project was to automate a process that otherwise required a staff member with a pair of binoculars to manually document activity. This staff member monitored two 'takeoff queing' areas, where aircraft would temporarily be staged before becoming airborne. Aircraft would stay in these areas for an average of 10 minutes, but sometimes for as little as 30 seconds depending on airfield traffic.

Because of the size of the stanging area, and the physical distance from the cameras, it was determined that fixed cameras were not sufficient to cover the area. The customer agreed that a dedicated camera for each takeoff area was optimal, however either camera needed to cover both takeoff areas, so an auto-zoom lens was also required.

Whatever solution was designed, the customer required that cameras must record activity to a 'high degree of confidence' so that no takeoff activities could be missed. The system had to pick up on registration numbers on every aircraft in all situations. Anecdotally, the customer said "Recording 99% of flights is passable, but only 98% is not acceptable." 

The Solution

After a successful 'proof of concept' test, it was determined a PTZ camera with autotracking would be provide a novel solution. The flexibility of the camera, coupled with the 'smartness' of the camera would enable the camera to be positioned to record tail numbers no matter where they might appear with range of the camera.

Two Axis PTZs with were installed onsite. The autotracking was configured within travel stops, so the camera would not be 'distrated' by insignificant movement outside of the intended FoV. Additionally, the camera was programed to return home after 3 minutes of inactivity to prepare for next flight.

Overview of Auto Tracking

The central element 'automating' the entire process was use of an autotracking feature. Here is a short marketing video demonstrating how the feature typically works:

The risk in relying on this feature is that like any analytic, it would not detect targets accurately. In this case, alternate plans to integrate motion sensors within a takeoff area were planned as an option. However, upon installation, the autotracing designed to be very successful and worked within the goal parameters. 

After a 'burn-in' period of two weeks, the customer was able to reassign the staff previously devoted to monitoring to other activities, and the project was considered a success.

Why It Worked

This deployment worked because auto-tracking was very tightly controled in a very controlled environment:

• Tight Scene: The tracking movements were limited to a rather tight space, approximately a 60 foot box. Rather than a situation where the camera is asked to detect movement in a broad area, the area defined in this application was only a fraction of the entire scene.
• No 'Interfering Noise': The only movement being tracked was within a restricted area with slow moving targets, generall moving in only one direction (away from camera). No 'movement noise' because it was restricted area, background was a tarmac with even lighting, color and no background movement.
• Images Only: Only visual 'picture' data was collected, no OCR or numerical data was gathered. Unlike LPR applications, the pictures did not need to be read for data, and only a focused image of the registration numbers were required.
• Home Position: Camera was programed to return 'home' after a period of time. Staging the camera in a central position allowed the camera to immediately pick up on new movement.

Conclusion

Surveillance cameras can be adapted into operational roles, and in some cirmstances provide unique solutions. In this situation, the combination of a tighly controlled environment coupled with powerful camera features resulted in a success story.