臺灣博碩士論文加值系統

本研究以實驗室現有的14頻道燈箱為基礎，設計一款可以藉由輸入光參數後，快速得到符合輸入光參數光譜的光譜生成器。利用基因演算法來生成不同照度的數據集，使用數據集來訓練神經網路，透過不同超參數組合來評估表現良好的模型，並將該模型設計成光譜生成器。本研究的光譜生成器分成無拓展輸入與拓展輸入兩種，無拓展輸入情況下，光譜生成時間為0.01至0.02秒，而拓展輸入時，光譜生成時間為0.19秒，雖然時間比無拓展輸入久但可以預測較大的範圍。
透過基因演算法生成相關色溫(Correlated Color Temperature, CCT)為3000 K、4000 K、5000 K與6500 K，照度(Illuminance, Ev)為500 lux、600 lux、700 lux、750 lux、800 lux、900 lux以及1000 lux的數據集，其中會包含14頻道權重、CCT、色偏差值(Delta u-v, Duv)、顏色保真度指數(Color Fidelity Index, Rf)以及晝夜節律刺激值(Circadian Stimulus, CS)等光參數。
在訓練神經網路的部分，會先將各照度的數據集拆分成訓練集、驗證集以及測試集，再依照實驗規劃將不同照度的訓練、驗證與測試集組合成新的訓練、驗證與測試集，並對訓練集與驗證集進行二維過採樣，使用訓練集與驗證集進行訓練。為了使神經網路的訓練時間以及訓練誤差達到最小，本研究比較了不同照度的數據集組合以及不同的超參數組合，來使神經網路可以用最少的數據以及最短的訓練時間來達到最好的效果。
研究結果顯示，當改變Ev時，CS值也會跟著變大，但Ev與CS並不為線性關係，Ev越大時，CS值的最大值與最小值的範圍將會變小。CS值也會受到藍色-黃色機制(b-y)的影響，使4000 K的CS值低於其他三種色溫。訓練神經網路時，適當的正規化值(L2 regularization)可以有效的降低神經網路的預測誤差。最終光譜生成器將可預測不同照度下不同照明參數的光譜，只要是位於數據集內的光參數，光譜皆可以順利被預測生成。

In this study, we design a spectrum generator based on the 14-channel illuminator that can quickly produce spectra matching the input lighting parameters. A genetic algorithm is used to generate datasets of different illuminances, which are then used to train neural networks. Various hyperparameter combinations are evaluated to develop a well-performing model, which is then used to create the spectrum generator.
The spectrum generator in this study is divided into two types: one with non-extended input and one with extended input. For non-extended input, the spectrum generation time is 0.01 to 0.02 seconds, while for extended input, it is around 0.19 seconds. Although the extended input takes a longer time, it allows for the prediction of a broader range.
Using a genetic algorithm, datasets were generated with Correlated Color Temperatures (CCT) of 3000 K, 4000 K, 5000 K, and 6500 K, and Illuminance (Ev) of 500 lux, 600 lux, 700 lux, 750 lux, 800 lux, 900 lux, and 1000 lux. These datasets include lighting parameters such as 14-channel weights, CCT, Delta u-v (Duv), Color Fidelity Index (Rf), and Circadian Stimulus (CS).
For the neural network training, the dataset for each illuminance is first split into training, validation, and test sets. According to the experimental plan, these sets are then combined into new training, validation, and test sets with different illuminances. The training and validation sets undergo two-dimensional oversampling and are used for training. To minimize the training time and error of the neural network, this study compares different combinations of illuminance datasets and various hyperparameter configurations. The goal is to achieve the best performance with the least amount of data and the shortest training time.
The research results indicate that as the Ev increases, the CS value also increases. However, the relationship between Ev and CS is not linear. As Ev becomes larger, the range between the maximum and minimum CS values decreases. The CS value is also influenced by the blue versus yellow color mechanism (b-y), resulting in lower CS values for 4000 K compared to the other three CCTs. When training the neural network, an appropriate L2 regularization value effectively reduces the prediction error. Ultimately, the spectrum generator can predict the spectrum of different lighting parameters under various illuminances, and as long as the lighting parameters are within the dataset, the spectra can be successfully generated.

摘要 v
Abstract vii
致謝 ix
目錄 x
圖目錄 xiv
表目錄 xxii
第一章 緒論 1
1-1 研究背景 1
1-2 研究動機與目的 2
1-3 論文架構 2
第二章 文獻探討 4
2-1 光譜及基本照明光參數介紹 4
2-2 顏色保真度指數(Color Fidelity Index, Rf) 11
2-3 照明對生理影響 17
2-3-1 非視覺效應(Non-visual Effects) 18
2-3-2 晝夜節律刺激值(Circadian Stimulus, CS) 21
2-4 多頻道燈具之特性 24
2-5 基因演算法(Genetic Algorithm, GA) 28
2-6 神經網路(Neural Network, NN) 31
第三章 光譜生成器的方法與步驟 36
3-1 實驗設備 36
3-1-1 可調式光譜多頻道燈箱 36
3-1-2 光源照明分析軟體 Q_LEDNavigator 38
3-1-3 光源控制軟體照度計 THOUSLITE FS 39
3-1-4 色彩照度計 CL-70F 40
3-2 實驗設計 41
3-2-1 模擬環境配置 41
3-2-2 實驗流程 42
3-3 實驗方法 47
3-3-1 不同照度集生成方法 48
3-3-2 神經網路訓練方法 53
3-3-2-1 比較不同過採樣 58
3-3-2-2 比較不同照度數據集組合 60
3-3-2-3 神經網路超參數優化 62
3-3-3 建構光譜生成器方法 67
第四章 實驗結果與討論 70
4-1 數據集統計及分析結果 70
4-2 神經網路預測結果 86
4-2-1 比較過採樣結果 86
4-2-2 不同照度組合之神經網路預測結果 89
4-2-2-1 組合500、1000 lux照度數據集 90
4-2-2-2 組合500、750、1000 lux照度數據集 93
4-2-2-3 組合500、700、800、1000 lux照度數據集 96
4-2-2-4 不同照度組合結果討論 99
4-2-3 超參數優化結果 101
4-3 光譜生成器結果 106
4-3-1 無拓展輸入光譜生成器結果 108
4-3-2 拓展輸入光譜生成器結果 122
4-3-3 LEDCube實際照明結果 127
4-4 光譜生成器結果討論 138
第五章 結論與未來展望 142
5-1 結論 142
5-2 未來展望 144
參考文獻 145
附錄A 無拓展輸入光譜生成器預測結果 150
附錄B 拓展輸入光譜生成器預測結果 190

[1] Texas Instruments Incorporated Technical Document Report, Semiconductor Single-Crystal Circuit Development, Contract No. AF33(616)-6600, Rept. No ASD-TDR-63-281. 1963.
[2] T. Pulli, T. Donsberg, T. Poikonen, F. Manoocheri, P. Karha, and E. Ikonen, “Advantages of white LED lamps and new detector technology in photometry,” Light: Science & Applications, 4, e332, 2021.
[3] S. Nakamura, M. Senoh, and T. Mukai, “P-GaN/N-InGaN/N-GaN double-heterostructure blue-light-emitting diodes,” Japanese Journal of Applied Physics, 32(Part 2, No.1A/B), L8–L11, 1993.
[4] R. Shahidi, R. Golmohammadi, M. Babamiri, J. Faradmal, and M. Aliabadi, “Effect of warm/cool white lights on visual perception and mood in warm/cool color environments,” Experimental and Clinical Sciences, 20, 1379–1393, 2021.
[5] I. Hirotake et al., “Intellectual productivity under task ambient lighting,” Lighting Research and Technology, 50, 237–252, 2016.
[6] G. C. Brainard et al., “Action spectrum for melatonin regulation in humans: evidence for a novel circadian photoreceptor,” Journal of Neuroscience, 21, 6405–6412, 2001.
[7] K. Thapan, J. Arendt, and D.J. Skene, “An action spectrum for melatonin suppression: evidence for a novel non-rod, non-cone photoreceptor system in humans,” The Journal of Physiology, 535, 261–267, 2001.
[8] D. M. Berson, F. A. Dunn, and M. Takao, “Phototransduction by retinal ganglion cells that set the circadian clock,” Science, 295(5557), 1070–1073, 2002.
[9] P. R. Mills, S. C. Tomkins, and L. J. M. Schlangen, “The effect of high correlated colour temperature office lighting on employee wellbeing and work performance,” Journal of Circadian Rhythms, 5:2, 2007.
[10] Y. Liu et al., “The impact and model of CS and CCT on alertness,” 2020 17th China International Forum on Solid State Lighting, Shenzhen, China, 2020.
[11] 鄭羽禎. 基於神經網路之多頻道燈箱光譜生成器. 國立中央大學, 碩士論文, June 2023.
[12] A.K.R. Choudhury. Principles of Colour and Appearance Measurement, Woodhead Publishingm, 2014.
[13] WIKIMEDIA COMMONS. File:PlanckianLocus.png. Source：https://commons.wikimedia.org/wiki/File:PlanckianLocus.png, accessed on 2024/04/25.
[14] Y. Ohno. Fundamentals of Colorimetry and Practical Color Measurements. Source：
https://www.lites.asia/files/otherfiles/0000/0413/Session_2_Fundamentals_of_Colorimetry_and_Practical_Color_Measurements.pdf, accessed on 2024/04/25.
[15] WIKIMEDIA COMMONS. File:Planckian-locus.png. Source：https://commons.wikimedia.org/wiki/File:Planckian-locus.png, accessed on 2024/04/25.
[16] NVC Lighting. What is correlated colour temperature. Source：https://www.nvcuk.com/technical-support/view/what-is-correlated-colour-temperature-%28cct%29-44, accessed on 2024/04/25.
[17] Y. Ohno and M. Fein, “Vision experiment on acceptable and preferred white light chromaticity for lighting,” CIE x039:2014, Proceedings of CIE 2014 Lighting Quality and Energy Efficiency, Kuala Lumpur, Malaysia, 2014.
[18] Commission Internationale de l’Eclairage. CIE S 017/E:2020, ILV: International Lighting Vocabulary, 2nd edition. 2020. Source:
https://cie.co.at/e-ilv, accessed on 2024/04/25.
[19] WIKIMEDIA COMMONS. File:
Munsell_1943_color_solid_cylindrical_coordinates.png. Source：https://en.wikipedia.org/wiki/Munsell_color_system, accessed on 2024/04/25.
[20] Commission Internationale de l’Eclairage. CIE 224:2017, CIE 2017 Colour Fidelity Index for accurate scientific use. 2017.
[21] Commission Internationale de l’Eclairage. CIE 13.3, Method of Measuring and Specifying Colour Rendering Properties of Light Sources. 1995.
[22] Illuminating Engineering Society. IES-TM-30-15: Method for Evaluating Light Source Color Rendition. 2015.
[23] Illuminating Engineering Society. IES-TM-30-20: Method for Evaluating Light Source Color Rendition. 2020.
[24] M. P. Royer. “Tutorial: Background and Guidance for Using the ANSI_IES TM_30 Method for Evaluating Light Source Color Rendition,” LEUKOS, 18(2), 191-231, 2022.
[25] M. R. Luo, G. Cui, and C. Li. “Uniform colour spaces based on CIECAM02 colour appearance model,” Color Research & Application, 31(4), 320-330, 2006.
[26] H. T. Gu, M.R. Luo, and X. Y. Liu. “Testing different colour rendering metrics using colour difference data,” Lighting Research & Technology, 49(5), 539-560, 2017.
[27] M. Wei, M. Royer, and H. P. Huang. “Perceived colour fidelity under LEDs with similar Rf but different Ra,” Lighting Research & Technology, 51(6), 858-869, 2019.
[28] Y. Ohno. “Practical Use and Calculation of CCT and Duv,” LEUKOS, 10(1), 47-55, 2014.
[29] CNS 12112 中華民國國家標準. 室內工作場所照明. 經濟部標準檢驗局, January 2012.
[30] A. Kruithof, “Tubular luminescence lamps for general illumination,” Philips technical review, 6, 65–73, 1941.
[31] WIKIPEDIA. Kruithof curve. Source：
https://en.wikipedia.org/wiki/Kruithof_curve, accessed on 2024/04/27.
[32] H. Xiao, H. Cai, and X. Li. “Non-visual effects of indoor light environment on humans: A review,” Physiology & Behavior, 2020.
[33] J. L. Souman et al. “Spectral tuning of white light allows for strong reduction in melatonin suppression without changing illumination level or color temperature,” Journal of Biological Rhythms, 33(4), 420-431, 2018.
[34] T. M. Brown. “Melanopic illuminance defines the magnitude of human circadian light responses under a wide range of conditions,” Journal of Pineal Research, 69(1), e12655, 2020.
[35] C. Lok, “Vision science: Seeing without seeing,” Nature, 469:284–285, 2011.
[36] L. J. M. Schlangen, and L. L. A. Price. “The lighting environment, its metrology, and non-visual responses,” Front Neurol, 12(624861), 2021.
[37] LumosTech. Science behind adjusting your circadian rhythm with light. Source：https://lumos.tech/science-behind-adjusting-circadian-rhythm/ , accessed on 2024/04/27.
[38] M. S. Rea, M. G. Figueiro, J. D. Bullough, and A. Bierman, “A model of phototransduction by the human circadian system,” Brain Research Reviews, 50(2), 213–228, 2005.
[39] M. S. Rea and M. G. Figueiro, “Light as a circadian stimulus for architectural lighting,” Lighting Research & Technology, 0, 1–14, 2016.
[40] H. J. Tian, T. Chen, Y. Hu, T. Guan, and M. P. Cai, “Change of circadian effect with colour temperature and eye spectral transmittance at different ages,” Lighting Research & Technology, 53(1), 41–53, 2021.
[41] T. Esposito and K. Houser, “Correlated color temperature is not a suitable proxy for the biological potency of light,” Scientific Reports, 12(1), 20223, 2022. 
[42] Commission Internationale de l’Eclairage. CIE S 026/E:2018, CIE System for Metrology of Optical Radiation for ipRGC-Influenced Responses to Light. 2018.
[43] M. L. Lo, T. H. Yang, and C. C. Lee, “Fabrication of a tunable daylight simulator,” Applied Optics, 50(9), C95–9, 2011.
[44] F. Zhang, H. Xu, and Z. Wang, “Spectral design methods for multi-channel LED light sources based on differential evolution,” Applied Optics, 55(28), 7771-7781, 2016.
[45] A. Llenas and J. Carreras, “Arbitrary spectral matching using multi-LED lighting systems,” Optical Engineering, 58(3), 035105, 2019.
[46] B. Zandi, A. Eissfeldt, A. Herzog, and T. Q. Khanh, “Melanopic limits of metamer spectral optimisation in multi-channel smart lighting systems,” Energies, 14(3), 527, 2021.
[47] J. H. Holland, Adaptation in Natural and Artificial Systems, The MIT Press, 1992.
[48] M. Warren, and W. Pitts, “A logical calculus of ideas immanent in nervous activity,” Bulletin of Mathematical Biophysics, 5(4), 115–133, 1943.
[49] AIM. Artificial Neural Networks 101. Source：
https://analyticsindiamag.com/artificial-neural-networks-101/, accessed on 2024/05/04.
[50] Deep Learning with R. Source：
https://livebook.manning.com/book/deep-learning-with-r/chapter-3/v-1/20, accessed on 2024/05/04.
[51] D. Kingma and J. Ba, “Adam: A method for stochastic optimization,” arXiv preprint arXiv:1412.6980, 2014.
[52] K. P. Murphy, Machine learning - a probabilistic perspective, The MIT press, Adaptive computation and machine learning series, 2012.
[53] J. Bergstra, and Y. Bengio, “Random search for hyper-parameter optimization,” Journal of Machine Learning Research, 13, 281–305, 2012.
[54] LEDCube. Source：
https://www.thouslite.com/product_detail/124.html , accessed on 2024/05/13.
[55] THOUSLITE LEDCube中文使用手冊V3, 2015.
[56] CRI Illuminance Meter CL-70F. Source：
https://www.konicaminolta.com/instruments/download/catalog/light/pdf/cl70f_catalog_eng.pdf, accessed on 2024/05/13.
 
[57] D. Durmus, ANSI/IES TM-30-20. MATLAB Central File Exchange.
https://www.mathworks.com/matlabcentral/fileexchange/95378-ansi-ies-tm-30-20 , accessed on 2024/03/27.
[58] National Electrical Manufacturers Association, American National Standard for Electric Lamps Specifications for the Chromaticity of Solid-state Lighting Products, ANSI-C78.377, 2015.
[59] J. C. Duchi, H. Elad, and S. Yoram, “Adaptive subgradient methods for online learning and stochastic optimization,” Journal of Machine Learning Research, 12, 2121-2159, 2011.
[60] T. Tieleman, and G. Hinton, “Lecture 6.5 - RMSProp, COURSERA:Neural Networks for Machine Learning,” Technical report, 2012.
[61] S. Loffe, and C. Szegedy, “Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift,” arXiv preprint arXiv:1502.03167, 2015.
[62] G. Obozinski, B. Taskar, and M. Jordan, “Joint covariate selection for grouped classification,” Technical Report 743, Dept. of Statistics, University of California Berkeley, 2007.
[63] L. N. Smith, “A disciplined approach to neural network hyper-parameters: Part 1 - learning rate, batch size, momentum, and weight decay,” arXiv preprint ArXiv: 1803.09820, 2018.
[64] C. Ferreira, “Gene expression programming: A new adaptive algorithm for solving problems,” Complex Systems, 13(2):87–129, 2001.