光谱学与光谱分析, 2018, 38 (9): 2757, 网络出版: 2018-10-02  

红外热成像与近红外光谱结合快速检测潜育期番茄花叶病

Rapid Detection of Tomato Mosaic Disease in Incubation Period by Infrared Thermal Imaging and Near Infrared Spectroscopy
朱文静 1,2,*李林 1,2李美清 1,2刘继展 1,2魏新华 1,2
作者单位
1 江苏大学现代农业装备与技术省部共建教育部/江苏省重点实验室, 江苏 镇江 212013
2 江苏大学农业装备工程学院, 江苏 镇江 212013
摘要
现有的番茄花叶病无损检测方法无法在潜育期内, 即显症之前进行早期识别导致施药不及时或者盲目过度施药。 设计与试制了红外热成像信息采集系统, 主要包括: 光箱、 红外热成像仪、 温度及升降控制器、 加热板和升降载物台。 该系统能够根据温度起止节点的要求, 人为调节拍摄温度。 在江苏大学现代农业装备与技术省部共建重点实验室Venlo型温室中进行非抗病性番茄品种辽宁农科院L-402的培育。 采用叶面摩擦接种花叶病毒(Tobacccco mosaic virus, ToMV), 分为轻度感染组(LI), 重度感染组(SI); LI组为磷酸缓冲液稀释500倍后的病毒液接种, SI组为病毒原液接种。 对照组(CG)喷施等量磷酸缓冲液。 接种10 d后叶片开始出现病斑, 证明接种后9 d为番茄花叶病的潜育期。 使用红外热成像系统采集了三个组共计144个样本的红外热成像图, 计算叶表最大温差(MTD) 以表征潜育期内连续9 d内的叶面温度变化情况。 CG组叶片的MTD值差异极小, 而接种后叶片MTD值随着病毒侵染时间的推进发生了显著的变化。 接种6 d后MTD值差异最大可达1.63 ℃, 第7 d开始差异逐步缩小, 表明病毒的扩散范围增大导致病叶越来越多的区域被侵染使得整体叶温上升。 光谱采集采用两种方法进行, 一种是根据热像图的MTD值计算判别出温度突变区域后采集光谱, 记为热像采集法(TCM); 另一种是不考虑病灶位置, 在叶尖、 叶中、 叶基三个区域分别随机选择一个点采集光谱后求平均值, 记为随机采集法(RCM)。 TCM确定三个光谱采集点的选择原则是: LI组接种后3, 6和9 d的温度突变区域平均MTD值比CG组温度分别高出0.3, 0.7和0.5 ℃。 SI组接种后3, 6和9 d的温度突变区域平均MTD值比CG组温度分别高出0.5, 1.2, 0.8 ℃。 差值达到此标准的病灶位置才定为TCM的可选区域。 对所有样本采用支持向量机(SVM)算法建立识别模型。 采用主成分分析对2 151个波长点的光谱信息进行压缩, 前6个主成分所对应的累积方差贡献率已到达99%。 分别对感病3, 6和9 d的样本按照2∶1的比例划分校正集和预测集, 对预测集样本的病害程度进行识别。 两种方法所建立的模型的总识别率分别为92.59%和99.77%。 采用TCM建立的光谱识别模型中仅有接种后3 d的一个LI组样本未能识别出来, 被误判成CG组样本外, 其余组识别率均达到了100%。 结果表明近红外光谱法识别番茄花叶病是可行的。 采用红外热成像结合近红外光谱法能够建立识别率更高的番茄花叶病潜育期识别模型, 克服点源采样随机性, 对后续管控流程和突破作物早期精准用药的关键技术探索, 建立更为精准的温室智能施药系统提供了新的思路。
Abstract
The lagging diagnosis method of tomato mosaic disease results in untimely and excessive application of pesticide. The conventional nondestructive testing methods were unable to be applied at early recognition in the incubation period. In this study, the infrared thermal imaging information acquisition system was designed. The efficiency and accuracy of this system were also tested. The main components of the system included a shell box, an infrared thermal image acquirer, a temperature and lift controller, a heating plate and a lift load table. The system developed in this study has the capacity to adjust the shooting temperature manually according to the requirements of the temperature range in a typical experiment. To test the precision of the system, the non-resistant tomatos variety L-402 were cultivated by Institute of Vegetables of Liaoning Academy of Agricultural Sciences in the Venlo type greenhouse of the Ministry & Provinces?Co - construction Key Laboratory of modern agricultural equipment and technology of Jiangsu University. The virus (Tobacccco mosaic virus, ToMV) infection experiment was conducted by using the method of leaf surface friction before the flowering stage. In the virus infection experiment, tomato plants were divided into three groups. The severe infection group (SI) was inoculated with the original virus solution. The Low-grade infection group (LI) was inoculated with diluted virus solution (500 times dilution by phosphate buffer). The control group (CG) was sprayed with equal amount of phosphate buffer. After 10 days of inoculation, spots began to appear on leaves of tomato plants in SI group, suggesting that the first 9 days were the incubation period of tomato mosaic disease. Infrared thermal imaging system was used to collect infrared thermal imaging of those three groups with a total sample size of 144 during the incubation period. The maximum temperature difference (MTD) of the leaf table was calculated to characterize the change of leaf temperature in continuous 9 days during the incubation period. The MTD value of the leaves in the CG group was statistical non-significant, but the MTD value of the leaves in both LI and SI groups was significantly changed after inoculation with the infection time of the virus. After six days of inoculation, the maximum difference of MTD value can reached 1.63 ℃. The difference gradually narrowed down from 7 days, indicating that the virus were spread to more and more regions on the infected leaves and raised the temperature of the whole leaf. Two spectral acquisition methods were conducted. The first one was Thermal-imaging collection method (TCM). During TCM, spectra were intensively acquired during the temperature mutation region which was calculated based on the MTD value from the infrared thermograph. The second method was to acquire spectrum on randomly selected points on the tip, middle, and base of leaves without focusing on the location of the lesion. This spectrum acquisition method was recognized as random collection method (RCM). The principle of TCM to select the effective position for the three spectral acquisition points was that the average MTD value of the mutation zone in the LI group was 0.3, 0.7 and 0.5 ℃ higher than those in the CG group on the 3rd, 6th and 9th day after inoculation respectively.The average MTD value of the mutation zone in the SI group was 0.5, 1.2 and 0.8 ℃ higher than those in the CG group on the 3rd, 6th and 9th day after inoculation, respectively. Lesion position met the above criteria could be considered as an optional area for TCM. All samples were identified by using Support Vector Machine (SVM) algorithm for discriminant analysis. The principal component analysis (PCA) was used to compress the spectral information of 2 151 wavelength points. The cumulative variance contribution rate of the first six principal components has reached 99%. The samples of 3, 6 and 9 d were divided into the correction set and the prediction set at the ratio of 2∶1, and the disease degree of the prediction set samples was identified. The total recognition rates of the models established by the two methods are 92.59% and 99.77%, respectively. In the spectral recognition model established by TCM, only one sample from LI group after 3 d was unable to be identified and mistaken into CG group. Despite this sample, the remaining recognition rate reached 100%. The results showed that it is feasible to use near infrared spectroscopy to identify tomato mosaic disease at early stage. Using infrared thermal imagingin combination with near-infrared spectroscopy technique allows us to establish higher recognition rate models for identification of tomato mosaic disease during incubation period. This study provided an alternative method for the development of follow-up control process, and created a new model to break through the bottleneck of the early precise pesticide spraying of crops. It overcame the point source of NIR sampling randomness and helped to establish a more accurate intelligent pesticide application system in greenhouse.

朱文静, 李林, 李美清, 刘继展, 魏新华. 红外热成像与近红外光谱结合快速检测潜育期番茄花叶病[J]. 光谱学与光谱分析, 2018, 38(9): 2757. ZHU Wen-jing, LI Lin, LI Mei-qing, LIU Ji-zhan, WEI Xin-hua. Rapid Detection of Tomato Mosaic Disease in Incubation Period by Infrared Thermal Imaging and Near Infrared Spectroscopy[J]. Spectroscopy and Spectral Analysis, 2018, 38(9): 2757.

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!