Pix4D is a company that has been providing photogrammetry software since it was founded in 2011. It is one of the names that pops up frequently when talking to drone operators and farmers and contractors operating their own drones.
In fact, drone imagery is just one of the images that can be viewed, processed and analysed in the Pix4D software. Others include images taken by hand or by plane with which the software creates customisable results that complement a wide range of applications and software. In this article we present four different user cases of Pix4D software, in 4 different crops and on 4 different continents:
Avocados are one of the few crops in Chile that can be cultivated commercially and economically on hills. On one hand, hills don’t contain pests or diseases that require frequent control measures. On the other hand however, crop care and protection and harvesting costs tend to be higher on hillsides. If the size of the avocado orchards is taken into account and the time spent on inspecting every plant: can avocado orchard management on hillsides be more feasible using new technologies?
The common way of managing avocado plants is to scout them on ground and to examine each tree to see if there are visible indications of a problem. In practice this would end up being a timely and a costly process, and still not everything can be seen by human eyes. Some plant diseases show no sign of an outbreak to the human eye. However, with reflectance maps created from crop indices, those plants might tell a different story. Having the advantage of drones as well as aerial and multispectral images becomes an important factor in deciding the viability of a project.
Chilean company Krops, together with Pix4D and one of their clients, developed a methodology for remote monitoring and care of the avocado orchards using drones, a MicaSense RedEdge camera and multispectral imagery. The aerial mapping started in August 2017 and was repeated mid-October and mid-April 2018. The overall surface, 1,100 hectares, was divided into four parts/farms with 200 to 380 hectares each to have comparable images for the software. During each flight, the drone follows the gradients of the hills level in order to maintain a similar altitude between higher and lower points of the field.
The processing of the vast amount of daily collected data proved to be a big challenge. The initial idea included processing all flight data captured on one day during the evening of the same day. As soon as the processing was finished on one (of the four), the individual projects were merged into bigger ones that represented between 90 to 100 hectares. This way of working resulted in reduced lead times of the flight results. In August 2017 the delivery time was 2 weeks, in October 5 days, and in April 2018 it was just 3 days. Having timely results was especially important for the client to compare the results after each flight. Aerial scouting resulted in supporting decisions such as:
Carrots are a key high value root crop in Dutch agriculture, with net yields up to 100 t/ha and a high nutrient requirement, especially for potassium. As farms increase in size and technologies evolve, it pays off to use new technologies for better and more precise management of larger fields. This helps both ways: more efficient operation saves money on inputs and it increases the total yield by the end of the year.
To assess the carrot plant health, quickly detect deficiencies and problems, and improve cultivation, the farmer together with agronomy firm and input provider Agrifirm, asked Aurea Imaging to acquire drone imagery and help with the analysis and interpretation. The project included image acquisition and processing, selection of soil sampling locations based on the imagery and advising on the implications of the results.
The field had 3 different crop zones with 3 different carrot varieties. Initially, although the NDRE (Normalised Difference Red Edge) map clearly showed variability, during scouting the field by the Agrifirm agronomists no clear differences in crop conditions were seen. This can be explained by the fact that plants show differences on their reflection to light for colours the human eye can’t see but multispectral cameras can.
Based on the NDRE imagery, 2 locations were selected for soil sampling, one sample on a ‘good = green’ location and one on a ‘bad = red’ location. The soil analysis showed large differences, especially the soil potassium (K) reserve, expressed as the ‘P-value’ which was 9 in the ‘bad = red’ location and 55 in the ‘good = green’ location. The optimal P-value for carrots is 30 – 40. Common fertilisation recommendations are based on a mixture of soil taken from across the field and the information derived from the NDRE map. Extra potassium fertiliser on the good locations might just flush to the groundwater while the worse locations will benefit from a much larger dosage of potassium.
The NDRE map assisted in selecting soil sampling locations and the discovery of soil potassium variations, which have a potentially large influence on crop growth and quality. The targeted potassium fertilisation reduced the overall fertiliser input and helped preventing the negative environmental impact of potassium percolation into groundwater. Innovative technologies using drones and multispectral imagery have a lot of potential in complementing field scouting and the art of agronomy.
Soil surveying, or classifying areas that share similar soil properties, plays an important role in farming. Properties such as soil pH, salinity, texture, slope, water availability, and erosion hazard can be estimated based on the mapping results.
This project was conducted by delineating the field according to two significant soil properties: organic matter and texture. The presence and quantity of organic matter in soil is important, because it can hold up to a thousand times more water and nutrients than other minerals. With RGB images from drones and Pix4D software, the visual differences were captured for further analysis with the goal of achieving better yield.
The entire region was divided into several small sections, with a target area of 80 acres mapped by Anez Consulting LLC using drone technology. RGB images were acquired with a SenseFly eBee drone and a RGB camera. 274 images were taken, covering 160 acres in total with an average ground sampling distance of 3.75 cm.
An up-to-date orthomosaic image (a combined aerial image from single drone images stitched together) was generated from the single RGB images and it was then used as a base for running a client-customised algorithm. By collecting and optimising from previous experiences and soil sampling tests, the client developed an algorithm that performs a bare earth analysis, which categorises the surface regions according to their soil properties. The results were used to create prescription maps for variable rate seeding of maize.
Full-field orthomosaic images from drone imagery were essential to the success of this project. It allowed Anez consulting to fill the remote sensing ‘information gap’ that was missing from satellite and manned aircraft imagery.
Alike in many other countries, current and future labour shortage imposes an increasing threat for agricultural production in Japan as well. A serious problem looms in the near future as Japan’s farming population age, with very few in line to take their place. The Japanese government, in response, is both encouraging labour-saving measures and actively looking for new ways to attract younger generations to the farming industry. One of the projects was implemented in Minamisanriku-cho, Miyagi prefecture, which is known for its high-quality rice. Space Entertainment Laboratory (Selab) and farmers from the area teamed with one of the largest cell phone carriers in Japan, NTT DOCOMO, to complete it.
In order to obtain a precise analysis of the small-scale rice fields, Selab used the Parrot Sequoia multispectral camera, which captures four discrete narrow bands and comes with a sensor to correct illumination in real time. A 3DR Solo drone was chosen to carry the multispectral camera and to map 15 hectares of the high-quality rice fields, at a ground sampling distance (pixel size) of 5 centimetres. The high resolution was chosen to detect problems in small clusters of rice crops. The imagery was used to create a reflectance map for each individual band, then the NDVI map was calculated based on the red and infrared bands.
Satellite imagery has not been applicable for the Japanese agriculture market because farmland distribution is much more concentrated and of smaller size than that of other countries. The average farm size is 2 to 3 hectares and the average rice patch is 10 to 30 acres.
Another benefit of drone mapping technology in the Japanese farming industry is the minimisation of field scouting labour for potential insect-damaged areas. While at the same, time encouraging young people to join farming by applying quantitative data to assist crop analysis and thus improve the growth.
As a result, there are promising opportunities in surveying Japanese rice fields with drones, due to the unique demographic characteristics of the country and the high resolution multispectral data suitable for analysing the farmland.
Read all the latest articles from the newest issue, just log in with your current details!