The art of farming has been all but lost in today's world of mass-production. Instead of individual crops being given exact amounts of water, nutrients, and pesticides, entire fields are treated uniformly. Even distribution of water and nutrients means that some crops may receive too much while others may be malnourished. The broadcasting of fertilizers over entire fields wastes fertilizer, which costs farmers money, and also wastes tilling, which intensifies erosion. The application of pesticides and herbicides over entire fields can result in over-treatment that damages crops and polluted watersheds.
An alternative to the mass-treatment of crops in agriculture is called precision agriculture. Precision agriculture uses sensing and computing technology to allow farmers to perform site-specific management with variable-rate technology. This has given rise to the need for crop monitoring systems, and hence our team's interest in UAVs. UAVs are relatively inexpensive, can be controlled remotely, and can even be programmed to fly autonomously, which makes them an ideal platform for a crop monitoring system.
Robotics Vision's goals for a crop monitoring and assessment platform:
- Low – mid altitude remote sensing
- Survey large areas quickly
- Identify crop condition
The current UAV we are using is a Mikrokopter.
The multispectral camera mounted underneath the UAV is a vegetation stress camera, which has three filters: blue, green, and near-infrared. With this type of camera, crop stress can be detected because healthy plants tend to reflect more light in the near-infrared spectrum than unhealthy plants.
During the summer of 2013, the Mikrokopter was flown weekly over an apple orchard at the University of Idaho Research and Extension Center in Parma Idaho. Below are a few of the images captured from the orchard. The images are taken from an altitude of 100 m.
The water and nitrogen data for each tree in the orchard is recorded by the research facility. We are currently attempting to find a correlation between the normalized difference vegetative index (NDVI) values from the images.
The graph below is a comparison of the average NDVI values from the multispectral images of the trees with the watering methods and nitrogen content of five rows of trees over a period of ten weeks.
Preliminary results show that water-stressed trees could be identified by lower NDVI values. It can also be observed from the figure above that there is no correlation between NDVI and nitrogen content. Additional imaging will be conducted this summer to validate the water-NDVI correlation and to investigate the effect of nitrogen content on NDVI.