Leaf vector的問題，透過圖書和論文來找解法和答案更準確安心。 我們找到下列包括價格和評價等資訊懶人包

應用高光譜影像檢測技術開發蓮霧果實病害預測之演算法

為了解決Leaf vector的問題，作者楊典穎 這樣論述：

蓮霧為台灣重要的經濟果樹，在採收後的貯藏期間可能會發生各種的病害，這些受到病菌感染的果實在初期不會表現出病徵，導致選別人員沒能分辨出來，若果實在沒被察覺病徵的情況下進行販售，待病害症狀加劇後，後果將不堪設想。由於高光譜成像技術能夠提早偵測植物病害的潛伏期，所以可以利用此技術來輔佐分級人員控管蓮霧的品質，不僅能夠更客觀地評斷果實，還能即早發現未顯現病徵的果實，因此，本研究應用高光譜影像數據建立一套蓮霧果實病害預測的演算法流程，將蓮霧的高光譜影像數據經由預處理、數據提取、降維、回歸和分類來達到對蓮霧果實病害預測的目的，降維使用了連續投影算法(Successive Projections Algo

rithm, SPA)和主成分分析(Principal Component Analysis, PCA)來減少高光譜影像的龐大數據，回歸使用了最小平方支持向量回歸(Least Squares Support Vector Regression, LS-SVR)來預測蓮霧在發病過程中的水分散失，分類使用了K-近鄰(K-Nearest Neighbors, KNN)演算法來建立蓮霧果實病害的分類模型，分類模型對感興趣區域(Region Of Interests, ROI)的光譜進行分類，其模型的準確率達99%，最後，利用視覺化的方式來呈現蓮霧果實的病害發生情形，這項研究可能有助於減少蓮霧的採後處理

損失。

深度學習方法應用於高光譜之特徵多光譜萃取-以蓮霧糖度預測為例

為了解決Leaf vector的問題，作者陳致融 這樣論述：

摘要 iAbstract iiAcknowledge ivContent vList of Figures viiiList of Tables xiiiChapter 1 Introduction 11.1 Smart Agriculture 11.2 Motivation 21.3 Introduction of Syzygium samarangense 5Chapter 2 Hyperspectral Imaging Technique 62.1 Basic Theory 62.1.1 Electromagnet

ic radiation 62.1.2 Radiometry 62.1.3 Geometrical optics 82.1.4 Absorbance, Transmittance, and Reflectance of material 132.2 Optical and Infrared Spectroscopy 142.2.1 Spectroscopy 142.2.2 Hyperspectral Spectroscopy 142.2.3 Multispectral Spectroscopy 152.2.2 Hyperspectral Imag

ing 162.3 Hyperspectral Instrument 172.3.1 Structure of hyperspectral imaging system 172.3.2 Sensing element of spectrometer 182.3.3 Types of Hyperspectrometers 192.4 Applications of Hyperspectral Imaging 23Chapter 3 Methodology of Hyperspectral Analysis and Deep Learning Models

263.1 Calibration of Instrument 263.1.1 Spatial calibration 263.1.2 White and Dark calibration 263.1.3 Spectral calibration 273.1.4 Spectra smoothing 283.2 Modeling of Hyperspectral Data 293.2.1 Dimension Reduction 293.2.2 Deep Learning Regression Modeling for Hyperspectral Ima

ges 303.2.4 Loss Function 363.2.5 Optimizer 373.3 Error evaluation 393.4 Related Research of Deep Learning Model to Hyperspectral Imaging 40Chapter 4 Methodology of Hyperspectral Conversion and Signature-band Extraction 434.1 Hyperspectral to Multispectral Conversion 434.2 Ex

plainable Artificial Intelligence 444.2.1 Occlusion 444.2.2 Saliency Map 444.2.3 Integrated Gradient 454.3 Related Research of Signature-band Extraction on Spectral Data 464.4 Multispectral Imaging Instrument 48Chapter 5 Experimental Design of Sugariness Prediction of Syzygium sama

rangense with Hyperspectral Data 495.1 Hyperspectral Data Preparation 495.1.1 Preparing Samples 505.1.2 The Procedure of Hyperspectral Measurement 505.1.3 Sugariness Measurement – Labelling 505.2 Hyperspectral Data Pre-processing 515.2.1 White/Dark Calibration on the Hyperspectral

Data 515.2.2 3-Dimensional and 1-Dimensional Data Type – ROI Sampling 515.2.3 Data sampling and splitting for modeling 525.3 Evaluation of Modeling of Hyperspectral datasets 545.3.1 Evaluation by Hyperspectral Data Visualization Using t-SNE 545.3.2 Evaluation Over Deep Learning Regres

sion Models 56Chapter 6 Results of Sugariness Prediction of Syzygium samarangense with Hyperspectral Data 606.1 The Data Visualization Results of the HSIs dataset 606.2 The Hyperspectral Modeling Results 616.3 Evaluation of the Learning Results of Hyperspectral data Modeling by Visualizi

ng the Inputs fed to the Last Layer 64Chapter 7 Experimental Design of Sugariness Prediction of Syzygium samarangense with Multispectral Data and Verification using the Hand-Held Device 677.1 Verification of the Modeling of Multispectral datasets 677.1.1 Multispectral Data Preparation 68

7.1.2 FNN Modeling and Verification using Multispectral Data 697.2 Verifications of the Bands Selection using Multispectral Datasets 707.2.1 Data Preparation of Re-sampled Multispectral Data 707.2.2 FNN Modeling and Verification using re-sampled Multispectral Data 717.3 Verifications of

the Modeling using Datasets Collected from the Hand-Held Device 727.3.1 Sample Preparation 727.3.2 Data Pre-processing 757.3.3 FNN modeling and Verification 79Chapter 8 Results of Sugariness Prediction of Syzygium samarangense with Multispectral Data and Hand-Held Device Datasets 828.

1 Results of the Modeling using Multispectral Data 828.2 Results of Bands Selection 858.3 Results of the Modeling using re-sampled Multispectral Data 868.4 Results of the Bands Selection using the Datasets Collected from the Hand-Held Device 928.5 Results of Eliminating the Outliers 9

58.6 Model Implementation with Coding from Scratch 97Chapter 9 Discussion, Conclusion and Future Work 1009.1 Discussion 1009.1.1 The Modeling Results on Hyperspectral and Multispectral Datasets 1009.1.2 Verification of Band Selection Results 1009.1.3 Modeling of the Data Collected fro

m Hand-Held Device 1019.2 Conclusion 1049.3 Future work 105References 106Publication 115