【作者】 王桂琴；

【导师】 郑丽敏；

【作者基本信息】 中国农业大学 ， 作物栽培与耕作学， 2004， 硕士

【摘要】 本研究通过田间实验数据及历史相关数据采集以及图像处理技术的应用，分析了不同施肥处理对产量形成过程的影响，建立产量形成过程的动态监测模型和产量预测模型。通过四大模块：图像预处理子系统、图像处理子系统、产量动态监测模型子系统及产量预测模型子系统构建成完整的系统，并利用VC++的编码技术实现了整个系统的功能。得到的主要结果如下： 1、建立中优9507和京411在产量形成过程中的各生理变化因子的动态监测模型，这此生理指标主要包括叶面积指数、干物重和分蘖数；绿度因子有植株SPAD值和倒三叶SPAD值；以及植株氮累积量、叶片氮含量(％)和籽粒氮含量(％)。通过上述各因子在整个生育期内的动态监测模型，达到对最终形成产量的动态监测。 2、据本试验实际情况，对中优9507和京411的五个施氮处理(高肥、中高、中肥、中低和低肥)，依不同的生育期，利用逐步回归分析及分段控制逐步逼近的原则，分别建立不同生理指标(干物重、叶面积指数和叶绿素含量)对产量的动态预测模型。利用统计方法检验了模型对产量预测的准确程度。 3、通过图像处理技术的应用，提取小麦图像形态特征参数和颜色特征参数，形态特征参数主要指小麦ILAI(图像叶面积指数)；颜色特征参数包括图像的R(红)、G(绿)、B(蓝)、H(色调)、I(亮度)、S(饱和度)等与颜色相关的参数及其它们的统计参量，结合与之相关生理指标对产量的动态预测模型，间接实现了图像统计参数(ILAI和相对H值)对小麦产量的预测。 4、欧氏距离法、直方图交叉法和直接差值法的比较应用，选取欧氏距离法为图像颜色模式匹配的方法，通过图像的直接匹配达到对小麦的产量预测，对其预测结果进行检验可达到图像匹配要求的精度。并且分别就上述三种方法：ILAI产量预测法、相对H值产量预测法和颜色匹配产量预测法进行统计检验分析及其方法评价。 5、为了削减由于外界环境条件引起的图像噪声，本研究据小麦图像的具体特点，采用相对色差法来较正：图像色度值减去由标准绿色比色卡计算出的环境误差值来代替图像中原来的色度值。检验证明该方法在对室外图像的处理问题上有重要的借鉴意义。 6、在VC++6.0编程环境中实现系统的所有功能，该系统可在同一界面内根据用户的不同需要，有选择性的实现小麦产量的动态预测不同方法的选择，不同生理指标的动态监测以及图像处理技术部分和图像预处理技术部分的实际应用。更多还原

【Abstract】 The paper is to obtain data from the test and relative fields and applicate image processing technology, in the same time analyse the relation about different nitrogen between wheat yields. The models about wheat yield building and preview were build The whole system was founded of basing on VC++6.0 language, it includes four parts: image pretreatment subsystem, imgae processing subsystem, wheat yield dynamic forecast subsystem and yield prediction subsystem.Conclusions are drawed as following:1, In the study, two wheat(Zhongyou9507 and Jing411) are used as the ecperimental varieties. Building the dynamical inspect model about some physiology indexes. Their yield building course is inspected by different physiology index. These physiology indexes main are leaf area index, biomass, tilling, plant SPAD, the last three leaf SPAD and plant N cumulate quantity, leaf Nitrogen(%) and grain Nitrogen(%) content. The wheat yield is dynamic inspect by above physiology indexes in whole course of wheat growth.2, By the exeperiment, according to the different Nitrogen content of Zhongyou9507 and Jing411(high Nitrogen, middle high Nitrogen, middle Nitrogen, middle low Nitrogen and low Nitrogen) builds dynamic forecast models of wheat leaf area index and chlorophll. The models applicate regressive analytical method step by step and subsection control and approach slowly and slowly. The two way veracity is tested about wheat yield forecast by statistical methods.3 , By image processing technology, obtain wheat’s shape and color character parameter. The shape character parameter main is wheat leaf area index in image, and the color character parameters are red, green, blue, hue, lightness and saturation and so on. The system obtain these parameters and their statistical numbers, then combine the physiology index model at last can predict wheat yield indirectly.4, The study use the way of Euclid distances, cross of maps of distribution and direct dispersion way of wheat image, and compare the result of the three ways. In the system the way of Euclid distances is used to match image. By the detected image matching the last similar image in the image database, the system offer the forecast result of wheat yield. In the paper test some forecast data indicate the way can be used forecast wheat yield. ILAI to forecast wheat yield, relative hue value and wheat image color matching three methods are tested and appraised by statistical way.5, The study use relative color dispersion for deduce the image yawp of environment. The method is that the number of the dispersion between the hue of original image and environment error obtained by compare with uniform green bar stead of the hue original image. The statistical test proves this way has important meaning to processing image of extraventricular image.6,The system is come true by VC++6.0 language. In this system the user can choice different part by different needs. It includes different methods are choiced to wheat yield forecast, different physiology index dynamic inspect and forecast, image processing technology and image pretreatment technology part.更多还原