Analysis of Growth Characteristics Using Plant Height and NDVI of Four Waxy Corn Varieties Based on UAV Imagery

Abstract

Although waxy corn varieties developed after the 1980s show differences depending on development stages and conditions, studies on the characteristics of waxy corn during the growth stage are rare. The subject of this study was a field survey and unmanned aerial vehicle (UAV) image acquisition of four waxy corn varieties cultivated in Idam-ri, Gammul-myeon, Goesan-gun, Korea. The study was conducted in four stages at intervals of two weeks after planting in 2019. The growth characteristics of each of the four varieties were analyzed using growth curves obtained based on field survey and UAV imagery data. The characteristics of each growth stage of the four varieties of corn, as assessed using normalized difference vegetation index (NDVI) and plant height (P.H.) values, were as follows. The growth model was identified as a model in which three-parameter logistic (3PL) curves reflect the growth characteristics of corn well. In particular, it was found that the variations in growth rate shown by P.H. and NDVI values clearly explain the differences between corn varieties. Among the four cultivars, growth and development first occurred at the early vegetative stage in Daehakchal, followed by Mibaek 2, Miheukchal, and finally Hwanggeummatchal. The variationsin P.H. and NDVI were achieved quickly and earlier in Daehakchal, followed by Mibaek 2, Hwanggeummatchal, and Miheukchal. It was confirmed that these results reflected the characteristics of the fast white-type varieties, while the black-type varieties were delayed, as in a previous study. These results reflect the resistance to lodging that affects the cultivation environment and the response characteristics to nutrients and moisture. It was confirmed that UAV accurately provides growth information that is very useful for analyzing the growth characteristics of each corn variety.

1. Introduction

Corn is one of the world’s three major food crops, along with rice and wheat, and is widely used not only for food, but also for animal feed and bioenergy crops (FAO, 2021). Additionally, corn is one of the crops that can affect food supply and security; the international demand has increased, while the supply has not kept up with the increase in biofuel use. The total global cultivated area is increasing, with varieties adapting depending on the environment and conditions of each region. In addition, because of its high adaptability to different climates, corn is grown across a wider variety of climatic conditions than rice or wheat, and has the advantage of very fast growth due to its high photosynthetic capacity. Another advantage of corn is that it has a strong soil absorption capacity and a large fertilization effect, while the yield varies greatly depending on the amount of fertilization and the fertilization method.

In South Korea, where there are many areas of reclaimed land, plans to expand the corn cultivation are in progress in order to diversify the use of reclaimed land and to ensure food self-sufficiency. In domestic environments, various varieties have been developed and grown due to the continuous supply and distribution of edible corn (Lee et al., 2018; Baek et al., 2020). Corn is classified into general corn, sweet corn, waxy corn, and fried corn, among others, depending on the characteristics of the ears. Also, waxy corn is commonly used to produce starch which is used as an industrial raw material in the form of amylopectin together with glue. In South Korea, waxy corn has a good texture and palatability, so it is harvested as green corn and is widely used as a snack. With the development of corn breeding research, various types of waxy corn, such as “Daekchal”, which is widely cultivated in two crop cycles, have been developed (Kim et al., 2014).

From 1989, hybrid sweet corn varieties with good quality and high yield, waxy corn, and Chodang corn began to be cultivated and distributed. In particular, the National Institute of Crop Science (NICS) and Rural Development Administration (RDA), through continuous breeding research, have been developing and distributing different varieties, such as “Chalok 1” (Park et al., 2007a; Park et al., 2007b; Lee et al., 2018). At the beginning of the breeding development, corn was cultivated in the highlands, but due to new cultivation technologies and improved varieties, corn is now grown in lowlands. For the successful cultivation of corn, it is important to implement cultivation methods, selecting suitable fertilization amounts, fertilization methods, and planting densities for the growing environment (NICS, 2019). In addition, it is important to follow a high-value-added agriculture approach involving growth management and quality control, so that corn can be shipped before the rainy season begins; therefore, analyzing the characteristics related to the growth and development of corn and suggesting a cultivation method suitable for each variety will contribute to accelerating growth and increasing yield.

Conventional growth models, such as the exponential (logistic) function and Gompertz and Richards equations, have been widely used to explain various biological growth processes (Richards, 1959; Ayiomamitis, 1986; Werker and Jaggard, 1997). Among them, the logistic growth model has various applications, ranging from population dynamics to the dose-response relationship (Oliver, 1964; Vandamme et al., 2021). The yield and marketability of cultivated corn are affected by regional and environmental conditions, such as climate, as well as cultivation techniques; therefore, in order to suggest a cultivation method for each variety suitable for local conditions, it is necessary to prepare countermeasures through analysis using the corn growth model of variables such as the growth characteristics of each variety, environmental conditions, and cultivation methods.

Major corn exporting countries such as the United States are implementing corn production management and quality control methods through various support services at universities in each production region of each state (Ransom and Endres, 2020; Cartwright et al., 2021); however, the reality is that Korea has limited support systems because of the low corn production rates and small cultivation area. So far, corn growth characteristics have been assessed using point or area methods to analyze environmental growth conditions and growth status (NICS, 2019); however, for the management of corn crops distributed and cultivated over a wide area, it is desirable to diagnose the effects on corn growth at each growth stage to determine solutions and countermeasures. For this purpose, an appropriate method should be used to identify, diagnose, and determine spatial corn growth information. Methods involving the use of unmanned aerial vehicles (UAVs) and sensors, which have been recently activated and used in various fields in agriculture, will be useful for corn management (Han et al., 2018; Niu et al., 2019; Lee et al., 2020; Zhou et al., 2020).

The purpose of this study was to select a suitable model for the growth characteristics of corn and to find a method for analyzing differences in growth characteristics during the growth process for each variety using field surveys and UAVs.

2. Materials and Methods

1) Study Area

This study was conducted in Idam-ri (36°52′03″N, 127°52′02″), Gammul-myeon, located in the northeastern part of Goesan-gun, Chungcheongbuk-do (Fig. 1). The target area measures approximately 580 ha. The Dalcheon River flows through the study site, so there is no difficulty in accessing agricultural water. The average annual temperature for the target area is 11.2℃, the average monthly temperature in August is 25.2℃, and for January is -4.3℃, showing a distinct continental climate with a temperature difference of 29.5℃. The average annual precipitation in Goesan is 1271.4 mm, which is similar to that for the rest of Korea. By season, most of the precipitation is concentrated in summer, with 204.0 mm precipitation (16%) in spring, 747.4 mm (58.7%) in summer, 258.7 mm (20.3%) in autumn, and 63.1 mm (5.0%) in winter (KMA, 2019).

Fig. 1. The location of the study area. The red solid line indicates the area where the research and field survey were conducted.

Goesan-gun is an area where Daehakchal was developed, and currently edible corn types such as Mibaekchal and Miheukchal are mainly grown. The study site was selected as a test bed for the field cultivation of Hwanggeummatchal, a corn variety newly developed by the RDA, and the local adaptability was evaluated. The corn cultivation area of Gammul-myeon, Goesan-gun, is 141.4 ha based on the 2017 agricultural business registration. This region has one of the highest ratios of corn cultivation per area (MAFRA, 2017).

2) Materials

In this study, a test field was constructed to compare the growth characteristics of 4 waxy corn varieties. The experimental materials were four types of corn grown from domestic genetic corn resources: Hwanggeummatchal, Mibaek 2, Miheukchal, and Daehakchal. The cultivation packaging and characteristics for each variety are shown in Table 1 and Fig. 2. Planting was carried out for 6 days from April 15 to 20, 2019. The planting distance at the time of corn planting was 90 cm × 40 cm, and two-grain seeding was carried out.

Table 1. Cultivation area, planting distance, and grain color characteristics for the four corn varieties

Fig. 2. Cultivation field locations for each of the four corn varieties. The red symbols indicates the cultivar for each field. The image used in the figure is an RGB image of the research target area using a UAV.

Here, Hwanggeummatchal was cultivated in a relatively narrow area compared to the other varieties because it was jointly tested and cultivated at the Goesan-gun Agricultural Technology Center and National Institute of Crop Science of the RDA. Since the waxy corn genetic factor is recessive compared to that of general corn, isolation cultivation is recommended to maintain cultivar purity (NICS, 2019). Therefore, in this study, as far as possible, the cultivation sites for each variety were separated and selected so that there was little interaction between the varieties.

In this study, UAV images were acquired using a fixed-wing UAV, eBee Plus (Sensefly, Cheseaux-sur-Lausanne, Switzerland). The UAV-mounted sensor used Sequoia (Parrot, Paris, France) consisting of four spectral bands of Green, Red, Red-edge, and Near Infrared (NIR). UAV image data acquisition was conducted four times on May 15, May 30, June 25, and July 9, 2019. Image preprocessing and matching were performed using the Pix4D Mapper (Sensefly, Cheseaux-sur-Lausanne, Switzerland) program. For radiating reflectance correction was performed before and after observation using the calibrated reflectance panel provided by Sequoia with the multi-spectral sensor, and the spatial resolution of the matched image is 10 cm/pixel.

Field growth survey was conducted for 6 fields shown in Fig. 2. First, one field was divided into 5 sectors. Second, the growth data was measured after selecting 4 corns in one sector as shown in Fig. 2. Third, the location of the growth survey was matched with the UAV image using the Real Time Kinematic-Global Positioning System (RTK-GPS) (HI-TARGET, Guangzhou, China). The corresponding area was set in consideration of the planting interval of corn, and the average NDVI value was extracted. Growth survey was conducted simultaneously with the acquisition of UAV image data.

NDVI images were obtained using the method performed described by Rouse et al. (1974), and Lee et al. (2021) (Table 2). Where, ρNIRis reflectance of near-infrared wavelength and ρREDis reflectance of red wavelength.

Table 2. NDVI used in this study

In the central region, which was the cultivation area, in the case of ordinary cultivation (direct sowing), sowing is generally carried out in mid-April, while early-ripening varieties are harvested in mid-July and mid- and late-ripening varieties are harvested from the end of July to early August.

3) Corn growth model

The reason for using the logistic model growth to model the growth characteristics of corn is because corn grows very fast in the initial stages and then decreases in the late stage of the growth phase (Pienaar and Turnbull, 1973; Birch and Shaw, 1997). In this study, a three-parameter logistic (3PL) equation in which the minimum value is 0 was used, assuming that the initial corn growth was 0 (Zeide, 1993).

As a growth condition, P.H. and NDVI were used, whereby t was the elapsed time (days) after planting. The model adopted as the growth model is shown in Equation (1).

P.H.(t) or NDVI(t) = c(1 + ae^–bt)^{1 /(1–m)}       (1)

Here, if m is expressed in logistic form by applying (2), the modified logistic growth model is the same as Equation (2).

\(\text { P.H.(t) or } \operatorname{NDVI}(t)=\frac{c}{1+a \times \exp ^{(-b t)}}\)       (2)

Here, c is the maximum growth value of corn, and a and b are the parameters for each variety (a>0, b>0).

The growth rates of corn (y′(t)) were obtained using the first derivative of Equation (2).

\(\mathrm{y}^{\prime}(\mathrm{t})=\frac{d \mathrm{P} \cdot \mathrm{H}}{d t}=b(1-\mathrm{P} \cdot \mathrm{H} \cdot(t) / c)\)       (3)

\(\mathrm{y}^{\prime}(\mathrm{t})=\frac{d \mathrm{NDVI}}{d t}=b(1-\mathrm{NDVI}(t) / c)\)       (4)

4) Evaluation of the accuracy

Being a variable in plant growth, using P.H. and NDVI, the accuracy was compared using a measurement value based on the absolute or square error between the actual value and the predicted value (Calka and Bielecka, 2020).

(i) The mean absolute error (MAE) and root mean squared error (RMSE) were used to compare prediction methods for the same data set.

\(\mathrm{MAE}=\frac{1}{N} \sum_{i-1}^{N}\left|p_{i}-\widehat{p}_{i}\right|\)       (5)

\(\operatorname{RMSE}=\sqrt{\frac{1}{N} \sum_{i=1}^{N}\left(p_{i}-\widehat{p}_{i}\right)^{2}}\)       (6)

Here, p_i is the measured value and \( \widehat{p}_{i}\) is the estimated value obtained by the predictive model at itime step. Nis the estimated number during the growing period.

(ii) The mean absolute percentage error (MAPE) was used for percentage accuracy based on measurements to maintain scale independence.

\(\text { MAPE }=\frac{100}{N} \sum_{i=1}^{N}\left|\frac{p_{i}-\hat{p}_{i}}{p_{i}}\right|\)       (7)

3. Results and Discussion

1) The P.H. and NDVI variation characteristics

Corn is very adaptable to different soil and climate conditions after planting, so it grows rapidly over time. Corn also goes through several growth stages before it penetrates the soil surface, matures, and is released. The corn growth stage is divided into two stages: a vegetative stage (V) and a reproductive stage (R) (Ransom and Endres, 2020; Cartwright et al., 2021). The vegetative stage (V) is further subdivided depending on the number of leaf appearances, while the reproductive stage (R) is divided into stages based on changes to the corn ears.

In this study, P.H. and NDVI results were obtained using data obtained from UAV and field growth surveys. The obtained results are shown in Fig. 3 and Fig. 4, respectively, using a boxplot for each variety.

Fig. 3. The Plant Height (P.H.) variation characteristics are shown depending on the days after planting (DAP). The number, n, of used datapoints for each field is 80.

Fig. 4. NDVI variation characteristics depending on the days after planting (DAP). The number, n, of used datapoints for each field is 80.

As shown in Fig. 3, P.H. values, which were investigated for six fields containing four varieties, were initially homogeneous, with no differences between varieties in the early vegetative stage (V). At 45 days after planting, differences by cultivar began to appear. The white varieties Mibaek 2 (F3) and Daehakchal (F4 and F5) showed rapid P.H. growth, while the colored varieties Miheukchal (F6) and Hwanggeummatchal (F1 and F2) showed low P.H. values. This trend continued until the second half of the vegetative stage (V). During the reproductive stage (R), there were no significant differences between any of the six plots of four varieties, with similar P.H. distributions being shown.

The NDVI variations over time showed a slightly different trend than those for P.H., as shown in Fig. 4. At the beginning of the vegetative stage (V), the colored strains Hwanggeummatchal (F1 and F2) and Miheukchal (F6) showed constant and low NDVI values. On the other hand, Mibaek 2 (F3) and Daehakchal (F4 and F5), which are white varieties, presented a relatively wide range of NDVI values and showed high-growth characteristics. In the middle of the vegetative stage (V), the colored strains Hwanggeummatchal (F1 and F2) and Miheukchal (F6) showed low NDVI values, while the white strains Mibaek 2 (F3) and Daehakchal (F4 and F5) tended to have a high NDVI distribution of 0.6 or higher. From the late vegetative stage (V) to the reproductive stage (R), the NDVI for Hwanggeummatchal (F1 and F2) was high, while the values for Mibaek 2 (F3) and Miheukchal (F6) were low.

The P.H. and NDVI results were interpreted, indicating that the characteristics set for the purpose of breed improvement at the time of breeding for each variety appeared as variations in the growth curve. In particular, Hwanggeummatchal (F1 and F2) and Miheukchal (F6), the colored strains, achieved the goals of lowering the P.H. to increase resistance to lodging during the growth process and of activating growth in the late vegetative stage (V).

2) Regression analysis of P.H. and NDVI values

Regression analysis of the NDVI and P.H. values was performed by applying the characteristics for the corn growth stages, divided into the vegetative stage (V) and reproductive stage (R). The analysis results for all four varieties showed a linear relationship with growth stage, as shown in Fig. 5. In the vegetative stage (V), for all four varieties, NDVI and P.H. the value was 0.93. However, a relatively low of 0.66 was shown during the reproductive stage (R), which represents the variation stage for the corn ears. As a result of the regression analysis, the initial growth changes for the four corn varieties were very similar, although the differences were reflected in the reproductive stage (R) due to differences in ear formation, harvest, and corn use.

Fig. 5. Regression analysis of Plant Height (P.H.) and NDVI values by vegetative (V) and reproductive (R) growth stages of four corn varieties for the time series data.

Table 3 and Fig. 6 show NDVI and P.H. regression analysis results. The results obtained for each cultivar were very high in the vegetative stage (V) and low in the reproductive stage (R). The colored strains Miheukchal (F6) and Hwanggeummatchal (F1 and F2) showed the highest values in the vegetative stage (V), while the white strains Mibaek 2 (F3) and Daehakchal (F4 and F5) showed relatively low values. In the reproductive stage (R), the correlation for Hwanggeummatchal (F1 and F2) was the highest, followed by Miheukchal (F6), while Daehakchal (F4 and F5) and Mibaek 2 (F3) were relatively low.

Table 3. Regression analysis results for Plant Height (P.H.) and NDVI for six fields containing four corn varieties

Fig. 6. Regression analysis of Plant Height (P.H.) and NDVI values for six fields and four corn varieties.

The obtained results reflect the process of corn growth, in which the NDVI and P.H. values increase as the corn grows rapidly in the vegetative stage (V). In addition, the saturation of NDVI values during the reproductive stage (R) is reflected in P.H. in the process leading to partial harvesting of corn and aging, which was interpreted as plant growth beginning to decrease.

3) Estimated corn growth model using P.H. and NDVI

The purpose of this study was to identify the growth characteristics of corn through minimal observation due to difficulties such as weather conditions. Most biologists use models that apply various functions to the growth models of animals and plants, including population growth (Zeide, 1993; Werker and Jaggard, 1997). Among them, many growth models involving logistic functions use 3 to 5 parameters (Zeide, 1993). In general, when three parameters (3PL) are used, the upper asymptote, the ratio parameter, and the time constant are applied. The rate parameter in Equation (2) determines the rate at which corn growth is accelerated at the beginning of the growth period.

The corn growth models with which P.H. and NDVI values were processed are shown in Fig. 7 and 8 and Tables 4 and 5, respectively. In Tables 4 and 5, a, b, and care the values of a, b, and cin Equation (2). The corn growth is characterized by rapid growth at first and decreasing growth in later stages as the plants approach maximum growth.

Fig. 7. Comparison of measured values and estimated corn growth model for each field using Plant Height (P.H.).

 

Fig. 8. Comparison of measured values and estimated corn growth models for each field using NDVI.

 Table 4. Evaluation of parameters and accuracy of estimated corn growth model for each field using Plant Height (P.H.)

 Table 5. Evaluation of parameters and accuracy of estimated corn growth models for each field using NDVI

Fig. 7 shows the estimated P.H. values in a corn growth model. The growth model using the 3PL from Equation (2) showed the variations in the vegetative stage (V) and the reproductive stage (R) well. In the vegetative stage (V), the growth rates for Mibaek 2 (F3) and Daehakchal (F4 and F5), which are white varieties, were very fast, while the P.H. values were also high. On the other hand, the colored strains Miheukchal (F6) and Hwanggeummatchal (F1 and F2) presented low P.H. values in the vegetative stage (V) and rapidly increased from the late vegetative stage (V) to the reproductive stage (R). In Table 4, the R², MAE, and RMSE for P.H. showed higher accuracy for the colored strains Miheukchal (F6) and Hwanggeummatchal (F1 and F2) than the white varieties Mibaek 2 (F3) and Daehakchal (F4 and F5). The reason why RMSE was used for accuracy evaluation here was that since the root is added to the square of the deviation, it is possible to directly determine how much the model differs from the actual value. On the other hand, MAPE values showed no significant differences between colored and white strains.

Fig. 8 shows the NDVI increase model estimates for six fields containing four corn varieties. As with the P.H. values, the NDVI showed growth well using 3PL. In the vegetative stage (V), the white varieties Mibaek 2 (F3) and Daehakchal (F4 and F5) showed remarkable growth and the growth curve was high and rapid, whereas the colored varieties Hwanggeummatchal (F1 and F2) and Miheukchal (F6) presented low values and slow progress. The reproductive stage (R) NDVI was saturated in the range of 0.7 to 0.8, and there were no significant variations in any of the four varieties.

Table 5 shows the relevant parameters and accuracy evaluation results for NDVI. For R², MAE, RMSE, and MAPE for NDVI, Daehakchal (F4 and F5) and Miheukchal (F6) showed very high accuracy. On the other hand, Hwanggeummatchal (F1 and F2) showed relatively low accuracy compared to other varieties. This trend was interpreted as reflecting the characteristics of Hwanggeummatchal growing at a time when growth is delayed and the growth variations of other varieties are small.

4) Corn growth rates using P.H. and NDVI

The corn growth rates using P.H. and NDVI were obtained by finding the first derivative of the corn growth model curves obtained in Fig. 7 and 8 and Tables 4 and 5, respectively. As shown in Fig. 9, the growth rate using P.H. was the fastest and highest for Daehakchal (F4), a white variety, with 7 cm growth at approximately 40 days after planting, followed by Daehakchal (F5) and Mibaek 2 (F3). Hwanggeummatchal (F1 and F2) and Miheukchal (F6), which are colored strains, presented high P.H. values in the latter half of the vegetative stage (V). As such, the P.H. growth rates for the white strains were high and fast, whereas the colored strains showed slow and low P.H. growth.

Fig. 9. Growth rate curves for corn obtained by finding the first derivative using Plant Height (P.H.).

Fig. 10 shows a graph of the corn growth rate using first derivative NDVI. As with P.H., the growth rates for NDVI were fast for the white varieties Mibaek 2 (F3) and Daehakchal (F4 and F5), followed by the colored varieties Hwanggeummatchal (F1 and F2) and Miheukchal (F6). The cultivar with the highest NDVI growth rate was Daehangchal (F5), which showed about 0.03 growth on the 37th day of planting, followed by Mibaek 2 (F3). Hwanggeummatchal (F1 and F2) and Miheukchal (F6), which are colored strains, showed NDVI growth rates of 0.02 and 0.014, respectively, about 45 days after planting. NDVI results were also similar to the P.H. growth rates, whereby the growth of the white varieties was high and fast but the colored varieties presented the maximum growth values about 20 days later.

Fig. 10. Growth rate curves for corn obtained by finding the first derivative using NDVI.

As such, the corn growth model and growth rate curves obtained in this study can be used as very useful tools to understand growth differences by corn variety. In particular, it is expected that the model will be used for cultivar improvement to respond to climate change and for cultivation management to improve farmer income and corn quality.

4. Conclusions

The purpose of this study was to analyze the growth characteristics of four corn varieties based on UAV and field survey data. For the growth characteristics of each cultivar, the estimated growth model was determined using P.H. and NDVI, and the growth rate for each cultivar was presented. The estimated corn growth curves were reflected the growth characteristics of the four varieties well using the 3PL (three-parameter logistic) approach.

The characteristics of each of the four corn varieties as indicated by the obtained P.H. and NDVI are summarized as follows. The four varieties of corn showed changes in vegetative stage (V) and reproductive stage (R) by growth stage when P.H. and NDVI values were used. The corn growth model using P.H. and NDVI data showed 3PL curve variation characteristics. The estimated corn growth model using P.H. and NDVI data and the growth rates obtained using the first derivative variation clearly explained the differences in growth characteristics between corn cultivars. The growth rate obtained using P.H. data was the fastest and highest for the white variety Daehakchal (F4), followed by Daehakchal (F5) and Mibaek 2 (F3). Hwanggeummatchal (F1 and F2) and Miheukchal (F6), which are colored strains, showed high formation in the second half of the vegetative stage (V). As such, the corn growth rates were consistent with the results of previous studies, in which white strains progressed quickly and showed high growth, whereas colored strains grew slowly. The variation in NDVI growth rate was rapid in Daehakchal (F4 and F5), although there was only a slight difference, followed by Mibaek 2 (F3), Hwanggeummatchal (F1 and F2), and Miheukchal (F6). These results reflected the growth characteristics of the white varieties being fast and the black varieties being slow.

Variations in P.H. and NDVI among the four corn cultivars were rapid and large in Daehakchal (F4 and F5) and Mibaek 2 (F3), whereas Hwanggeummatchal (F1 and F2) and Miheukchal (F6) progressed slowly. In particular, since the most important characteristics that affect yield are reflected in the maturation period, for the more recently developed varieties, the faster growth occurred over a short period before harvest.