| 摘要: |
| 目的 利用快速独立成分分析法， 实现有效利用色彩光谱信息进行色彩精度复制的关键技术。方法 利用快速独立成分分析法对孟塞尔colormatt光谱数据集进行光谱空间降维， 并利用选用的独立成分进行光谱空间重建； 从累积空间覆盖率和表色精度等方面对该方法进行评价。结果 随着选用独立成分个数的增加， 累积空间覆盖率和表色精度数据逐步增大， 当选用8个独立成分时， 累积空间覆盖率和表色精度逐步处于平稳状态； 根据需求选用5个基矢量对颜色进行准确光谱重建， 最终累积空间覆盖率达到97%， 99.92%的重构光谱拟合度达0.9以上， 100%的样品色差小于0.5。结论 利用快速独立成分分析法进行光谱空间降维， 能够高精度地表示原始光谱空间。 |
| 关键词: 快速独立成分分析法 宽带多光谱空间 累积空间覆盖率 色差 光谱拟合度 |
| DOI： |
| 分类号:TS801.3 |
| 基金项目:运城学院院级科研项目（CY-2013021） |
|
| Color Signal Wide-band Multispectral Space Research Based on FICA |
|
LI Mei1, KONG Ling-wang2
|
|
1.Yuncheng Institute， Yuncheng 044000， China;2.Wuhan University， Wuhan 430079， China
|
| Abstract: |
| The aim of this work was to use fast independent component analysis to achieve effective utilization of the key technology of colour spectrum information reproducing color accuracy. Fast independent component analysis was applied to munsell colormatt spectroscopy data set for dimension reduction of the space, and the chosen independent components were used to reconstruct spectral space, finally this method was evaluated from the two aspects of accumulated space coverage and colorful accuracy. With the increase of the chosen number of independent components, the cumulative spatial coverage and colorful accuracy data increased gradually. When eight independent components were selected, the cumulative space coverage and colorful precision tended to be steady gradually. According to the demand, five base vectors were selected for accurate color spectrum reconstruction, and eventually the cumulative space coverage reached 97%, the fitting degree of 99.92% of the reconstructed spectrum reached 0.9 and above, the color difference of 100% of samples was less than 0.5. Using fast independent component analysis for dimension reduction of the spectral space could achieve high-precision representation of the original spectral space. |
| Key words: fast independent component analysis wide-band multi-spectral space cumulative space covering ratio color difference fitting degree of the spectrum |
渝公网安备 50010702501716号