## [Case Study] NDVI Imaging In Plantation Management

by Afiq Iskandar

Image via Britannica

This week, we will be looking at the real-life applications of NDVI imaging usage in plantation management.

We have been going through a few articles about NDVI imaging, and now we will be looking at real-life applications that have proven to be effective in streamlining farming operations such as crop monitoring and yield forecasting.

This time around, we are looking at NDVI applications on two crops; sugarcane and oil palm (in the next article).

##### Multispectral Imaging

To understand how multispectral imaging can help us in crop monitoring and management, we first must understand the mechanisms behind multispectral imaging itself.

To start with, all objects absorb and reflect solar irradiance in different ways and of course, this holds true when it comes to plants. To simply put it, the ratio of the amount of light leaving a target to the amount of light striking the target is called reflectance.

Every plant has a specific amount of light reflectance which produces its unique spectral signature. Different types of vegetation differ in their reflectance values, which produces unique spectral signatures distinct to each of them.

Plants that are stressed or diseased can also be identified by their distinct spectral signatures. The leaf pigments, cell structure and water content are among the crucial factors that influence the spectral reflectance of vegetation.

##### Types of Vegetation Indices

###### NDVI (Normalised Difference Vegetation Index)
• NDVI quantifies vegetation by measuring the difference between NIR (which vegetation strongly reflects) and RED light (which vegetation absorbs).
• NDVI always ranges from -1 to +1. A higher value refers to greener and healthier vegetation. Lower values show unhealthy & sparse vegetation.

$$NDVI = {(NIR - Red) \over (NIR + Red)}$$
###### GNDVI (Green Normalised Difference Vegetation Index)
• Similar to NDVI except it measures green spectrum instead of red spectrum. This index is more sensitive to chlorophyll concentration as compared to NDVI and is commonly used to determine water and nitrogen uptake into the plant canopy.

$$GNDVI = {(NIR - Green) \over (NIR + Green)}$$
###### NDRE (Normalised Difference Red Edge Index)
• NDRE uses the method of NDVI to normalise the ratio of NIR radiation to Red Edge (RE) radiation
• It is sensitive to chlorophyll content in leaves (how green a leaf appears), variability in leaf area, and soil background effects.

$$NDRE = {(NIR - Red Edge) \over (NIR + RedEdge)}$$
###### LCI (Chlorophyll Index)
• The chlorophyll index is used to calculate the total chlorophyll content of the leaves. The total chlorophyll content is linearly correlated with the difference between the reciprocal reflectance of green/ red-edge bands and the NIR band.
• The green and red spectrum values are sensitive to small variations in the chlorophyll content and consistent across most plant species.

$$LCI = {{NIR \over Green}-1}$$
###### OSAVI (Optimised Soil Adjusted Vegetation Index)
• OSAVI introduces a constant value of 0.16 for the canopy background adjustment factor. This index determines soil background from vegetation cover.

$$OSAVI = {(NIR - R) \over (NIR + R + 0.16)}$$

In this case study, we are looking at the usage of the Phantom 4 Multispectral paired with an RTK base station as it is widely used for spectral imaging for agriculture purposes.

##### Tools and Software

Phantom 4 Multispectral

The DJI Phantom 4 Multispectral is equipped with sensor output of 1 visible light band and 5 light bands (R, G, B, NIR <Red-Edge).

Detecting infections or problems are important as farmers can administer preventive measure to avoid them from spreading further and doing more damage to the crops.

The P4 Multispectral consolidates the process of capturing data that gives insight into crop health and vegetation management. DJI has created this platform with the same powerful performance standards that DJI is known for, including 27 minutes max flight time and up to 7 km1 transmission range with the OcuSync system.

###### DJI Terra

DJI Terra is a software tool that transforms drone data into digital 3D models and maps for easy analysis and decision-making.

DJI Terra enables businesses and organisations using DJI drone technology to capture, visualise and analyse aerial images for a wide variety of applications across the public safety, construction, infrastructure, agriculture and film industries.

###### Airamap

An advanced aerial mapping analytics software, Airamap is used in industrial applications to monitor on-site situations so you can do swift and accurate decision-making. Airamap is equipped with an artificial intelligent (AI) data processing engine that cuts the processing time short while maintaining the quality of the processed results.

##### Case A: Sugarcane Brix Estimation

The scope of this case study is to determine the reliability of multispectral imaging to estimate Brix (sweetness) levels in sugarcane, validating it with the relevant ground data.

The aim of this operation was to mainly monitor the crops, forecasting yield, and extrapolate harvesting sequence using the Brix Prediction Model.

There are two different varieties that were planted in the area; Khong Kean 3 (KK3) and Thong Poom 6 (TP6).

The results delivery is broken down to:

1. 2D RGB orthomosaic
2. 2D Multispectral Orthomosaic (NDVI, NDRE, GNDVI, OSAVI, LCI)
3. Classification map
##### Workflow
###### Data capture

Data capture is the initial process of any drone operations. This is done by the pilot flying the platform (i.e drones) mounted with payloads (i.e. sensors/camera) that are able to capture specific set or format of data on the intended site.

This data capturing process normally is done automatically and the completion time normally depends on the size of the land itself.

In this case, Phantom 4 Multispectral was used to collect the data and was supported with Ground Control Points (GCP) for enhanced accuracy.

###### Data processing

After capturing the data, it would need to be sorted out as the raw data is a big chunk of visuals/images that need to be combined to form a whole picture. This is attributed to the way UAV platforms capture on-ground data.

This process of combining the data is called stitching, a process where overlapping sections in the images are cut out, forming one big, complete map.

Only after that can other analytical processes such as NDVI analysis, terrain modelling and tree counting take place.

In this study, we used DJI Terra and Airamap airamap.com for radiometric calibration and ArcMap for image classification, segmentation and zonal analysis. Radiometric correction is done to calibrate the pixel values and/ correct for errors in the values. The process improves the interpretability and quality of remote sensed data. Radiometric calibration and corrections are particularly important when comparing multiple data sets over a period of time (which is what .

Radiometric calibrations will result in models such as DSM, orthomosaic map and reflectance and vegetation index maps such as NDVI, NRE, LCI and many more.

In-depth analysis such as image classification involves extracting information classes from a multiband raster image. Other than that, there is also the segmentation analysis that is actually a process of grouping neighbouring pixels together that are similar in colour and have certain shape characteristics.

(Raster: In its simplest form, a raster consists of a matrix of cells (or pixels) organized into rows and columns (or a grid) where each cell contains a value representing information, such as temperature. Rasters are digital aerial photographs, imagery from satellites, digital pictures, or even scanned maps. Read more on raster: ArcGis)

###### Data visualisation & reporting

After data is processed and simplified, the complete information can be put together in a report to give the clients/users insights and important information which in turn, utilised for decision-making and preventive actions. See more on the results section.

###### Results
Image from DJI Terra processing

• Total Area: 5 ha
• Total images: 2,142 Image size: 7.27 GB
• Processing Time: ~10 minutes (Using processing specs of 128 GB RAM and Nvidia 2080Ti)
• Flying Altitude: ~70m

Read more: DJI Terra

2D Orthophoto Map
Vegetation Indices Map
##### Findings

Over a period of five weeks, multispectral data and ground data of the sugarcane plantation was collected and processed.

Later, it was found that, out of the other six vegetation indices maps, LCI or the chlorophyll index has the highest correlation to the Brix value when analysed using a logarithmic function model.

The data model from week 1 until week 4, was later on, used as a basis for week 5 prediction. The discrepancy error from week 5 was 1.01% (KK3) and 2.09% (TP6) from the original data.

The discrepancy can be caused from a various number of factors, one of the major factors would be the weather, temperature and soil moisture.

As we all know, chlorophyll is affected by the weather, which determines the sunlight coverage of an area fluctuating the temperature and soil moisture content.

These three factors are continually fluctuating throughout the day, despite the consistent time of day when the data collection process was executed, which in turn, directly affected the indices calculation.

##### Conclusion

In summary, it was found that the pre-harvest prediction model is useful to predict the optimum level of sugarcane sweetness for future harvest cycles. This method requires a full crop cycle of sample data to be used effectively.

Other than that, Phantom 4 Multispectral has been proven useful to perform crop analysis, particularly in determining the harvesting sequence, sugar estimation in sugarcane, crop condition, problematic areas and yield forecasting.

Furthermore, NDRE and LCI indices can be used to manage vegetation based on variables such as chlorophyll content to evaluate sugarcane growth and early detection.

Different vegetation indices can give different types of crop analysis. NDVI and GNDVI are useful for sugarcane greenness that provides information on crop growth and nut

In the second part of this article, we will be looking at the usage of the NDVI and other vegetation indices in determining and analysing palm oil crop health and harvest sequence.