基于Faster R-CNN的田间西兰花幼苗图像检测方法
|
摘要
为解决自然环境下作物识别率不高、鲁棒性不强等问题,以西兰花幼苗为研究对象,提出了一种基于Faster R-CNN模型的作物检测方法。根据田间环境特点,采集不同光照强度、不同地面含水率和不同杂草密度下的西兰花幼苗图像,以确保样本多样性,并通过数据增强手段扩大样本量,制作PASCAL VOC格式数据集。针对此数据集训练Faster R-CNN模型,通过设计ResNet101、ResNet50与VGG16网络的对比试验,确定ResNet101网络为最优特征提取网络,其平均精度为90. 89%,平均检测时间249 ms。在此基础上优化网络超参数,确定Dropout值为0. 6时,模型识别效果最佳,其平均精度达到91. 73%。结果表明,本文方法能够对自然环境下的西兰花幼苗进行有效检测,可为农业智能除草作业中的作物识别提供借鉴。
Traditional methods of image processing for crop detection under agricultural natural environment are easily affected by small samples and subjective judgment, so they have many disadvantages such as low recognition rate and low robustness. Deep learning can self-study according to data set,and has a strong ability to express feature. Therefore,a new broccoli seedlings detection approach based on Faster R-CNN model was proposed. Data acquisition was the first step to build deep learning model,and the diversity of data can improve the generalization ability of the model. According to the characteristics of field environment,broccoli seedlings images with different light intensities,different ground moisture contents and different weed densities were collected. The sample size was expanded by images rotation and noise enhancement,and data set was transformed as PASCAL VOC format. And then the Faster R-CNN model was trained by using data set. Contrast experiment was designed on ResNet101,ResNet50 and VGG16 networks. The results showed that ResNet101 network with the deepest network layer and smaller parameter space was the best feature extraction network. The average detection accuracy was 90. 89%,and the average time-consuming was 249 ms. Based on that,the network super-parameters were optimized and the average accuracy of model detection reached 91. 73%,when Dropout value was 0. 6. The results showed that this approach can effectively detect broccoli seedlings in agricultural natural environment,and provided a hopeful solution for crop detection in the field of agriculture.
Traditional methods of image processing for crop detection under agricultural natural environment are easily affected by small samples and subjective judgment, so they have many disadvantages such as low recognition rate and low robustness. Deep learning can self-study according to data set,and has a strong ability to express feature. Therefore,a new broccoli seedlings detection approach based on Faster R-CNN model was proposed. Data acquisition was the first step to build deep learning model,and the diversity of data can improve the generalization ability of the model. According to the characteristics of field environment,broccoli seedlings images with different light intensities,different ground moisture contents and different weed densities were collected. The sample size was expanded by images rotation and noise enhancement,and data set was transformed as PASCAL VOC format. And then the Faster R-CNN model was trained by using data set. Contrast experiment was designed on ResNet101,ResNet50 and VGG16 networks. The results showed that ResNet101 network with the deepest network layer and smaller parameter space was the best feature extraction network. The average detection accuracy was 90. 89%,and the average time-consuming was 249 ms. Based on that,the network super-parameters were optimized and the average accuracy of model detection reached 91. 73%,when Dropout value was 0. 6. The results showed that this approach can effectively detect broccoli seedlings in agricultural natural environment,and provided a hopeful solution for crop detection in the field of agriculture.
引文
[1]刘伟明,汪恩国,蔡建军,等.西兰花早中熟品种不同苗龄高密度栽培技术研究[J].农学学报,2018,8(11):49-53.LIU Weiming,WANG Enguo,CAI Jianjun,et al. High density cultivation techniques of early-middle maturing broccoli at different seedling ages[J]. Journal of Agriculture,2018,8(11):49-53.(in Chinese)
[2]马爱民,刘宗陈,陆亚琴,等.秋播西兰花大面积生产667 m2产量超1 000 kg高产栽培技术[J].长江蔬菜,2018(23):20-21.
[3]赵国庆.农业机械自动化的现状与推进模式[J].湖南农机,2014,41(12):6-7.
[4] PIRON A. Selection of the most efficient wavelength bands for discriminating weeds from crop[J]. Computers and Electronics in Agriculture,2008,62(2):141-148.
[5]李先锋,朱伟兴,纪滨,等.基于特征优化和LS-SVM的棉田杂草识别[J].农业机械学报,2010,41(11):168-172.LI Xianfeng,ZHU Weixing,JI Bin,et al. Weed identification based on features optimization and LS-SVM in the cotton field[J].Transactions of the Chinese Society for Agricultural Machinery,2010,41(11):168-172.(in Chinese)
[6]胡炼,罗锡文,曾山,等.基于机器视觉的株间机械除草装置的作物识别与定位方法[J].农业工程学报,2013,29(10):12-18.HU Lian,LUO Xiwen,ZENG Shan,et al. Plant recognition and localization for intra-row mechanical weeding device based on machine vision[J]. Transactions of the CSAE,2013,29(10):12-18.(in Chinese)
[7] LI N,GRIFT T E,YUAN T,et al. Image processing for crop/weed discrimination in fields with high weed pressure[C]∥Michigan:American Society of Agricultural and Biological Engineers,2016.
[8] CHRISTINE M,ONYANGO J A. Segmentation of row crop plants from weeds using colour and morphology[J]. Computers and Electronics in Agriculture,2003,39(3):141-155.
[9] SI C. Weed-plant discrimination by machine vision and artificial neural network[J]. Biosystems Engineering,2002,83(3):275-280.
[10]刘洪臣,陈忠建,冯勇.结合颜色和形态特征的杂草实时识别方法[J].光电工程,2006,33(7):96-100.LIU Hongchen,CHEN Zhongjian,FENG Yong. Real-time weed recognition method based on color and morphological features[J]. Opto-Electronic Engineering,2006,33(7):96-100.(in Chinese)
[11]李颖,张立福,严薇,等.地面成像光谱数据的田间杂草识别[J].遥感学报,2013,17(4):855-871.LI Ying,ZHANG Lifu,YAN Wei,et al. Weed identification using imaging spectrometer data[J]. Journal of Remote Sensing,2013,17(4):855-871.(in Chinese)
[12]乔永亮,何东健,赵川源,等.基于多光谱图像和SVM的玉米田间杂草识别[J].农机化研究,2013,35(8):30-34.
[13] CHEN Y,JIN X,TANG L,et al. Intra-row weed recognition using plant spacing information in stereo images[C]∥Michigan:American Society of Agricultural and Biological Engineers,2013.
[14]陈树人,沈宝国,毛罕平,等.基于颜色特征的棉田中铁苋菜识别技术[J].农业机械学报,2009,40(5):149-152.CHEN Shuren,SHEN Baoguo,MAO Hanping,et al. Herb detection from cotton field based on color feature[J]. Transactions of the Chinese Society for Agricultural Machinery,2009,40(5):149-152.(in Chinese)
[15] HERRMANN I,SHAPIRA U,KINAST S,et al. Ground-level hyperspectral imagery for detecting weeds in wheat fields[J].Precision Agriculture,2013,14(6):637-659.
[16] GARCIA I D,MONTALVO M,GUERRERO J M,et al. Automatic detection of curved and straight crop rows from images in maize fields[J]. Biosystems Engineering,2017,156:61-79.
[17]张志斌,赵帅领,罗锡文,等.基于SURF算法的绿色作物特征提取与图像匹配方法[J].农业工程学报,2015,31(14):172-178.ZHANG Zhibin,ZHAO Shuailing,LUO Xiwen,et al. Matching method of green crops based on SURF feature extraction[J].Transactions of the CSAE,2015,31(14):172-178.(in Chinese)
[18] NIEUWENHUIZEN A T,HOFSTEE J W,HENTEN E J. Adaptive detection of volunteer potato plants in sugar beet fields[J].Precision Agriculture,2010,11(5):433-447.
[19]傅隆生,冯亚利,ELKAMIL T,等.基于卷积神经网络的田间多簇猕猴桃图像识别方法[J].农业工程学报,2018,34(2):205-211.FU Longsheng,FENG Yali,ELKAMIL T,et al. Image recognition method of multi-cluster kiwifruit in field based on convolutional neural networks[J]. Transactions of the CSAE,2018,34(2):205-211.(in Chinese)
[20]彭红星,黄博,邵园园,等.自然环境下多类水果采摘目标识别的通用改进SSD模型[J].农业工程学报,2018,34(16):155-162.PENG Hongxing,HUANG Bo,SHAO Yuanyuan,et al. General improved SSD model for picking object recognition of multiple fruits in natural environment[J]. Transactions of the CSAE,2018,34(16):155-162.(in Chinese)
[21]周云成,许童羽,郑伟,等.基于深度卷积神经网络的番茄主要器官分类识别方法[J].农业工程学报,2017,33(15):219-226.ZHOU Yuncheng,XU Tongyu,ZHENG Wei,et al. Classification and recognition approaches of tomato main organs based on DCNN[J]. Transactions of the CSAE,2017,33(15):219-226.(in Chinese)
[22] REN S,HE K,GIRSHISK R,et al. Faster R-CNN:towards real-time object detection with region proposal networks[C]∥Montréal:International Conference on Neural Information Processing Systems,2015.
[23] GIRSHISK R,DONAHUE J,DARRELL T,et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]∥Ohio:Conference on Computer Vision and Pattern Recognition,2014.
[24] GIRSHISK R. Fast R-CNN[C]∥Santiago:International Conference on Computer Vision,2015.
[25] HE K,ZHANG X,REN S,et al. Identity mappings in deep residual networks[C]∥European Conference on Computer Vision,2016:630-645.
[26] SRIVASTAVA N,HINTON G,KRIZHEVSKY A,et al. Dropout:a simple way to prevent neural networks from overfitting[J].Journal of Machine Learning Research,2014,15(1):1929-1958.
[2]马爱民,刘宗陈,陆亚琴,等.秋播西兰花大面积生产667 m2产量超1 000 kg高产栽培技术[J].长江蔬菜,2018(23):20-21.
[3]赵国庆.农业机械自动化的现状与推进模式[J].湖南农机,2014,41(12):6-7.
[4] PIRON A. Selection of the most efficient wavelength bands for discriminating weeds from crop[J]. Computers and Electronics in Agriculture,2008,62(2):141-148.
[5]李先锋,朱伟兴,纪滨,等.基于特征优化和LS-SVM的棉田杂草识别[J].农业机械学报,2010,41(11):168-172.LI Xianfeng,ZHU Weixing,JI Bin,et al. Weed identification based on features optimization and LS-SVM in the cotton field[J].Transactions of the Chinese Society for Agricultural Machinery,2010,41(11):168-172.(in Chinese)
[6]胡炼,罗锡文,曾山,等.基于机器视觉的株间机械除草装置的作物识别与定位方法[J].农业工程学报,2013,29(10):12-18.HU Lian,LUO Xiwen,ZENG Shan,et al. Plant recognition and localization for intra-row mechanical weeding device based on machine vision[J]. Transactions of the CSAE,2013,29(10):12-18.(in Chinese)
[7] LI N,GRIFT T E,YUAN T,et al. Image processing for crop/weed discrimination in fields with high weed pressure[C]∥Michigan:American Society of Agricultural and Biological Engineers,2016.
[8] CHRISTINE M,ONYANGO J A. Segmentation of row crop plants from weeds using colour and morphology[J]. Computers and Electronics in Agriculture,2003,39(3):141-155.
[9] SI C. Weed-plant discrimination by machine vision and artificial neural network[J]. Biosystems Engineering,2002,83(3):275-280.
[10]刘洪臣,陈忠建,冯勇.结合颜色和形态特征的杂草实时识别方法[J].光电工程,2006,33(7):96-100.LIU Hongchen,CHEN Zhongjian,FENG Yong. Real-time weed recognition method based on color and morphological features[J]. Opto-Electronic Engineering,2006,33(7):96-100.(in Chinese)
[11]李颖,张立福,严薇,等.地面成像光谱数据的田间杂草识别[J].遥感学报,2013,17(4):855-871.LI Ying,ZHANG Lifu,YAN Wei,et al. Weed identification using imaging spectrometer data[J]. Journal of Remote Sensing,2013,17(4):855-871.(in Chinese)
[12]乔永亮,何东健,赵川源,等.基于多光谱图像和SVM的玉米田间杂草识别[J].农机化研究,2013,35(8):30-34.
[13] CHEN Y,JIN X,TANG L,et al. Intra-row weed recognition using plant spacing information in stereo images[C]∥Michigan:American Society of Agricultural and Biological Engineers,2013.
[14]陈树人,沈宝国,毛罕平,等.基于颜色特征的棉田中铁苋菜识别技术[J].农业机械学报,2009,40(5):149-152.CHEN Shuren,SHEN Baoguo,MAO Hanping,et al. Herb detection from cotton field based on color feature[J]. Transactions of the Chinese Society for Agricultural Machinery,2009,40(5):149-152.(in Chinese)
[15] HERRMANN I,SHAPIRA U,KINAST S,et al. Ground-level hyperspectral imagery for detecting weeds in wheat fields[J].Precision Agriculture,2013,14(6):637-659.
[16] GARCIA I D,MONTALVO M,GUERRERO J M,et al. Automatic detection of curved and straight crop rows from images in maize fields[J]. Biosystems Engineering,2017,156:61-79.
[17]张志斌,赵帅领,罗锡文,等.基于SURF算法的绿色作物特征提取与图像匹配方法[J].农业工程学报,2015,31(14):172-178.ZHANG Zhibin,ZHAO Shuailing,LUO Xiwen,et al. Matching method of green crops based on SURF feature extraction[J].Transactions of the CSAE,2015,31(14):172-178.(in Chinese)
[18] NIEUWENHUIZEN A T,HOFSTEE J W,HENTEN E J. Adaptive detection of volunteer potato plants in sugar beet fields[J].Precision Agriculture,2010,11(5):433-447.
[19]傅隆生,冯亚利,ELKAMIL T,等.基于卷积神经网络的田间多簇猕猴桃图像识别方法[J].农业工程学报,2018,34(2):205-211.FU Longsheng,FENG Yali,ELKAMIL T,et al. Image recognition method of multi-cluster kiwifruit in field based on convolutional neural networks[J]. Transactions of the CSAE,2018,34(2):205-211.(in Chinese)
[20]彭红星,黄博,邵园园,等.自然环境下多类水果采摘目标识别的通用改进SSD模型[J].农业工程学报,2018,34(16):155-162.PENG Hongxing,HUANG Bo,SHAO Yuanyuan,et al. General improved SSD model for picking object recognition of multiple fruits in natural environment[J]. Transactions of the CSAE,2018,34(16):155-162.(in Chinese)
[21]周云成,许童羽,郑伟,等.基于深度卷积神经网络的番茄主要器官分类识别方法[J].农业工程学报,2017,33(15):219-226.ZHOU Yuncheng,XU Tongyu,ZHENG Wei,et al. Classification and recognition approaches of tomato main organs based on DCNN[J]. Transactions of the CSAE,2017,33(15):219-226.(in Chinese)
[22] REN S,HE K,GIRSHISK R,et al. Faster R-CNN:towards real-time object detection with region proposal networks[C]∥Montréal:International Conference on Neural Information Processing Systems,2015.
[23] GIRSHISK R,DONAHUE J,DARRELL T,et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]∥Ohio:Conference on Computer Vision and Pattern Recognition,2014.
[24] GIRSHISK R. Fast R-CNN[C]∥Santiago:International Conference on Computer Vision,2015.
[25] HE K,ZHANG X,REN S,et al. Identity mappings in deep residual networks[C]∥European Conference on Computer Vision,2016:630-645.
[26] SRIVASTAVA N,HINTON G,KRIZHEVSKY A,et al. Dropout:a simple way to prevent neural networks from overfitting[J].Journal of Machine Learning Research,2014,15(1):1929-1958.