可见/近红外光谱技术无损识别苹果品种的研究

为了能快速无损鉴别苹果的品种,基于高光谱成像技术结合模式识别算法,分别建立了判别苹果品种K最近邻(KNN)识别模型与偏最小二乘判别分析(PLS-DA)识别模型。综合比较了不同光谱预处理方法(二阶微分(SD)、标准正态变换(SNV)和多元散射校正(MSC))对各模型识别效果的影响,并利用主成分分析方法对预处理后的光谱数据进行降维,以提取能反映苹果品种的特征光谱。结果表明:采用主成分分析法选择了累计贡献率超过99.9%的前10个主成分作为样本集特征光谱数据,很好地实现了光谱数据的降维;MSC预处理方法对光谱反射率预处理的效果最好;2种判别模型均能基本满足实际要求,且MSC+KNN识别模型的识别性能最优,对预测集样本的正确识别率高达100%。     

Hyperspectral signal decomposition and symptom detection of wheat rust disease at the leaf scale using pure fungal spore spectra as reference

This study establishes a method to detect and distinguish between brown rust and yellow rust on wheat leaves based on hyperspectral imaging at the leaf scale under controlled laboratory conditions. A major problem at this scale is the generation of representative and correctly labelled training data, as only mixed spectra comprising plant and fungal material are observed. For this purpose, the pure spectra of rust spores of Puccinia triticina and P. striiformis, causal agents of brown and yellow rust, respectively, were used to serve as a spectral fingerprint for the detection of a specific leaf rust disease. A least-squares factorization was used on hyperspectral images to unveil the presence of the spectral signal of rust spores in mixed spectra on wheat leaves. A quantification of yellow and brown rust, chlorosis and healthy tissue was verified in time series experiments on inoculated plants. The detection of fungal crop diseases by hyperspectral imaging was enabled without pixel-wise labelling at the leaf scale by using reference spectra from spore-scale observations. For the first time, this study shows an interpretable decomposition of the spectral reflectance mixture during pathogenesis. This novel approach will support a more sophisticated and precise detection of foliar diseases of wheat by hyperspectral imaging.     

Developing a spectral disease index for myrtle rust (Austropuccinia psidii)

Since 2010 Australian ecosystems and managed landscapes have been severely threatened by the invasive fungal pathogen Austropuccinia psidii. Detecting and monitoring disease outbreaks is currently only possible by human assessors, which is slow and labour intensive. Over the last 25 years, spectral vegetation indices (SVIs) have been designed to assess variation in biochemical or biophysical traits of vegetation. However, diagnosis of individual diseases based on classical SVIs is currently not possible because they lack disease specificity. Here, a novel spectral disease index (SDI), the lemon myrtle–myrtle rust index (LMMR), has been developed. The index was designed from hyperspectral leaf-clip data collected at a lemon myrtle plantation in New South Wales, Australia. A total of 236 fungicide-treated (disease free) and 228 untreated (diseased) lemon myrtle leaves were sampled and a random forest classifier was used to show that the LMMR discriminates those classes with an overall accuracy of 90%. Compared to three classical SVIs (PRI, MCARI, NBNDVI), commonly applied for stress detection, the LMMR clearly improved classification accuracies (58%, 67%, 60%, respectively). If the LMMR can be validated on independent datasets from similar and different host species, it could enable land managers to reduce disease impact by earlier control. There might also be potential to collect useful data for epidemiology models. Calculating the LMMR based on hyperspectral data collected from aerial platforms (e.g. drones) would allow for rapid and high-capacity screening for disease outbreaks.     

Effect of tetramethylpyrazine on neurocognitive impairment and functional imaging changes induced by lipopolysaccharide

AIM: To investigate the effect of traditional Chinese medicine tetramethylpyrazine on neurocognitive impairment and functional imaging changes induced by lipopolysaccharide (LPS). METHODS: Sprague-Dawley (SD) rats (〖WTBX〗n=36) were randomly divided into 6 treatment groups (n=6 for each group):control group, sham operation group, LPS group, and low dose, medium dose and high dose of tetramethylpyrazine groups. The rats in LPS group and tetramethylpyrazine groups were intracerebroventricularly injected with LPS (150 μg per rat, dissolved in cerebrospinal fluid), whereas the rats in sham operation group were administered the same volume of cerebrospinal fluid in a similar manner. The rats in low dose group, medium dose group and high dose group were intraperitoneally injected with tetra-methylpyrazine at 50 mg/kg, 100 mg/kg and 200 mg/kg, respectively. The neurocognitive impairment was accessed using Morris water maze. The protein levels of interleukin-1β (IL-1β) and tumor necrosis factor-α (TNF-α) were measured by ELISA. The functional imaging changes induced by LPS, including the diffusion motion component of the imaging pure water molecule and the blood perfusion-related diffusion movement, were determined using the imaging IVIM D and IVIM D* functional sequence. RESULTS: In Morris water maze, LPS significantly increased the escape latency, and decreased the number of crossing platform and the time in the target quadrant compared with sham operation group (P<0.05). In addition, low dose of tetramethylpyrazine treatment obviously shortened the escape latency, and increased the number of cros-sing platform and the time in the target quadrant compared with LPS group (P<0.05), but no significant difference between LPS group and medium dose or high dose group was observed. LPS exposure significantly increased IL-1β and TNF-α levels in the cortex and hippocampus compared with sham operation group (P<0.05), while low dose of tetrame-thylpyrazine treatment obviously attenuated the elevated IL-1β and TNF-α levels in the cortex and hippocampus induced by LPS (P<0.05). The imaging results showed that LPS exposure significantly decreased the IVIM D, IVIM D*, and f va-lues of the cortex and hippocampus compared with sham operation group (P<0.05), whereas low dose of tetrame-thylpyrazine treatment obviously inhibited the decreases in IVIM D, IVIM D*, and f values induced by LPS (P<0.05). However, no significant difference of the IVIM D, IVIM D*, and f values between LPS group and midium dose or high dose group was found. CONCLUSION: Tetramethylprazine attenuates neurocognitive impairment induced by LPS in the rats. The mechanism may be related to the inhibition of inflammatory response through the increases in the water molecule diffusion and the perfusion of cerebral blood flow in the brain.     

基于高光谱成像技术的泾源黄牛肉色度PLSR预测模型构建

利用高光谱成像技术对泾源黄牛肉色度的PLSR预测模型进行构建。通过可见近红外高光谱成像系统获取牛肉样本的高光谱图像,提取感兴趣区域(ROI)的光谱信息并计算平均光谱,采用蒙特卡洛法剔除异常样本后进行样本集划分,并对划分后的样本数据进行预处理。其中,亮度(L^*)经Deresolve法预处理的模型结果最好,其$R_{C}^{2}$为0.979 0,预测集相关系数$R_{P}^{2}$为0.976 6;红度(a^*)经卷积平滑法预处理的模型结果最好,其$R_{C}^{2}$和$R_{P}^{2}$分别为0.807 0、0.915 5;黄度(b^*)经卷积平滑法预处理的模型结果最好,其$R_{C}^{2}$和$R_{P}^{2}$分别为0.931 1、0.950 6。分别利用竞争性自适应重加权法(CARS)、连续投影算法(SPA)和无信息变量消除算法(UVE)进行特征波长提取,建立基于特征波段的偏最小二乘回归(PLSR)模型,进而优选出最佳预测模型,结合视觉的空间深度、立体程度,对样本的形态和色觉感知进行提取与辨别。结果表明,利用高光谱成像技术所构建的色度PLSR模型均是可行的,研究结果为牛肉品质在线快速检测提供了理论依据。     

基于高光谱成像技术的羊肉新鲜度预测

为实现羊肉新鲜度的快速、无损检测,应用高光谱成像技术对不同存储天数的羊肉建立挥发性盐基氮TVB-N(Total volatile basic nitrogen,TVB-N)含量预测模型。通过高光谱成像系统获取羊肉样本935～2 539nm的高光谱图像,选取样本左上、左下、右上、右下和中间5个位置20×20像素的方形作为感兴趣区域(Region of interesting,ROI),提取ROI并计算区域内样本平均光谱。利用二进制粒子群优化算法(Binary particle swarm optimization,BPSO)和相关系数分析法(Correlation coefficient,CC)提取羊肉TVB-N高光谱特征变量,结合偏最小二乘回归(Partial least squares regression,PLSR)和随机森林回归(Random forest regression,RFR)建模算法,分别建立表征羊肉TVB-N含量的BPSO-PLSR、BPSO-RFR、CC-PLSR、CC-RFR预测模型。依据袋外均方根误差RMSEOOB最小原则,对最佳回归子树和分裂特征2个主要参数进行寻优以提高RFR建模算法的预测精度。比较4个模型的预测效果发现,BPSO-RFR模型的预测精度最高,其校正集决定系数R_c~2和均方根误差RMSEC分别为0.87和2.99,预测集决定系数R_p~2和均方根误差RMSEP分别为0.86和3.36。综上,高光谱成像技术和机器学习算法的有机结合为快速有效检测肉品新鲜度提供了理论依据。     

基于激光图像的水果表面农药残留检测试验研究

初步探讨了运用激光图像技术来检测水果表面农药残留的方法。应用自行研制的激光图像采集系统获取水果光谱图像,通过对比研究水果表面不喷洒农药、喷洒农药、喷洒不同浓度农药的水果图像特征分析可知,一般未喷洒农药的水果灰度图像灰度值分布在1~100范围内像素点总数低于喷洒农药后的水果灰度图像;喷洒不同浓度农药水果的图像像素灰度值在1~100范围内像素点个数也有明显的差别;但喷洒农药的浓度与像素值聚集在0~100范围内的像素个数之间规律性不明显,有待进一步研究。试验研究表明,用激光图像技术检测水果表面的是否存在农药残留是可行的。     

基于热红外成像的坡面薄层水流流速测量方法

流速是表征水流水力学特性的重要物理量。为了准确获取薄层水流流速,基于热红外成像技术、计算机视觉识别技术,设计了一种薄层水流流速测量系统。该系统通过对热示踪剂的自动控制以及对其热成像图的瞬时采集、影像校正、噪点去除、质心确定等手段,获取薄层水流的流速等参数,从而实现对坡面薄层水流流速的动态观测。该系统精确度和准确度高,可从不同时间和空间尺度上更加准确地观测热示踪剂动态运移过程。系统的测量标准差为0.020 m/s,观测精度可达到98.33%,观测的时间分辨率为1/9 s,空间分辨率为2 mm。为了验证该系统的准确度,以流量法为基准,与传统示踪法（染料示踪法、盐示踪法）比较,结果表明,该系统的准确度高于染料示踪法和盐示踪法。其中热红外成像观测系统的相对误差均在?10%以内;染料示踪法的相对误差均大于10%;盐示踪法52%的相对误差在?10%以内。利用热红外成像系统获取的影像,可对不同时刻示踪段水流的发生、发展的过程进行追溯,也可以测量示踪剂沿水流方向的运移速度及垂直于水流方向的扩散速度,计算热示踪剂的弥散系数等。该技术可应用于降雨侵蚀、径流冲刷等方面的研究,对于进一步深化土壤侵蚀过程与机理研究具有重要的意义。     

基于小麦籽粒高光谱特征的品种鉴别研究

为探究高光谱成像技术在小麦籽粒品种鉴别等的研究,以15种品种已知的小麦籽粒为对象,分别采集样品的光谱和图像信息。运用主成分分析法优选三个特征波长(447nm、615nm、955nm),提取特征波长下小麦籽粒图像的形态特征(面积、周长、圆度、长轴长度、短轴长度)和纹理特征(均值、标准差、熵)。应用Bayes判别分析法进行多元统计分析,建立判别函数判别回代准确率为99.9%,交叉验证的准确率为98%,模型的判别效果良好。研究表明利用高光谱成像技术结合Bayes判别分析的方法可用于小麦籽粒品种的鉴别。