Current status of low-altitude drone use
With the development of science and technology, drones have been employed in all walks of life, but there are still some limitations in the use of drones. Based on feedback from all parties, our company has integrated software and hardware resources and developed a "drone airport comprehensive management platform". This solution can realize the dynamic allocation of the use of multi-type mission drones, allowing data resources to be shared in a centralized manner, making flights safer and data more secure, thereby "reducing costs and increasing efficiency" for government governance and making law enforcement "accurate and reliable."
Traditional working methods
Currently, most drone users in the industry still use manual or semi-automatic operation modes. On the one hand, the operation process requires personnel intervention and has high manpower requirements. On the other hand, the operation results depend on the pilot's skill level, and the operation threshold is high.
New way of working
7*24 hours unmanned operation.
Unmanned operation: After the whole system is installed and deployed, the operation can be completed remotely without the need for on-site personnel.
Fully automated: After the first configuration through the cloud platform, the airport can automatically complete the inspection work at a fixed time and location.
Fully automated airport system architecture
Natural Resources - Farmland Supervision
It can collect high-precision orthophoto data for the covered area in a very short time. Compared with the quarterly updates of satellites, the high-frequency (7-10 days) orthophoto production capacity makes farmland supervision possible.
Combined with AI algorithms, it can realize the ability to automatically supervise whether farmland is occupied, whether law enforcement is in place, and check the restoration of farmland, which greatly improves the supervision of land by natural resources departments.
Illegal construction investigation
Three-dimensional violation inspection; through full-link automated production capabilities, regularly collect oblique photography models of areas with high incidence of illegal construction. This model can truly reflect geographical details, fix the construction situation in the jurisdiction, and avoid buck-passing.
Two-dimensional illegal construction inspection: frequently collect high-precision orthophotos of areas with high incidence of illegal construction, use AI algorithms for comparison, and automatically detect changes in buildings.
Pre-disaster and post-disaster patrols
In areas prone to typhoons and heavy rains, patrols are carried out in the jurisdiction based on the warning information from the meteorological station.
Through automated and high-precision image analysis, the impact of disasters can be quickly identified and assessed, and the efficiency and accuracy of patrols can be improved, so as to better conduct disaster assessment, resource allocation and emergency response.
The AI detection model uses deep learning technology to train a large amount of labeled pre-disaster and post-disaster image data, and learns to identify and locate disaster-related features (such as flames, broken windows, collapsed buildings, etc.)
Urban fire reconnaissance/wildfire inspection
Fire detection is greatly advanced
The application of drone airports in advance reconnaissance scenarios at fire scenes has initially achieved "fast dispatch, fast reconnaissance, fast early warning, fast response, and fast image transmission", deepening and expanding the "one short and three fast" concept of fire fighting and rescue.
Fast dispatch: Take off within 1 minute after sending the command.
Fast arrival: After taking off from the hangar, the drone quickly flies to the disaster site at high altitude.
Fast handling: Continuously transmit visible light, thermal imaging and other information to the command center and commanders on the way.
Through the combination of visible light, infrared and other sensors with AI algorithms, forest fire patrol warnings can be realized in mountainous areas.
The input image is divided into grids, and each grid is predicted by the AI model to have a bounding box and category probability of a potential flame area. The model has been trained with a large amount of labeled flame data, and can quickly identify flames in images and output detection results including flame location and confidence.
Crowd flow detection (security guarantee for major events)
Urban areas with large passenger flow (hub squares, port squares, scenic spot entrances, etc.) need to realize coherent passenger flow information, and when the passenger flow is too large, the emergency evacuation mechanism needs to be activated.
The Shenzhen Fei platform has deployed an AI algorithm for passenger flow monitoring to realize instantaneous passenger flow viewing in the monitoring area. Its mobile capability is a supplement to the existing monitoring.
The passenger flow AI detection model is mainly used to monitor the number of pedestrians in a specific area in real time. It is based on computer vision and deep learning technology. The data collected by the drone lens is input into the model. Each detection task is regarded as a single regression problem. The bounding box and category probability are directly predicted from the input image through a single forward propagation, and the purpose of real-time target detection is finally achieved.
Deployment site selection conditions
- Open space
The airport can be deployed on the ground, roof, tripod, etc. The installation location should be open, without buildings or mountains, and avoid strong interference signal sources such as operator base stations and high-voltage power poles.
- Multi-dimensional safety testing
Including magnetic field detection to confirm that there is no magnetic field anomaly;
Signal quality detection to ensure signal stability within the communication distance;
Navigation positioning signal detection to evaluate satellite signal strength and interference;
Obstacle detection to avoid obstacles around the airport that affect drone takeoff and landing.
- Stable power and network support
Generally speaking, an automated airport needs to provide 220V/16A AC power, reliable grounding, and a wired network of 10Mbps or above.
Implementation plan - GeoDock Pro V4 + M350
