The project idea is making a relatively low-cost system to monitor plants and cultivation areas to detect the grown state as well as potential parasite and damaged plants.
The device can be easily installed on a semi-autonomous UAV to cover large areas of terrain, as well as installed on a ground robotised device. The use of a UAV represents the most flexible solution compared to the ground mobile unit.
Some of the most challenging issues if adopting the ground mobile solution:
- Difficulty to move on the non-regular terrain
- Limitations due to the kind of plants
- Slower and more difficult mobility
- Reduced field of operation
- Slower back-to-home maneuvres
The video below shows the first proof of concept of the nanodrone, developed with an Arduino Nano 33 implementing visual recognition with Tensorflow Lite installed on a DJI Mavic Mini drone.
Drafting the Project
The image above shows the draft of the project through a typical workflow:
- A drone with onboard the collection/inspection device will move along a field following a predefined path.
- The collected information – including visual data – coming from several sensors are integrated realtime and saved locally on a microSD card.
- The device has a GPS to save the acquisition points, independent by the drone navigation system
- Every return-to-home cycle of the drone data the information are updated via BLE to the PSoC6 Pioneer Kit (a small mobile station to the ground) that collect every inspection fly session (minimum one).
- At the end of the series of inspections of the field (may need more fly cycles, depending on the extension) the whole inspection set of retrieved information are sent to the AWS IoT Console via MQTT, certificates and several shadows to monitor several inspections acquired along the time.
- The full data retained by the PSoC6 Pioneer Kit station are sent (when the boards are on the same network) via WiFi to a Raspberry Pi that can process more detailed information.
Real-World Project Applications
Accordingly to the kind of data it is possible to acquire and the position repatability of the sampling there are at least three main areas of application of this project, that in my opinion can offer the opportunity to grow the prototype to a product level:
- Plants and cultivated trees inspection for small and medium-size farming
- Architectural structures variation on time and deformation analysis.
- Environmental impact changes
Pest, parasites, growing stage of fruits and grasps, and more can take advantage from this kind of local, medium-range inspection where – in a similar way – satellite specific-range visual information are acquired for large terrain areas, wildlife zones etc.
Integrating the visual inspection information together with the environmental conditions (temperature, humidity, etc.), weather conditions, and time-of-day it is possible to track growing curves of the evolution of some phenomenons curves that for some reason are impacting the productivity level of the cultivations.
Sensors data collected can be integrated locally (on the ground Raspberry Pi machine) with drone photos acquired in the same position to provide more specific and detailed information, as well as a visual history of the acquisition. The core information collected and pre-processed by the PSoC6 Pioneer Kit ground unit instead, are sent to the AWS IoT Console for changes over time analysis.
Above: the PSoC6 Pioneer Kit box case designed with Fusion360 and 3D printed with the Elegoo LCD Saturn 3D printer.
Structural Variations and Environmental Impact Changes
The possibility to precisely repeat along a timeline (maybe daily lot less frequent, depending on the kind of inspection) gives the nanodrone project the possibility to acquire comparable series of data during periods. Based on this...Read more »