I. Introduction

During the last years, the availability in the natural resources such as water, air, and soil has been reduced in quality and quantity; on the other hand, the climatic variability has been increased and this has caused several economic losses to both small and large agricultural actors in Cuba. Likewise, the insufficient planning of the crops in most parts of the agricultural Cuban sector joined with the cost overruns associated with the diseases attention and the poor technical assistance have entailed effects involving low competitivity of the national agricultural sector and low market penetration of Cuban products in the international markets, where high quality standards related with a low use of agrochemical materials and sustainable practices are required.

There is a large need to implement efficient and precise techniques in agriculture, since that will lead tothe producers to reduce their supplies expenses and will increase the yield production (Saxena & Armstrong, 2014). The so-called Precision Agriculture [PA] is an agronomic concept related with crop field management based on the analysis of the crop variability, where its implementation implies the use of current technologies such as Global Positioning Systems [GPS]; sensors; satellites and aerial images; Geographic Information Systems [GIS]; among others to estimate, assess, and understand such variations (Marote, 2010).

Within PA, a cycle of agricultural practices is implemented that replaces the current supplies recommendation —where their amounts are calculated based on average values—, with another one where the formulated amounts are more precise through a localized treatment and considering the yield variations in all the cultivated area. In consequence, PA is focused on the optimization of the resources usage by locating in the soil the number of seeds that each point supports together with the right amount of supplies and water. Furthermore, the weed control is performed only where required and not indiscriminately (Best & Zamora, 2008).

The multispectral photogrammetry is one of the main tools for PA, it is a passive technique (in the infrared range between 0.76 and 14 µm) focused on water, element affecting the technical properties of the plants; this makes this technique useful where the leaf has several amount of water per surface unit (Vibhute & Bodhe, 2012).

The crop monitoring activity from the imagegathering of has been performed by using sensors transported in the air via manned and unmanned vehicles ( Guo, Kujirai, & Watanabe, 2012; Torres, Gómez, & Jimenez, 2015; Campo, Corrales, & Ledezma, 2015), which entails advantages relative to the use of satellite images. This last because satellite images do not have the temporal and spatial resolution that the crop monitoring activity requires; furthermore, data can be affected by meteorological aspects —such as the presence of clouds— complicating the observation of the coverage at the ground level. The images taken from aerial platforms comply with the need of immediate image gathering with superior characteristics than the ones obtained from satellites (Basso, 2014; Gago et al., 2015; Hernández et al., 2016).

The multispectral images obtained by using Unmanned Aerial Vehicles [UAV] allow the obtaining of vegetation indexes [VI], where these last are algebraic combinations of several spectral bands designed to highlight the strength and properties of the vegetation such as the biomass, absorbed radiation, and the chlorophyll content (Gutiérrez-Rodriguez, Escalante-Estrada, & Rodriguez-Gonzalez, 2005). The calculation of several VI allows to understand the information of the Near Infrared Spectroscopy [NIR] and Red Green Blue [RGB] images obtainedby using photogrammetry techniques. Some of the achieved parameters using this technique are the Normalized Difference Vegetation Index [NDVI], plant strength, Green Normalized Difference Vegetation Index [GNDVI], nitrogen level, Crop Water Stress Index [CWSI], and Soil Adjusted Vegetation Index[SAVI]. This last is an adjustment of the agronomic studies to the soil type in the crops(Candiago, Remondino, De Giglio, Dubbini, & Gattelli, 2015).

Previous research works have documented the capability of vegetation indexes to estimate the potential crop yield of the sugarcane; nevertheless, most of them have been focused on the use of satellite and passive sensors platforms (Lofton et al., 2012; Zhao et al., 2016; Johansen et al., 2017; Martínez, 2017)and very few have demonstrated the ability of an active remote sensor on the ground to estimate variables of interest in a sugarcane crop located in a tropical zone.

Within the Cuban sugar company, there is a growing interest for the application of PA techniques looking foran increase in the crop yields. Be able to delimitate environments with several productive potentials in a same sugarcane field would be a good starting point for the implementation of the site-specific handling in this crop.

The main objective of this work is to perform a multispectral photogrammetry work by using an UAV in the fields of the Territorial Research Station of the Sugarcane [ETICA, Estación Territorial de Investigaciones de la Caña de Azúcar] to elaborate maps with vegetation indexes allowing to assess the vegetal strength and the foliar mass of sugarcane crops sown in a pellic vertisol soil in dry conditions and under a tropical weather.

II. Vegetation Indexes

The four VI most employed and derivable from a tri-band multispectral sensor are NDVI, GNDVI, and SAVI (Candiago et al., 2015). The research works related with these indexes indicate that the reflectance in the blue and red zones of the spectrum are low (Bachmann, Herbst, Gebbers, & Hafner, 2013), whilst in the green region, a peak is observed. Within the infrared range, the reflectance is higher than in the visible band, as Figure 1 presents.

Figure 1
Reflectance of the green vegetation depending on its strength

The NDVI is a concept integrating the contrast of the high absorbance (low reflectance) in the red band of the visible spectrum with the high reflectance of the near infrared, such as Equation 1 presents.

(1)

where:

NIR = spectral value of the near infrared band; and

R= spectral value of the red band.

The NDVI values are between -1 and 1, in this range, a value near 1 corresponds to a vegetal specimen with high strength and a value of -1 to areas without vegetation —such water, ice, sand, snow, or clouds— (Best, León, & Claret, 2005; Chuvieco, 2000). The common range for vegetation is from 0.2 to 0.9 as follow: 0.2 to 0.3 for brushes and grass; and 0.4 to 0.9 for forests and crops (Pettorelli et al., 2005). Due to the ease of use and its relation with many parameters of the ecosystem, the NDVI usage has been generalized in several ecosystems for supervising labors related with the vegetation dynamics or phonologic changes of plants through time, with the biomass production, with the changes in the pasture conditions, and with the classification of the soil occupancy and moisture (Virlet, Costes, Martínez, Kelner, & Regnar, 2015; Lopes & Reynolds, 2012; García & Herrera, 2015).

On the other hand, the GNDVI is an index of the “green” plant or its photosynthetic activity commonly used to determine the water absorption and the nitrogen in the crop foliage. This index is defined by Equation 2.

(2)

where:

NIR = spectral value of the near infrared band; and

green= spectral value of the green band.

The GNDVI replaces the red band (R) for the data gathering process for NDVI; it performs this replacement with a very specific light band in the green range, entailing in the gathering of additional information. In opposition to indexes such as NDVI that cannot be determined without a red band, alternative indexes such as GNDVI have information content and similar value and they do not need this band (Gitelson, Kaufman, & Merzlyak, 1996).

The gathering of frequent GNDVI allow the watering optimization and indicates when there is a lack of this resource or when it varies through the field. This information —together with a GPS— can be used to determine efficient solutions in some problematic areas of the crop, it also helps to provide application and correction material to make the fields more uniform relative to the water retention and usage. Furthermore, it also allows to implement an accurate variable rate diffusion technique (Dennis, Wright, & Philip, 2003; Hunt et al., 2010).

Some alternative VI such as the SAVI (Huete, 1988)and the Optimized Soil Adjusted Vegetation Index [OSAVI] (Rondeaux, Steven, & Baret, 1996) have been proposed to minimize the soil impact in the calculation of the VI in such areas where the vegetal coverage is poor and the soil surface is exposed. These variations are part of the vegetation indexes commonly used in remote sensing applications with the normalized vegetation index (Salamí, Barrado, & Pastor, 2014; Trotter, Frazier, Trotter, & Lamb, 2008; Hatfield & Prueger, 2010). The SAVI varies in the ranges of -1 and 1; values trending to -1 correspond to a low coverage of the green vegetation and —as expected— values near 1 represent high coverage. Its calculation is made via Equation 3.

(3)

where:

NIR = spectral value of the near infrared band;

R = spectral value of the red band; and

L = corrective factor of the soil reflectance.

It is important to mention that L is a parameter with extreme values from -1 to 1, where -1 is used in areas with a dense vegetation coverage; the 1 value is used when there is no green coverage in the terrain. The value of 0.5 is used for a moderate coverage, but it is possible to use intermediate values depending if the trend is going to one extreme or the other (Candiago et al., 2015). In this project, we used a value of L=0,2.

III. Equipment for the gathering and processing of agricultural images

The gathering of aerial images in the NIR and IR spectrum requires a UAV and a multispectral sensor. The UAV we used was the USENSE-X8, which has a solid reputation relative to reliability and robustness in the market. It is a fixed-wing vehicle, it has an automatic pilot system based on open source technology, and it has a design allowing a flexible operation and a secure implementation under several environments and meteorological conditions. The selection of this UAV was validated with AZCUBA, the sugar company of the country. The selected multispectral camera was the Parrot Sequoia one with a weight of only 107 grams, making it a compact but powerful element. It has four filters to analyze nutrients and biomass of the crops and it also has an RGB lens of 16 megapixels. This last allow the generation of highly detailed maps, ideal for counting the small and sprout plants, it also has a sensor capturing the incident light, useful to store the illumination conditions and to calibrate the four multispectral sensors. These multispectral sensors allow the gathering of precise vegetation indexes such as the ones required for this project.

IV. Procedure for the gathering and processing of the multispectral images

In general, the working flow that eases the gathering and processing of the multispectral images (Figure 2) is as follow: when receiving a service request, the area to overfly and take the photos is studied, if the terrain has unfavorable features for the development of the mission —such as large ruggedness or strong winds—, the mission is discarded; otherwise, the flight plan is traced using MissionPlanner. Once the area is accepted, the status of the UAV and of the multispectral camera are checked; if some issues arise, the possibility to fix them in the terrain is assessed and if this is not possible, the mission is cancelled. If both elements are ready and the weather conditions are good, the flight and the concurrent image gathering are performed; after, the information is extracted from the camera to preprocess it and the multispectral orthomosaic is generated by using Agisoft’s Photo Scan. Finally, the vegetation indexes are calculated with the corresponding bands for each formula.

Figure 2
Workflow for the calculation of vegetation indexes

MissionPlanner —the software used to plan the flight— is a ground control station with all the functions of the Multi-Platform Autopilot [MPA] open source project; it has an interface to establish the gain and damping of the vehicle following strategy, which is programmed in the ArduPilot (Hernández-Morales, Valeriano-Medina, Hernández-Julián, & Hernández-Santana, 2017). It also proposes a series of cameras —included the one employed in this project— to calculate the flight path and the altitude the UAV should fly, which is fundamental to obtain images with the adequate resolution.

The polygons to photograph, the flyby to follow, the picture overlapping requirements in the flight direction (longitudinal) and parallel to the flight (transversal), and the camera configuration (see Figure 3) are established in the software screen. In this project, the flights were performed at an altitude of 40 meters with a maximum speed of 6 meters per second; this led to a Ground Sample Distance [GSD] of 3 cm/pixel and an adequate overlapping.

Figure 3
Planning screen of the MissionPlanner mission

In this particular project, the overfly terrain was a sugarcane plot with a plant age of 6 months (large growing period) sowed with 25 diversities in a random block design in a non-irrigated land condition. Furthermore, the soil was pellic vertisol under a tropical climate (typical of Cuba) and each diversity takes up to 41 square meters (6.4 x 6.4 m) made by 5 furrows with a border furrow in each side. The field dosage was 212.76 kg/ha of triple superphosphate and 333.33 kg/ha of KCl.

The tool selected for the calculation of the mentioned VIwas QGIS on its version 2.4[Quantum GIS] together withan open source GIS for GNU/Linux, Unix, Mac OS, Windows, and Android platforms, allowing to handle vector raster formats through its libraries and with databases. The fact of being an open source project allows the integrations of plugins developed in C++ and Python. The employed version also has full support for multi-thread processing (processors with at least 4 cores are recommended). QGIS calculates the VI from the SAGA ToolBox following this sequence: The multispectral data —where the individual bands are separated raster files— are unified in a unique pile of layers having all the bands; the image stacked in layers is loaded in the SAGA radiometric indexes tool; the employed bands to calculate the indexes are defined; the index to use is selected (in this case is NDVI); and the calculation of the index is performed.

The obtained images present three spectral bands: NIR, green, and red (in this order). First, the multispectral image of the field where the VI will be calculated is obtained (Figure 4), after, the spectral bands to use are extracted via the raster calculator (Figure 5). In this case, our interest is in the red, green, and NIR bands, which are required to calculate the NDVI, GNDVI, and SAVI. Figure 4 corresponds to the multispectral image of the overflown sugarcane field at an altitude of 40 meters and a speed of 6 meters per second. The red pixels are cultivated parcels, whilst the dark sectors are zones without plants or correspond to soil.

Figure 4
Multispectral image of the overflown field to calculate the vegetation indexes

Figure 5
QGIS’s raster calculator

V. Relation between the spectral reflectance of the sown fields and the amount of sugarcane stalks

In Figure 6, the reader can appreciate how the reflectance is low for plants in the red spectrum (a), larger in the green spectrum (b), and even more in the NIR (c). The NIR band is fundamental in the calculation of the indexes; if a transformation towards a space with colors is performed, it would be possible to appreciate the variation through the strength and vitality of the plants. The task is to create a precise prescription for the field considering these variations by calculating a vegetation index —in this case, NDVI—. The real NDVI image in gray scale is shown in Figure 7.

Figure 6
Spectral bands extracted from the multispectral image: red (a); green (b); y NIR (c)

Figure 7
NDVI Image

Frequently, this image is transformed to a “pseudo-color” format to ease its interpretation. For our case, the colors have the trend to follow the standard color spectrum where the purple areas are soil or zones with dry grass and the red zones show the vegetal sections (Figure 8). Both parts of Figure 8 (a and b) present a high correlation level between the studied crop type and the NDVI readings. On its initial growing stage, sugarcane has elevated chlorophyll levels and show high NDVI values within its fundamental reflectance features. In this experiment, the maximum obtained NDVI value was 0.94, with an average of 0.7; which indicates a young and healthy crop. If the sugarcane had more time of being sowed, it would be in the ripening period and the NDVI values would be smaller. The high vegetal strength of the studied field and deducted from the NDVI image was validated by specialists in the terrain through the gathering and laboratory analysis of multiple samples This was done through the count in the number of stalks per parcel, and through the measurement of the sugarcane average height. Furthermore, we were able to observe the presence of weed in the furrows; it could be eliminated via the application of herbicides in the zones with larger grass reflectance. The image at the right (Figure 8, right) shows a legend equivalent to a vegetation general coverage; hence, the NDVI values of the young sugarcane are similar to the ones that the trees have, as per the consulted scientific literature

Figure 8
NDVI Orthophoto with predefined colors as per the scale generated by the software

The NDVI can be converted into variable application maps for its usage with farm equipment supporting the so-called Variable Rate Technologies [VRT]; i.e., the machinery is able to ration supplies —such as fertilizers to apply per zone— by using a GPS. One example is the nutrients application in a single field, where the farmer can apply up to 60 pounds of fertilizer for the zones with difficulties (green), 50 pounds to the medium areas (yellow), and 40 pounds to the healthy sections (yellow-red). This will entail in a considerable reduction of the associated costs in the fertilizers usage plus the reduction in the contamination of the water-table layer, and it would increase the crop yield.

The soils can negatively affect in the quality of the performed photogrammetry to a crop due to the reflectance that they have in the NIR band of the spectrum. In order to counteract this measurement error, we used the SAVI index with the parameter L = 0.2. This, since the photographed crops presented a high strength as Figure 9 (left) depicts. Right side of Figure 9 shows the equivalent legend to a general vegetation coverage.

Figure 9
SAVI images with predefined colors as per the scale generated by the software

Finally, we calculated the GNDVI index (Figure 10); this index takes advantage of the existing difference between the green and NIR reflectance of plants and provides a measurement with less range but tightly related with its photosynthetic activity. This allows to perform several studies to the vegetation such as the case of detecting nitrogen lacks. In this case, the reading values were up to 0.5 with an average of 0.35; i.e., a good photosynthetic activity translated in a level within the standard nitrogen range in the plantation.

Figure 10
GNDVI images with the predefined colors as per the scale generated by the software

The relation between the maps with NDVI and GNDVI indexes and the number of sugarcane stalks per parcel was validated via the dispersion graphs on each experiment. Figure 11 and Figure 12 correspond to dispersion graphs with the following values: R2=0.702and R2=0.685 respectively. Both graphs were generated by using SPSS. On the other hand, the correlation between the SAVI maps for a value of L = 0.2 and the number of stalks was poor, i.e., the calculation of this index should be done again by reducing the value of L towards -1 (due to the high strength level of the sugarcane parcels) until getting better results. Nevertheless, given the high vegetal coverage level present in the field, the soil reflectance effect is not large; hence, the calculation of the SAVI index is not indispensable to minimize the impact of the soil in the calculation of the vegetation indexes.

The reader should consider that the L factor does not have any effect on the NDVI and GNDVI indexes —the ones showing a good correlation between them and the number of stalks per parcel—. Furthermore, it is important to focus that no more tests were carried out with different L values to modify the SAVI index, since the incidence of the soil in the reflectance is low, the calculation of the NDVI and GNDVI indexes is enough to identify parcels with high vegetal density.

Figure 11
Dispersion graph(R2=0.702) –Correlation between the NDVI map and the number of stalks per parcel

Figure 12
Dispersion graph(R2=0.685)- Correlation between the GNDVI map and the number of stalks per parcel

Besides, we performed some linear regressions between the NDVI and GNDVI indexes and the number of stalks in 25 parcels in order to assess the linear regression degree between these variables, as Table 1 shows. The selected method was the Kendall’s rank correlation coefficient. Given the high Kendall Tau values obtained, we can affirm that these two indexes can be directly correlated with the amount of sugarcane stalks and indirectly with the agricultural production. The NDVI and GNDVI index maps also allowed the punctual identification of the parcels with an average of stalks under the rest of the parcels.

Table 1
Kendall correlation between the NDVI and GNDVI indexes and the number of stalks per parcel

VI. Conclusions and Future Work

The NDVI and GNDVI orthophotos generated from the photographs of the Sequoia camera allowed to identify sugarcane parcels with a high density and vegetal strength, results that were corroborated in the terrain through the gathering and laboratory analysis of multiple vegetal samples and through the count of the number of stalks per parcel and average sugarcane height. Consequently, it is possible to affirm that the UAV represents an excellent tool for the evaluation of sugarcane crops due to the ease in mounting multispectral cameras and obtaining high-resolution images represented in 3 cm/pixel.

The obtained index maps allowed us to identify growing problems in the furrows and take actions in punctual areas of the crop, which can reduce the time and work not only in hard-to-reach working zones, but also in large sugarcane plantations. The identification of sugarcane parcels with a low vegetal density from the spectral response ease the farmers the design of fertilizing variable dosage strategiesdepending of the sowing needs. This seeking towards the homogenization of performance and quality.

In order to improve the obtained results in the cultivated zones, the need to perform complementary studies including variables such as the soil analysis supporting the identification of specific spectral response zones has arisen. Furthermore, studies about the monitoring in different period of times and supervision in the evolution of the detected zones as per the corrective measurements applied is also another need that has appeared.

References

Bachmann, F., Herbst, R., Gebbers, R., & Hafner, V.V. (2013). Micro UAV based georeferenced orthophoto generation in VIS+NIR for precision agriculture. In: Proceedings of the UAV. Remote Sensing and Spatial Information Sciences, (Vol. 40. pp. 11-16).

Basso, B. (2014). Perspectivas y avances del uso de UAV en AP en USA. Retrieved from: https://inta.gob.ar/sites/default/files/script-tmp-inta_g1-perspectivas_y_avances_del_uso_de_uav_en_ap_e.pdf

Best, S. & Zamora, I. (2008). Tecnologías aplicables en agricultura de precisión: uso de tecnología de precisión en evaluación, diagnóstico y solución de problemas productivos. Santiago de Chile: Fundación para la Innovación Agraria.

Best, S., León, L., & Claret, M. (2005). Use of precision viticulture tools to optimize the harvest of high quality grapes. Proceedings of the fruits and nuts and vegetable production engineering TIC (Frutic05), (pp. 249-258).

Campo, L., Corrales, J. & Ledezma, A. (2015). Remote sensing for agricultural crops based on a low cost quadcopter. Sistemas & Telemática, 13(34), 49-63. https://doi.org/10.18046/syt.v13i34.2092

Candiago, S., Remondino, F., De Giglio, M., Dubbini, M., & Gattelli, M. (2015). Evaluating multispectral images and vegetation indices for precision farming applications from UAV images. Remote Sensing, 7(4), 4026-4047.

Chuvieco, L. (2000). The use of qualitative airbone multispectral imaging for managing agricultural crops: a case study in South- Eastern Australia. Aust. J. Exp. Agric, 40, 725-738.

Dennis, L., Wright, J. & Philip, R. (2003). Managing protein in hard red spring wheat with remote sensing [paper in The 6th Annual National Wheat Industry Research Forum, 2003. Retrieved from: https://www.researchgate.net/publication/252140884_Managing_Grain_Protein_in_Wheat_Using_Remote_Sensing

Gago, J., Douthe, C., Coopman, R., Gallego, P., Ribas-carbo, M., ... & Medrano, H. (2015). UAVs challenge to assess water stress for sustainable agriculture. Agricultural Water Management, 153, 9-19.

García, C. & Herrera, F. (2015). Percepción remota en cultivos de caña de azúcar usando una cámara multiespectral en vehículos aéreos no tripulados [paper in: Anais XVII Simpósio Brasileiro de Sensoriamento Remoto-SBSR. Retrieved from: http://www.dsr.inpe.br/sbsr2015/files/p0873.pdf

Gitelson, A. A., Kaufman, Y. J., & Merzlyak, M. N. (1996). Use of a green channel in remote sensing of global vegetation from EOS-MODIS. Remote sensing of Environment, 58(3), 289-298.

Guo, T., Kujirai, T.. & Watanabe, T. (2012). Mapping crop status from an unmanned aerial vehicle for precision agriculture applications. In: International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, (Vol. 39-B1 pp.485-490). ISPRS.

Gutierrez-Rodriguez, M., Escalante-Estrada, J. A., & Rodriguez-Gonzalez, M. T. (2005). Canopy reflectance, stomatal conductance, and yield of Phaseolus vulgaris L. and Phaseolus coccinues L. under saline field conditions. Int. J. Agric. Biol, 7, 491-494.

Hatfield, J. L. & Prueger, J. H. (2010). Value of using different vegetative indices to quantify agricultural crop characteristics at different growth stages under varying management practices. Remote Sensing, 2, 562-578.

Hernández, L., Rodríguez, E., Martínez, A., Álvarez, H., Kharuf, S., & Morales, L. H. (2016). Levantamiento fotogramétrico de la UBPC “Desembarco del Granma” utilizando aviones no tripulados, solución de bajo costo para la agricultura nacional. In: VII Edición de la Conferencia Científica Internacional sobre Desarrollo Agropecuario y Sostenibilidad 2016. Santa Clara, Cuba: UCLV.

Hernández-Morales, L., Valeriano-Medina, Y., Hernández-Julián, A. & Hernández-Santana, L. (2017). Estudio sobre la estrategia de guiado L1 para el seguimiento de caminos rectos y curvos en UAV. Ingeniería Electrónica, Automática y Comunicaciones, 38, 14-25

Huete, A.R. (1988). A soil-adjusted vegetation index (SAVI). Remote Sensing of Environment, 25, 295-309.

Hunt, E. R., Hively, W. D., Fujikawa, S. J., Linden, D. S., Daughtry, C. S., & McCarty, G. W. (2010). Acquisition of NIR-green-blue digital photographs from unmanned aircraft for crop monitoring. Remote Sensing, 2, 290-305.

Johansen, K., Sallam, N., Robson, A., Samson, P., Chandler, K., Derby, L., ... & Jennings, J. (2018). Using GeoEye-1 Imagery for Multi-Temporal Object-Based Detection of Canegrub Damage in Sugarcane Fields in Queensland, Australia. GIScience & Remote Sensing, 55(2), 285-305.

Lofton, J., Tubana, B. S., Kanke, Y., Teboh, J., Viator, H., & Dalen, M. (2012). Estimating sugarcane yield potential using an in-season determination of normalized difference vegetative index. Sensors, 12, 7529-7547.

Lopes, M. S. & Reynolds, M. P. (2012). Stay-green in spring wheat can be determined by spectral reflectance measurements (normalized difference vegetation index) independently from phenology. Journal of Experimental Botany, 63, 3789-3798.

Marote, M. (2010). Agricultura de Precisión. Ciencia y Tecnología, 10, 151.

Martínez, L. J. (2017). Relationship between crop nutritional status, spectral measurements and Sentinel 2 images. Agronomía Colombiana, 35, 205-215.

Pettorelli, N., Vik, J. O., Mysterud, A., Gaillard, J. M., Tucker, C. J., & Stenseth, N. C. (2005). Using the satellite-derived NDVI to assess ecological responses to environmental change. Trends in Ecology & Evolution, 20(9), 503-510.

Rondeaux, G., Steven, M., & Baret, F. (1996). Optimization of soil-adjusted vegetation indices. Remote Sensing of Environment, 55(2), 95-107.

Salamí, E., Barrado, C., & Pastor, E. (2014). UAV flight experiments applied to the remote sensing of vegetated areas. Remote Sensing, 6(11), 11051-11081

Saxena, L., & Armstrong, L. (2014). A survey of image processing techniques for agriculture. In: Proceedings of Asian Federation for Information Technology in Agriculture (pp. 401-413). Perth, W.A: Australian Society of Information and Communication Technologies in Agriculture. Retrieved from: https://ro.ecu.edu.au/ecuworkspost2013/854

Torres, A., Gómez, A., & Jiménez, A. (2015). Development of a multispectral system for precision agriculture applications using embedded devices. Sistemas & Telemática, 13(33), 27-44. https://doi.org/10.18046/syt.v13i33.2079

Trotter, T. F., Frazier, P., Trotter, M. G. & Lamb, D. W. (2008). Objective biomass assessment using an active plant sensor (Crop Circle), preliminary experiences on a variety of agricultural landscapes [white paper]. Retrieved from: https://www.researchgate.net/profile/Paul_Frazier2

Vibhute, B. S. & Bodhe, S. K. (2012). Applications of Image Processing in Agriculture: A Survey. International Journal of Computer Applications, 52, 34 - 40.

Virlet, N., Costes, E., Martinez, S., Kelner, J. J., & Regnard, J. L. (2015). Multispectral airborne imagery in the field reveals genetic determinisms of morphological and transpiration traits of an apple tree hybrid population in response to water deficit. Journal of Experimental Botany, 66(18), 5453-5465.

Zhao, Y., Della-Justina, D., Kazama, Y., Rocha, J., Graziano, P., & Camargo, R. (2016). Dynamics modeling for sugar cane sucrose estimation using time series satellite imagery. In: Proceedings of the Remote Sensing for Agriculture, Ecosystems, and Hydrology XVIII (pp. 99980J). International Society for Optics and Photonics. https://doi.org/10.1117/12.2242490

Author notes

Additional information

How to cite: Kharuf-Gutierrez, S., Orozco-Morale, R., Aday, O., & Pineda, E. (2018). Multispectral aerial image processing system for precision agriculture. Sistemas & Telemática, 16(47), 45-58. https://doi.org/10.18046/syt.v16i47.3221