基于机器学习和环境因子的温室网纹甜瓜纹理特征的预测
|
摘要
为了对网纹甜瓜纹理特征进行定量预测,本文结合多个环境因子(基质含水量、温度、湿度、光合有效辐射),使用支持向量机(SVM)与随机森林(RF)多元回归分析方法,对基于灰度共生矩阵提取的4个果实表面纹理特征——对比度(contrast)、熵(entropy)、相关性(correlation)、角二阶矩(ASM)进行了预测。结果表明,纹理特征受环境影响敏感度依次为基质水分、光合有效辐射、空气湿度或有效积温。对比RF模型和SVR模型,发现在不同环境因子输入条件下,RF模型均优于SVM模型,其中输入全部环境因子后RF模型的预测精度最高,对比度、熵、角二阶矩的模拟精度均达到了0.90,分别为R~2=0.945(RMSE=0.243)、R~2=0.940(RMSE=0.235)、R~2=0.934(RMSE=0.248)。上述结果表明,RF模型对于网纹甜瓜纹理特征具有较好的预测结果,本研究结果可为温室网纹甜瓜栽培过程中的质量监控和栽培管理提供全新的思路。
In order to quantitatively predict the texture characteristics of netted melon,the predictive models of four fruit surface texture features were constructed,based on gray-level co-occurrence matrix—contrast,entropy,correlation,and angular second moment(ASM),combined with multiple environmental factors(matrix moisture,temperature,humidity,photosynthetically active radiation) and support vector machine(SVR) and random forest(RF) multiple regression analysis method.The results show that the sensitivity of the texture feature values to the environment is matrix moisture,photosynthetically active radiation,humidity and temperature.Comparing the RF model with the SVR model,it is found that the RF model is superior to the SVR model under different environmental factor input conditions.The RF model has the highest prediction accuracy after inputting all environmental factors,and the simulation accuracies of contrast,entropy and ASM have reached 0.9,R~2=0.945(RMSE=0.243),R~2=0.940(RMSE=0.235),and R~2=0.934(RMSE=0.248),respectively.The above results show that the RF model has a good performance for predicting the texture characteristics of netted melon.This research can provide a new idea for the quality monitoring and cultivation management of greenhouse muskmelon.
In order to quantitatively predict the texture characteristics of netted melon,the predictive models of four fruit surface texture features were constructed,based on gray-level co-occurrence matrix—contrast,entropy,correlation,and angular second moment(ASM),combined with multiple environmental factors(matrix moisture,temperature,humidity,photosynthetically active radiation) and support vector machine(SVR) and random forest(RF) multiple regression analysis method.The results show that the sensitivity of the texture feature values to the environment is matrix moisture,photosynthetically active radiation,humidity and temperature.Comparing the RF model with the SVR model,it is found that the RF model is superior to the SVR model under different environmental factor input conditions.The RF model has the highest prediction accuracy after inputting all environmental factors,and the simulation accuracies of contrast,entropy and ASM have reached 0.9,R~2=0.945(RMSE=0.243),R~2=0.940(RMSE=0.235),and R~2=0.934(RMSE=0.248),respectively.The above results show that the RF model has a good performance for predicting the texture characteristics of netted melon.This research can provide a new idea for the quality monitoring and cultivation management of greenhouse muskmelon.
引文
[1] 吴振磊,刘秀杰.甜瓜生产现状与发展策略研究[J].新农村(黑龙江),2017(23):173-173.
[2] 何三鹏.基于计算机视觉技术的温室网纹甜瓜果实形态模拟研究[D].上海:上海交通大学,2011.
[3] Kacira M,Ling P P.Design and development of an automated and non-contact sensing system for continuous monitoring of plant health and growth[J].Transactions of the Asae American Society of Agricultural Engineers,2001,44(4):989-1001.
[4] 黄丹枫,李磊,柯顺魁,等.网纹甜瓜果实表型提取与量化方法[P].中国专利:ZL2012105638560,2013-05-04.
[5] 王书志,张建华,冯全.基于纹理和颜色特征的甜瓜缺陷识别[J].农业机械学报,2011,42(3):175-179.
[6] 张建华.基于计算机视觉的硬皮甜瓜外部品质自动化分级研究[D].兰州:甘肃农业大学,2009.
[7] 管学玉.网纹甜瓜品质形成特点的研究[D].杭州:浙江大学,2006.
[8] 张容.优质网纹甜瓜春夏栽培指南[J].中国蔬菜,2012,1(12):49-50.
[9] 王怀松,周晓星.网纹甜瓜果实发育规律的观察[J].中国瓜菜,2002(2):20-21.
[10] 毛莎莎,曾明,何绍兰,等.近红外光谱技术在水果成熟期预测中的应用(综述)[J].亚热带植物科学,2010,39(1):82-87.
[11] 周志华.机器学习:=Machine learning[M].北京:清华大学出版社,2016.
[12] 王升,付智勇,陈洪松,等.基于随机森林算法的参考作物蒸发蒸腾量模拟计算[J].农业机械学报,2017,48(3):302-309.
[13] 韩兆迎,朱西存,房贤一,等.基于SVM与RF的苹果树冠LAI高光谱估测[J].光谱学与光谱分析,2016(3):800-805.
[14] Li Zhen-wang,Xin Xiao-ping,Tang Huan,et al.Estimating grassland LAI using the Random Forests approach and Landsat imagery in the meadow steppe of Hulunber,China[J].Journal of Integrative Agriculture,2017,16(2):286-297.
[15] Lu Sheng,Cai Zhong-jian,Zhang Xiao-bin.Forecasting agriculture water consumption based on PSO and SVM[C].2009:147-150.
[16] 薛大伟,方茂庭,钱前.有效积温在水稻生产中的应用[J].中国稻米,2004(4):47-48.
[17] 刘敏,郎荣玲,曹永斌.随机森林中树的数量[J].计算机工程与应用,2015,51(5):126-131.
[18] 付阳,李昆仑.支持向量机模型参数选择方法综述[J].电脑知识与技术,2010,6(28):8081-8082.
[19] 徐寿军,杨恒山,郑根昌,等.大麦植株干物质分配模型[J].麦类作物学报,2010,30(4):735-740.
[20] 施泽平,郭世荣,康云艳,等.基于生长度日的温室甜瓜发育模拟模型的研究[J].南京农业大学学报,2005,28(2):129-132.
[21] 高汉峰,房俊龙,陈月华.作物生长信息智能监测方法研究进展[J].农机化研究,2010,32(4):1-5.
[22] 武军,谢英丽,安丙俭.我国精准农业的研究现状与发展对策[J].山东农业科学,2013,45(9):118-121.
[23] Thibault Nordey,Jacques Joas,Fabrice Davrieux,et.al.Robust NIRS models for non-destructive prediction of mango internal quality[J].Scientia Horticulturae,2017,216:51-57.
[24] Gabriel A Leiva-Valenzuela,Renfu Lu,José Miguel Aguilera.Prediction of firmness and soluble solids content of blueberries using hyperspectral reflectance imaging[J].Journal of Food Engineering,2013,115(1):91-98.
[25] 崔冲.基质含水量与温室网纹甜瓜果实品质形成的模拟[D].上海:上海交通大学,2012.
[26] 魏亦榕.基于有机氮营养的温室网纹甜瓜果实品质形成模型研究[D].上海:上海交通大学,2013.
[2] 何三鹏.基于计算机视觉技术的温室网纹甜瓜果实形态模拟研究[D].上海:上海交通大学,2011.
[3] Kacira M,Ling P P.Design and development of an automated and non-contact sensing system for continuous monitoring of plant health and growth[J].Transactions of the Asae American Society of Agricultural Engineers,2001,44(4):989-1001.
[4] 黄丹枫,李磊,柯顺魁,等.网纹甜瓜果实表型提取与量化方法[P].中国专利:ZL2012105638560,2013-05-04.
[5] 王书志,张建华,冯全.基于纹理和颜色特征的甜瓜缺陷识别[J].农业机械学报,2011,42(3):175-179.
[6] 张建华.基于计算机视觉的硬皮甜瓜外部品质自动化分级研究[D].兰州:甘肃农业大学,2009.
[7] 管学玉.网纹甜瓜品质形成特点的研究[D].杭州:浙江大学,2006.
[8] 张容.优质网纹甜瓜春夏栽培指南[J].中国蔬菜,2012,1(12):49-50.
[9] 王怀松,周晓星.网纹甜瓜果实发育规律的观察[J].中国瓜菜,2002(2):20-21.
[10] 毛莎莎,曾明,何绍兰,等.近红外光谱技术在水果成熟期预测中的应用(综述)[J].亚热带植物科学,2010,39(1):82-87.
[11] 周志华.机器学习:=Machine learning[M].北京:清华大学出版社,2016.
[12] 王升,付智勇,陈洪松,等.基于随机森林算法的参考作物蒸发蒸腾量模拟计算[J].农业机械学报,2017,48(3):302-309.
[13] 韩兆迎,朱西存,房贤一,等.基于SVM与RF的苹果树冠LAI高光谱估测[J].光谱学与光谱分析,2016(3):800-805.
[14] Li Zhen-wang,Xin Xiao-ping,Tang Huan,et al.Estimating grassland LAI using the Random Forests approach and Landsat imagery in the meadow steppe of Hulunber,China[J].Journal of Integrative Agriculture,2017,16(2):286-297.
[15] Lu Sheng,Cai Zhong-jian,Zhang Xiao-bin.Forecasting agriculture water consumption based on PSO and SVM[C].2009:147-150.
[16] 薛大伟,方茂庭,钱前.有效积温在水稻生产中的应用[J].中国稻米,2004(4):47-48.
[17] 刘敏,郎荣玲,曹永斌.随机森林中树的数量[J].计算机工程与应用,2015,51(5):126-131.
[18] 付阳,李昆仑.支持向量机模型参数选择方法综述[J].电脑知识与技术,2010,6(28):8081-8082.
[19] 徐寿军,杨恒山,郑根昌,等.大麦植株干物质分配模型[J].麦类作物学报,2010,30(4):735-740.
[20] 施泽平,郭世荣,康云艳,等.基于生长度日的温室甜瓜发育模拟模型的研究[J].南京农业大学学报,2005,28(2):129-132.
[21] 高汉峰,房俊龙,陈月华.作物生长信息智能监测方法研究进展[J].农机化研究,2010,32(4):1-5.
[22] 武军,谢英丽,安丙俭.我国精准农业的研究现状与发展对策[J].山东农业科学,2013,45(9):118-121.
[23] Thibault Nordey,Jacques Joas,Fabrice Davrieux,et.al.Robust NIRS models for non-destructive prediction of mango internal quality[J].Scientia Horticulturae,2017,216:51-57.
[24] Gabriel A Leiva-Valenzuela,Renfu Lu,José Miguel Aguilera.Prediction of firmness and soluble solids content of blueberries using hyperspectral reflectance imaging[J].Journal of Food Engineering,2013,115(1):91-98.
[25] 崔冲.基质含水量与温室网纹甜瓜果实品质形成的模拟[D].上海:上海交通大学,2012.
[26] 魏亦榕.基于有机氮营养的温室网纹甜瓜果实品质形成模型研究[D].上海:上海交通大学,2013.