Research on Monitoring Methods for the Appropriate Rice Harvest Period Based on Multispectral Remote Sensing

Cong, Chen; Guangqiao, Cao; Yibai, Li; Dong, Liu; Bin, Ma; Jinlong, Zhang; Liang, Li; Jianping, Hu

doi:https://doi.org/10.1155/2022/1519667

Discrete Dynamics in Nature and Society

On this page

Abstract Introduction Materials and Methods Discussion Conclusions Data Availability Conflicts of Interest Authors’ Contributions Acknowledgments References Copyright Related Articles

Research Article | Open Access

Volume 2022 | Article ID 1519667 | https://doi.org/10.1155/2022/1519667

Research on Monitoring Methods for the Appropriate Rice Harvest Period Based on Multispectral Remote Sensing

Chen Cong,^1,2Cao Guangqiao,²Li Yibai,²Liu Dong,²Ma Bin,²Zhang Jinlong,²Li Liang,²and Hu Jianping¹

Academic Editor: Juan L. G. Guirao

Received03 Mar 2022

Accepted25 Jun 2022

Published30 Jul 2022

Abstract

Rice production is highly seasonal, and the timing for conducting harvest operations has a relatively significant impact on yields, especially the appropriate timing of the harvesting operations. Using scientific methods to monitor the appropriate harvest period can effectively reduce harvesting losses. Most of the existing studies in this field focus on yield changes that occur during the harvest period of rice and also use multispectral remote sensing for inversion of the crop canopy. However, there has been no research on combining the yields and spectral characteristics of the harvest period. In this article, a monitoring method for the appropriate harvest period of rice based on multispectral remote sensing was established. Remote sensing data of the rice canopy were acquired using UAV equipped with multispectral (550 nm, 660 nm, 735 nm, and 790 nm) cameras, and physiological characteristic data of rice moisture content, thousand-grain weight, and tassel ratios were determined simultaneously. Single-band and combined-band spectral reflectance were introduced as model input variables to compare the BP neural network, SVM, and decision tree; a quantitative inversion model of the physiological characteristics of rice was established. The accuracy of the inversion model with different input variables and model methods was evaluated, and the optimal inversion model for rice moisture content and thousand-grain weight was selected, which could be used to determine whether the crop was harvested during the appropriate period. The continuous monitoring of two rice varieties (South Japonica 46 and South Japonica 5055) indicated that the moisture content of South Japonica 46 decreased linearly during the 20-day trial period, while its thousand-grain weight showed a trend of increasing first and then decreasing. The moisture content of South Japonica 5055 showed a fluctuating downward trend, and its thousand-grain weight also showed a trend of increasing first and then decreasing. The spectral reflectance at 735 nm increased gradually, and that of both rice varieties increased from about 35% at the beginning of the trial to about 75% in the later stage. The spectral reflectance at the other three bands did not show significant change with the harvest date. The inverse model of spectral reflectance with rice thousand-grain weight and moisture content was established using the BP neural network, SVM, and decision tree. In single-band inversion models, the regression effect at 735 nm was relatively good in the BP neural network model, with the highest determination coefficients of spectral reflectance and thousand-grain weight of both rich varieties and the smallest RMSE of prediction results. In combined-band inversion models, the regression effects at 550 + 660+735 nm and 660 + 735+790 nm were the best in the SVM and thousand-grain weight inversion models.

1. Introduction

Since ancient times, “doing farm work in the right season” has been one of the basic principles agricultural production in China has followed. Lü’s Spring and Autumn Annals mentioned that “the right season is the most valuable where agricultural planting is concerned.” Carrying out field operations at the right time is a prerequisite for stable yields. Rice production is highly seasonal, and timely operation has a relatively significant impact on its yields, especially the appropriate timing of the harvest [1, 2]. Harvesting too early results in incomplete development of rice, high grain moisture content, low dry matter content such as starch, low thousand-grain weight, and a high percentage of broken kernels (PBK) during rice processing [3]. Harvesting too late, on the other hand, decreases the thousand-grain weight, increases the probability of yield reduction or even crop failure in case of adverse weather [4], and makes crops more vulnerable to attacks from insects, birds, mammals, and microorganisms [5], leading to yield reduction [6, 7]. Timely harvesting is essential to reducing the losses of rice during the harvest process and guaranteeing the quality of agricultural products [8]. The scientific method for monitoring the appropriate harvest period of rice is of great practical significance in promoting the fine management of rice harvesting operations and reducing harvesting losses.

The yields and quality of paddy rice harvested during the appropriate harvest period are significantly high. Li proposed that the harvest day with the lowest timeliness loss of harvesting crops was the best harvest day [9], and the harvest period with the lowest timeliness loss of total rice harvest was the appropriate harvest period [10, 11]. The academia has conducted some exploration in ascertaining an appropriate harvest period for crops. Regarding the harvest period and crop yield loss at harvest, Wang used the randomized block experiment method to determine the yields of rice planted in the same field and on the same day, measuring the output once every other day and discovered the quantitative relationship between the yield loss and harvest period for rice [1]. TLN Le Ngoc Huyen et al. measured the chemical composition of various organs of silver grass that could be converted into fermentable sugars at different harvest periods and found that greater benefits could be obtained when it was harvested earlier than later [12]. Godin et al. evaluated what effect the harvest period had on the yields of various energy crops converted to biofuels and found that biofuel yield decreased when the crops were harvested too early or too late [13]. Heidari et al. analyzed the effects of the harvest period on fennel yields through experiments, and the results indicated that fennel yields were highest when they were at a state of maturity [14]. Qiao et al. studied the relationship between the yields of soybeans at harvest and the harvest date and obtained the timeliness loss pattern in the mechanized harvesting of soybeans [15]. Du et al. found through experiments that when the harvest time was delayed, the brown rice rate and refined rice rate of japonica rice gradually increased first and significantly decreased after reaching the maximum at 65 days after full heading; the whole refined rice rate fluctuated with no obvious regularity before 65 days after full heading and then significantly decreased after 65 days after full heading [16]. Regarding the methods for determining the characteristics of an appropriate harvest period, spectral analysis techniques, having the advantages of fast analysis and low costs, can properly characterize the internal quality information of agricultural products and have been extensively applied in the nondestructive testing of agricultural product quality and safety [17–19], especially in the detection of fruit ripeness. Xue et al. conducted discriminant analysis on the ripeness of red apricots from Qingxu County using hyperspectral imaging (HSI) (400–1,000 nm), and the correct rate of discrimination of the ELM model established based on the fusion values of characteristic wavelengths and color features reached 93.33% [20]. Li et al. used HSI to collect image information of plum samples with different ripeness, and the discrimination model based on RGB-HSV phase fusion color eigenvalues achieved the correct rate of discrimination of 98.4%, 90.0%, 85.6%, and 90.9% for unripe, semiripe, ripe, and overripe plums, respectively [21]. Wan et al. designed a ripeness detection device based on computer vision technology to determine ripeness by extracting color feature values from the surface of tomatoes with an average accuracy of 99.31% [22]. Santos Pereira et al. predicted the ripeness of papayas using a digital imaging technique and random forest algorithms and then validated the predictions by measuring the hardness of papaya flesh [23]. The aforesaid literature mainly focused on yield changes during the harvest period of rice and also used multispectral remote sensing for inversion of the crop canopy. However, the output and spectral characteristics during the harvest period have not been combined, and the accuracy in the determination results of the appropriate harvest period is doubtful.

This article intends to collect the spectral characteristics of the rice canopy continuously during the harvest period using multispectral remote sensing while testing the thousand-grain weight through the randomized block experiment and, at the same time, establish a regression model of spectral characteristics and thousand-grain weight and, then based on this, determine the appropriate harvest period of rice.

2. Materials and Methods

2.1. Overview of the Study Area

Located at the eastern end of Qixia District, Nanjing is the field trial site of Taiping Village in the Longtan Subdistrict, at the junction of Nanjing, Zhenjiang, and Yangzhou. The surrounding environment is a northern subtropical humid climate zone and maritime climate zone of monsoon circulation, home to significant monsoons, cold winters, and hot summers and has four distinct seasons, sufficient sunshine, and abundant water resources. The annual number of sunshine hours is about 2,100, the annual sunshine rate is about 47%, the frost-free period is 7 months, the amount of annual precipitation is about 1,000 mm, and the annual average number of precipitation days is about 110. The soil mainly consists of horse liver soil with a pH value of 6.4–7.2, organic matter content of 1.80% in the cultivated layer, and rice-wheat rotation, in which rice varieties are mainly South Japonica series.

2.2. Test Materials

The trial was conducted from November 5 to 29, 2020. On November 5, all the rice ears in the trial field were drooping, with more than 90% of rice ears and 60% of rice leaves turning yellow. On November 29, all rice ears and more than 90% of rice leaves turned yellow. The weather conditions during the trial period are shown in Table 1. The trial was suspended during rain showers and the day following the rain. The rice varieties used in the trial were South Japonica 5055 and South Japonica 46. The equipment operated during the trial included XMission UAV (4 rotors and a flight height of 9–120 m) and four-channel multiband cameras produced by Guangzhou Xaircraft Technology Co., Ltd. (spectral bands of 550 nm, 650 nm, 735 nm, and 850 nm), diffuse reflectance white board, particle counter, Suncue TD-6 moisture meter (accuracy of 0.5%) produced by Shanghai Suncue Machinery Co., Ltd., and JY20002 electronic balance (accuracy of 0.01 g) produced by Shanghai Shangping Instrument Co., Ltd.

2.3. Experimental Methods

2.3.1. Multispectral Remote Sensing Experiment of Rice Canopy

Photographs were taken at 12:00 noon every day with XMission UAV equipped with multispectral cameras (550 nm, 660 nm, 735 nm, and 790 nm) at a vertically downward angle from a height of 10 m and 70% overlap, with the image resolution of 72 DPI. A diffuse reflectance white board was placed in the trial area to reflectively calibrate the remote sensing images collected and filter the effect of sunlight intensity differences on the image data. A route that got full coverage of the trial field was set by using the XMission flight control system. The UAV remote sensing platform flew along the route autonomously and took photos in a loop at a frequency of 3 s/photo. A specific view of the trial is shown in Figure 1.

2.3.2. Rice Moisture Content and Thousand-Grain Weight Test

Five plants from both South Japonica 5055 and South Japonica 46 were selected using a random sampling method at 12:00 noon every day, and the moisture content (%) of the rice was measured using a moisture meter. Each collected sample was stored in a refrigerator at 0°C. After all the samples were collected, they were placed in an oven at 60°C for 24 hrs; 1,000 grains of each variety were taken per day using a grain counter and weighed (g), and this process was repeated 3 times.

2.3.3. Multispectral Remote Sensing Image Processing and Spectral Reflectance Extraction

The remote sensing images were corrected and stitched using the XMission supporting image processing software. The areas filled with rice in the stitched remote sensing images were selected as the areas of interest (AOIs), and about 20 AOIs were randomly cropped in each plot image. Subsequently, the spectral reflectance (SR) of the AOI was calculated using the spectral data of the diffuse reflectance whiteboard according to the method in the following equation:where (%) represents the spectral reflectance rate of the d-th AOI; represents the spectral reflectance value in the i-th row and i-th column of the d-th AOI; represents the number of rows in the d-th AOI; represents the number of columns in the d-th AOI. represents the spectral reflectance value in the i-th row and j-th column of the diffuse reflectance whiteboard control area; represents the number of rows in the diffuse reflectance whiteboard control area; represents the number of columns in the diffuse reflectance whiteboard control area.

2.3.4. Numerical Regression Analysis

(1) Neural Network Model. The BP neural network uses vector multiplication and nonlinear activation function to multiply the input data with the connection weights, map the results into the nonlinear space, and adjust the connection weights by error back propagation to obtain a good fit. Its topological structure consists of an input layer, an implicit layer, and an output layer, as shown in Figure 2. The BP neural network automatically extracts the “reasonable rules” between the input and output data through learning and stores the adaptive learning content in the network weights. For problems with complex internal mechanisms, it can also approximate any nonlinear continuous function through self-learning. In this article, the model was trained and tested with the spectral reflectance at different bands as the input and the grain moisture content or thousand-grain weight as the output.

(2) Support Vector Machine. The support vector machine (SVM) is a machine learning method based on the statistical learning theory Vapnik–Chervonenkis dimension and the principle of structural risk minimization [24]. The linear or nonlinear kernel function can be selected to map the data into the high-dimensional space and extract the high-dimensional features of data by finding the hyperplane with the maximum interval.

(3) Decision Tree. The decision tree algorithm is a tree-structured prediction model that characterizes the mapping relationship between attributes and values of the predicted object. The algorithm structure contains nodes and directed edges, the nodes of which include a root node, child nodes, and leaf nodes (Figure 3). Child nodes are created based on the recursive principle from the top down in the decision tree algorithm until the dataset is indivisible and a regression tree is generated.

3. Result Analysis

3.1. Changes in Rice Moisture Content and Thousand-Grain Weight

Field sampling, moisture content testing, drying, and thousand-grain weight trials were carried out repeatedly according to the rice moisture content and thousand-grain weight test methods in 2.3.2. The moisture content and thousand-grain weight data of South Japonica 46 and South Japonica 5055 from Nov. 5 to 29 were obtained, as shown in Table 2.

As shown in Table 2, the moisture content of the test sample South Japonica 46 decreased linearly during the 20-day trial period from 30.93% on Nov. 5 to 23.57% on Nov. 29, while the thousand-grain weight of South Japonica 46 showed a trend of increasing first (from 23.57 g on Nov. 5 to 26.81 g on Nov. 20) and then decreasing to 25.21 g on Nov. 29. The moisture content of South Japonica 5055 showed a fluctuating downward trend, first increasing from 28.97% on Nov. 5 to 30.73% on Nov. 8, then decreasing to 25.93% on Nov. 29, during which time the moisture content fluctuated significantly due to the influence of weather factors. The thousand-grain weight of South Japonica 5055 also showed a trend of increasing first (from 24.60 g on Nov. 5 to 25.94 g on Nov. 24) and then decreasing to 24.81 g on Nov. 29.

3.2. Changes in Spectral Reflectance

By using the multispectral remote sensing test method to the rice canopy in 2.3.1 and applying the multispectral remote sensing image processing and spectral reflectance extraction method in 2.3.3, the spectral reflectance of South Japonica 46 and South Japonica 5055 at four bands from Nov. 5 to 29 were obtained, as shown in Table 3.

As shown in Table 3, the change in spectral reflectance at 550 nm with the passing date was not obvious, where the spectral reflectance of South Japonica 46 at 550 nm fluctuated around 25% during the trial period, while that of South Japonica 5055 fluctuated around 23%. The change in spectral reflectance at 660 nm with the passing date was not obvious either, where the spectral reflectance of South Japonica 46 at 660 nm fluctuated around 26% during the trial period, while that of South Japonica 5055 rice fluctuated around 25%. The spectral reflectance at 735 nm tended to increase with each date, from about 35% at the beginning of the trial to about 75% during the later stage for both rice varieties. The spectral reflectance at 790 nm fluctuated drastically. It could be preliminarily determined that the spectral band at 735 nm was more sensitive to changes in the physiological traits of rice maturation.

3.3. Regression Prediction Results

With spectral reflectance as the input and rice thousand-grain weight or moisture content as the output, 300 samples were randomly selected as the training set and the rest as the test set to perform regression prediction at single-band, two-band, three-band, and full-band combinations, respectively.

3.3.1. Single-Band Regression Results

Three numerical regression methods (BP neural network, SVM, and decision tree) proposed in 2.3.4 were used to perform regression on the spectral reflectance at four single bands and rice physiological indexes of South Japonica 46 and South Japonica 5055, respectively. The determination coefficient (R²) and root mean square error (RMSE) of each model were calculated. The results were obtained are shown in Table 4.

In terms of physiological indexes of rice, the variable determination coefficient (R²) of the regression model of rice grain moisture content and spectral reflectance was relatively low and generally lower than that of rice grain thousand-grain weight and spectral reflectance. In terms of regression methods, the regression model obtained from the decision tree had a good correlation between input and output variables, but the RMSE of the prediction results was much larger than that from the BP neural network and SVM. The variable determination coefficient (R²) of the regression model obtained from SVM was generally lower than those from the BP neural network and decision tree, and the RMSE of the prediction results was smaller than that from the decision tree but larger than that from the BP neural network. In terms of spectral bands, the regression effect at 735 nm was relatively good in the BP neural network model; the determination coefficients between the spectral reflectance at 735 nm and the thousand-grain weight of both rice varieties were the highest, and the RMSE of the prediction results was also the smallest. The regression effect at 790 nm was relatively good in the SVM model, while the R² was generally lower than that of the BP neural network; the RMSE was generally higher than that of the BP neural network. The regression effect at 790 nm was relatively good in the decision tree model, with higher R² than that in BP neural network and SVM, but the RMSE was also the highest.

3.3.2. Combined-Band Regression Results

The spectral data of two bands, three bands, and full band were combined, respectively, to obtain 11 band combinations, namely, 550 + 660, 550 + 735, 550 + 790, 660 + 735, 735 + 790, 550 + 660 + 735, 550 + 735 + 790, 660 + 735 + 790, 550 + 660 + 790, 550 + 660 + 790, and 550 + 660 + 735 + 790. According to the method in 2.3.4, regressions were performed on the spectral reflectance at combined bands and physiological indexes of South Japonica 46 and South Japonica 5055. The determination coefficient (R²) and RMSE of each model were calculated. The results that were obtained are shown in Table 5.

According to the combined-band regression analysis, the R² and RMSE of the regression model of rice moisture content and spectral reflectance were generally lower than those of rice thousand-grain weight and spectral reflectance. In terms of regression methods, the regression model that used the decision tree had the highest determination but also had the largest RMSE of prediction results; the regression model using the BP neural network had the highest correlation, but the RMSE was larger than that using SVM. In terms of band combinations, the regression effect at full band was relatively good in the BP neural network and SVP models, with relatively high determination coefficients of spectral reflectance with rice moisture content and thousand-grain weight and the smallest RMSE of prediction results in both models. The overall regression effect at the three-band combinations was relatively good, with the best regression effect at 550 + 660 + 735 nm and 660 + 735 + 790 nm in the SVM model. The regression effect at two-band combinations was relatively good in the decision tree model; the regression model had the highest variable determination coefficient (R²) but also had the largest RMSE of prediction results.

4. Discussion

Although spectral information is extensively used in crop detection, existing studies mostly use spectral features of crops to monitor various periods [25, 26], which has certain limitations. In this article, spectral features of the rice canopy were continuously collected by multispectral remote sensing cameras during the harvest period. Combined with the thousand-grain weight of rice, a regression model of spectral features and thousand-grain weight was established to determine the appropriate harvest period. By combining the multispectral data collected by UAV with different physiological indexes of rice, a rice maturity prediction model using different algorithms was established. Three machine learning algorithms (BP neural network, SVM, and the decision tree) were used for conducting regression analysis of spectral characteristics and physiological indexes of rice, respectively. The determination coefficient (R²) and RMSE under different prediction models were calculated. The correlation analysis of different band data indicated that the multiband regression model was significantly superior to the single-band one. In particular, the correlation between spectral data (660 + 735 + 790) and thousand-grain weight of 5055 (R² = 0.99, RMSE = 0.09) was superior to the other regression models. Thousand-grain weight is an essential basis for measuring rice yields, and the change in thousand-grain weight can reflect the maturity of rice during the growth process. The regression model established in this article has a high correlation between spectral data and thousand-grain weight. Hence, it can be used to reflect the maturity of rice.

5. Conclusions

(1)In this article, a monitoring method for the appropriate harvest period of rice based on multispectral remote sensing was established. Remote sensing data of the rice canopy were acquired by UAV equipped with multispectral (550 nm, 660 nm, 735 nm, and 790 nm) cameras, and physiological characteristic data of rice moisture content and thousand-grain weight, tassel ratio were determined simultaneously. Single-band and combined-band spectral reflectance were introduced as model input variables to compare the BP neural network, SVM, and decision tree; a quantitative inversion model of rice physiological characteristics was established. The accuracy of the inversion model with different input variables and model methods was evaluated, and the optimal inversion model for rice moisture content and thousand-grain weight was selected, which could be used to determine whether a crop is in the appropriate harvest period. The remote monitoring method for the rice harvesting period studied in this article can replace manual inspection in large-scale operation entities or “unmanned farms,” effectively reducing labor costs and improving monitoring efficiency.(2)The continuous monitoring of two rice varieties (South Japonica 46 and South Japonica 5055) indicated that the moisture content of South Japonica 46 decreased linearly during the 20-day trial period, while its thousand-grain weight showed a trend of increasing first and then decreasing. The moisture content of South Japonica 5055 showed a fluctuating downward trend, and its thousand-grain weight also showed a trend of increasing first and then decreasing. The spectral reflectance at the 735 nm band increased with each passing day, and that of both rice varieties increased from about 35% at the beginning of the trial to about 75% in the later stage. The spectral reflectance at the other three bands did not change significantly with the harvest date.(3)The inverse model of spectral reflectance with rice thousand-grain weight and moisture content was established by using the BP neural network, SVM, and decision tree. In single-band inversion models, the regression effect at 735 nm was relatively good in the BP neural network model, having the highest determination coefficients of spectral reflectance and thousand-grain weight of both rich varieties and the smallest RMSE of prediction results. In combined-band inversion models, the regression effects at 550 + 660 + 735 nm and 660 + 735 + 790 nm were the best in the SVM and thousand-grain weight inversion models.

Data Availability

The remote sensing data and codes used in the experiments to support the findings of this study are available from the corresponding author upon request.

Conflicts of Interest

The authors declare that they have no conflicts of interest.

Authors’ Contributions

Chen Cong and Hu Jianping contributed to the conceptualization; Chen Cong and Cao Guangqiao were responsible for the methodology; Liu Dong was responsible for the hardware; Li Yibai and Zhang Jinlong were responsible for the software; Cao Guangqiao and Chen Cong validated the data; Cao Guangqiao conducted the formal analysis; Li Liang conducted the investigation; Chen Cong was responsible for the resources, wrote the original draft preparation, and reviewed and edited the work; Ma Bin contributed to the data curation; Cao Guangqiao supervised the work; Hu Jianping was the project administrator; Zhang Jinlong was responsible for the funding acquisition. All authors have read and agreed to the published version of the manuscript.

Acknowledgments

This work was supported by a grant of the special funding for Basic Scientific Research Business Expenses of Central Public Welfare Scientific Research Institutes (S202010 and S202109-02) and Science and Technology Innovation Project of the Chinese Academy of Agricultural Sciences (Academy of Agricultural Sciences Office (2014) No. 216).

References

J. Wang, “Research of the rice harvesting season timeliness loss in Sanjiang plain zone,” Transactions of the Chinese Society for Agricultural Machinery, vol. 35, no. 2, pp. 175–177, 2004.
View at: Google Scholar
A. Kessler, S. V. Archontoulis, and M. A. Licht, “Soybean yield and crop stage response to planting date and cultivar maturity in Iowa, USA,” Agronomy Journal, vol. 112, no. 1, pp. 382–394, 2020.
View at: Google Scholar
W. Chen, Q. Zhou, and J. Huang, “Estimating pigment contents in leaves and panicles of rice after milky ripening by hyperspectral vegetation indices,” Chinese Journal of Rice Science, vol. 20, no. 4, pp. 434–439, 2006.
View at: Google Scholar
L. Wannasek, M. Ortner, H.-P. Kaul, B. Amon, and T. Amon, “Double-cropping systems based on rye, maize and sorghum: impact of variety and harvesting time on biomass and biogas yield,” European Journal of Agronomy, vol. 110, no. 10, Article ID 125934, 125934 pages, 2019.
View at: Publisher Site | Google Scholar
C. M. Boyce, K. C. VerCauteren, and J. C. Beasley, “Timing and extent of crop damage by wild pigs (Sus scrofa Linnaeus) to corn and peanut fields,” Crop Protection, vol. 133, no. 7, Article ID 105131, 2020.
View at: Publisher Site | Google Scholar
R. Lantin, Rice: post-harvest Operations, International rice research institute, Philippines, Asia, 1999.
W Y. MiaoDeyu and HeHaisheng, “Effect of harvesting time and pattern on milling quality and yield of rice,” North Rice, vol. 37, no. 4, pp. 25–27, 2007.
View at: Google Scholar
H. Gao and G. A. Moore, “Wheatquality losses due to untimely harvesting in Australia,” Journal of Beijing Agricultural Engineering University, vol. 11, no. 01, pp. 6–14, 1991.
View at: Google Scholar
Z. Li, “Function of operation’s timeliness and determinationofproper distribution of operations’duration,” Transactions of the Chinese Society for Agricultural Machinery, vol. 13, no. 4, pp. 69–77, 1986.
View at: Google Scholar
WangJinwu and G. Yang, “Research of the rice planting season timeliness loss in Sanjiang Plain zone,” Journal of Northeast Agricultural University, vol. 35, no. 4, pp. 472–475, 2008.
View at: Google Scholar
G. Wang, Z. Yi, C. Cong, and G. Cao, “Effect of harvesting date on loss component characteristics of rice mechanical harvested in rice and wheat rotation area,” Transactions of the Chinese Society of Agricultural Engineering, vol. 32, no. 2, pp. 36–42, 2016.
View at: Google Scholar
T. Le Ngoc Huyen, C. Rémond, R. M. Dheilly, and B. Chabbert, “Effect of harvesting date on the composition and saccharification of Miscanthus x giganteus,” Bioresource Technology, vol. 101, no. 21, pp. 8224–8231, 2010.
View at: Publisher Site | Google Scholar
B. Godin, S. Lamaudière, R. Agneessens et al., “Chemical characteristics and biofuels potentials of various plant biomasses: influence of the harvesting date,” Journal of the Science of Food and Agriculture, vol. 93, no. 13, pp. 3216–3224, 2013.
View at: Publisher Site | Google Scholar
N. Heidari, M. Pouryousef, A. Tavakkoli, and J. Saba, “Effect of drought stress and harvesting date on yield and essential oil production of anise,” Iranian Journal of Medicinal and Aromatic Plants, vol. 28, no. 1, pp. 121–130, 2012.
View at: Publisher Site | Google Scholar
QiaoJinyou, J. Sun, and F. Qu, “Study on regularities of timeliness loss changing with soybean harvesting date,” Journal of Northeast Agricultural University, vol. 48, no. 6, pp. 76–81, 2017.
View at: Google Scholar
Z. Du, Y. Yang, and Y. Xia, “Effects of harvest time on quality traits of hybrid japonica rice and inbred japonica rice in northern China,” Crops, vol. 34, no. 1, pp. 147–151, 2018.
View at: Google Scholar
P. Sirisomboon, M. Tanaka, T. Kojima, and P. Williams, “Nondestructive estimation of maturity and textural properties on tomato ‘Momotaro’ by near infrared spectroscopy,” Journal of Food Engineering, vol. 112, no. 3, pp. 218–226, 2012.
View at: Publisher Site | Google Scholar
X. Sun, Y. Liu, Y. Li, M. Wu, and D. Zhu, “Simultaneous measurement of brown core and soluble solids content in pear by on-line visible and near infrared spectroscopy,” Postharvest Biology and Technology, vol. 116, pp. 80–87, 2016.
View at: Publisher Site | Google Scholar
W. Guo, J. Gu, D. Liu, and L. Shang, “Peach variety identification using near-infrared diffuse reflectance spectroscopy,” Computers and Electronics in Agriculture, vol. 123, no. 4, pp. 297–303, 2016.
View at: Publisher Site | Google Scholar
XueJianxin, S. Zhang, and J. Zhang, “Ripeness classification of Shajin apricot using hyperspectral imaging technique,” Transactions of the Chinese Society of Agricultural Engineering, vol. 31, no. 11, pp. 300–307, 2015.
View at: Google Scholar
L. Li, B. Wang, X. Zhang, and S J. Zhang, “Discrimination on maturity of plums based on hyperspectral imaging information,” Modern Food Science & Technology, vol. 33, no. 12, pp. 228–232, 2017.
View at: Publisher Site | Google Scholar
P. Wan, A. Toudeshki, H. Tan, and R. Ehsani, “A methodology for fresh tomato maturity detection using computer vision,” Computers and Electronics in Agriculture, vol. 146, no. 3, pp. 43–50, 2018.
View at: Publisher Site | Google Scholar
L. F. Santos Pereira, S. Barbon, N. A. Valous, and D. F. Barbin, “Predicting the ripening of papaya fruit with digital imaging and random forests,” Computers and Electronics in Agriculture, vol. 145, no. 2, pp. 76–82, 2018.
View at: Publisher Site | Google Scholar
S. Ding, B. Qi, and H. Tan, “An overview on theory and algorithm of support vector machines,” Journal of University of Electronic Science and Technology of China, vol. 40, no. 1, pp. 2–10, 2011.
View at: Google Scholar
J. Long, C. Zhao, L. Sen, W. Guo, C. Wen, and Y. Zhang, “Segmentation method of the tomato fruits with different maturities under greenhouse environment based on improved Mask R-CNN,” Transactions of the Chinese Society of Agricultural Engineering, vol. 37, no. 18, pp. 100–108, 2021.
View at: Publisher Site | Google Scholar
R. Yang, B. Jun, and SuJiaen, “Identification of tobacco leaf maturity based on the fusion of near infrared spectroscopy and image recognition,” Journal of Hunan Agricultural University, vol. 47, no. 04, pp. 406–411+418, 2021.
View at: Google Scholar

Copyright

Copyright © 2022 Chen Cong et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

PDF Download Citation

Download other formats

Order printed copies