Autonomous flight in GPS-denied environments(under bridges, indoors)
ACSL has developed a proprietary visual SLAM (Simultaneous Localization And Mapping) system using cutting edge machine vision technology. With this system, ACSL's PF1-Vision drone can operate autonomously in a variety of settings that are not feasible using traditional GPS-based position sensing. Adding a camera best suited for the purpose of inspection makes PF1-Vision the ultimate inspection machine.
Manual flight / Attitude hold mode / GPS position hold mode / GPS waypoint flight, planned-route flight / 920MHz data link (usable up to ≈7km distance) Alerts for all critical systems / "Go home" on user command or data signal loss / Flight recorder
- CUSTOMIZED PARTS
- GPS denied environment / Laser SLAM / Autonomous flight
Machine vision technology for localization and orientation recognition
Typical drones rely on GPS to achieve autonomous flight. However, GPS / GNSS signals are often too weak or unreliable near high structures, under bridges, indoors, etc. ACSL has developed a proprietary machine vision (visual SLAM) algorithm that enables PF1-Vision to compute its position and orientation from the natural landmarks that it discovers with its downward-facing camera.A forward-facing stereo camera combined with ACSL proprietary machine vision algorithms enables PF1-Vision to sense the distance to objects in front of the camera. With this information PF1-Vision can avoid collisions or inspect an object of interest at a desired distance.
- Bridge inspection
- Enabling detection of cracks as small as 0.1mm. Using PF1-Vision for primary screening helps users to limit arears requiring detailed inspection, thereby minimizing total cost.
- Factory equipment inspection
- PF1-Vision can fly near pipes and tanks. Using PF1-Vision for regular inspections and high-level screening reduces need for inspecting high and dangerous locations.
- Building inspection
- GPS/GNSS data is unreliable near high-rise buildings. PF1-Vision can fly in these environments.
Drones must recognize its own position to achieve autonomous flight. PF1-Vision recognizes its position and orientation with machine vision technology called visual SLAM (Simultaneous Localization and Mapping). The visual SLAM system compares pixel by pixel and selects the most unique visual features in real time (shown as green dots on the photo). Then computes the locations of those features to determine its own position and orientation.
Visual SLAM is realized by utilizing cameras and a GPU on board the drone. PF1-Vision is equipped with a forward-facing stereo camera, a downward-facing camera combined with an on-board NVIDIA JETSON™ TX2 for real time visual SLAM.
Distance control with machine vision
Just like human eyes, a forward-facing stereo camera combined with ACSL proprietary machine vision algorithms enables PF1-Vision to sense the distance to objects, as small as 3cm, in front of the camera. With this information PF1-Vision can avoid collisions or inspect an object of interest at a desired distance.