Automating the greenhouse of the future.
Growing plants more efficiently is one of the world’s great challenges, when you consider the global need for food, both now and in the future. Another great agricultural challenge is managing the ever-changing inventory of plants in a typical, large-scale greenhouse. From seed to harvest, plants must be tracked and managed throughout the process to meet government regulations and maximize yield. To automate what has been traditionally been an error-prone, manual process, a global agriculture leader partnered with RIIS.
The client is a worldwide leader in the creation, support, and advancement of agricultural hardware, including just-in-time crop health management, crop sensing, nutrition measurement, connectivity and data management. In addition to precision agriculture, their products advance industries such as construction, geo-positioning, mapping, surveying, forestry, mining, utilities, forensics and education.
Lights, cameras and a lot of action.
In an industrial greenhouse there is a lot of action. Throughout the four growth stages of germination, seedling, vegetative and flowering, plants move to different rooms based on the lighting conditions required, current health status, nutritional need and time until harvest. To maximize yield, plants need constant attention, care and nutrients. To handle these conditions, the plants are physically moved from room to room in the greenhouse. Management must know exactly where each plant is within each room at all times. And to revolutionize the process we utilized high-tech cameras and custom software.
Dude, where’s my plant?
For this project, we created a hardware prototype capable of collecting up-to-the-second plant location and health data from inside a greenhouse to automate what has traditionally been done by hand. We used several off the shelf hardware pieces to create a stand-alone sensor device that can connect to the internet and take vital readings in real-time. After developing firmware to integrate all of the on-board sensors, we produced a new user interface for basic operation, debugging, and data collection.
Putting the pro in prototype.
After compiling drivers from source, as well as vetting SDKs and utilities, we were able to provide all-new functionality to the prototype sensor. Using traditional and infrared cameras, the sensor records the exact location and health status of each plant. Additionally, it reads the temperature and humidity where the plant is located. By reading from RFID tags, the greenhouse can track each plant and all its movements from original planting all the way to harvest with previously unimaginable efficiency.
Poetry in motion.
The ultimate success of this project lies in the accuracy of the data. From seedling all the way to graduation from the greenhouse, plants are in constant motion. They move from room to room, and floor to floor, based on a host of factors. Now, with the new prototype sensor, they can arrange plants in a known grid layout within the greenhouse, and the RFID tag reader verifies plant location and keeps all tracking data up-to-date, right up-to-the-second.
Key technologies used.
The initial prototype was built with Raspberry Pi hardware and the Jetson Nano Develop Kit from Nvidia. Additional sensors and hardware included the Intel® RealSense™ D435i camera, the FLIR Lepton® Radiometric camera, the Raspberry Pi NoIR Camera Board v2, a Hamamatsu Micro-Spectrometer and breakout board, as well as a RFID reader and both internal and external ambient air sensors. The software included a custom solution built around SDKs from RealSense and Jetson. The operating system is Ubuntu and all development happened in Python and C++, with code built via Jenkins.
How did we do it?
We work in a similar way to the greenhouse. We strive to be as efficient as humanly possible. Scrum and Agile methodologies were used extensively with daily scrums and two-week sprints. We also performed reviews and retrospectives at the end of every sprint, planning session and backlog grooming. There were 3 total sprints for this project, and they were executed over a period of six weeks.
The not at all gory details.
If you’re still reading, you’re probably a lot like us. Interested in every little detail. It’s how we continue to be successful for our clients and partners. We leave nothing to chance, right down to the last detail. For this project, a custom sensor bus was created to handle all in-coming and out-going data channels from each sensor and to manage the priority of each input and output stream. It was written in C++ to make it fast and responsive. Several sensors had base level drivers in Python or C++, and testing scripts were created to individually test each sensor and report results, issues or fatal errors.
Once the prototype passed the overall testing script, standard output was connected via the HDMI port out to a monitor. Live camera footage and still photos could be seen, as well as real-time temperature, humidity and location data. The whole system can be controlled remotely by a mobile app, allowing the user full access to visualization of the greenhouse via the cameras or 3D modeling to monitor real-time data streams for plant health, temperature, humidity and over-all history right down to the individual plant.
Key Services Provided:
- Agile Methodology
- Web Development
Lake Trust
ETA Detroit
culturecliQ
PillCast
AiM Verify
Wellzie
BCBSM
DTE Energy
CliqBit
