Overview / Usage

Tea estates are spread across vast areas of land and manually going through each crop isn't really possible as it will require an extensive amount of manpower. The same can be solved by the use of autonomous drones. Anomalies can be detected by observing the IR spectrum and by the use of autonomous drones programmed to follow a certain pattern/waypoints provides an easy way around.
Our proposed technology is focussed on the use of autonomous drone swarm and combine the IR data with a custom model to be optimized using the OpenVino distribution.

Methodology / Approach

Our proposed technology is focussed on the use of autonomous drone swarm and combine the IR data with a custom model to be optimized using the OpenVino distribution.
The system will be based on wireless sensor network where each drone will represent a sensor node. The visual data will then be either processed at edge level where the visual data will be passed through an inference engine to determine if an anomaly is present.
Additionally, telemetry data will also be channelized and a base station will be responsible for the handling of both visual and sensor data.
Visual processing is to be handled using tensorflow custom model based on mobilenet ssd and ultimately the model is to be optimised using OpenVino toolkit. Finally the inference engine will be responsible for the final output.

Update 1 (20-03-2019):
Work is in progress to setup the drone platform with a high level of stability. We are currently experimenting with the PID tuning and several modes of operation. Once we attain perfect stability, we will focus on the wireless RGB+IR video feed transmission.

Technologies Used

Drone control
1: GCS with Ardupilot based control board
2: PixHawk based developer board
3: Peripheral components for drone which includes GPS, IMU, RF link

Data processing
1: Python
2: Tensorflow
3: Intel distribution for OpenVino
4: Intel DevCloud

Collaborators