基于ICESat-GLAS数据和模糊模式识别算法识别森林类型

【目的】利用冰、云和陆地高程卫星-地球科学激光测高系统(ICESat-GLAS)回波波形数据,通过模糊模式识别算法,提出波形特征参数组合,对森林类型进行识别研究,以期提高森林类型分类精度。【方法】利用不同森林类型冠层在GLAS回波波形上表现出的差异性,提取波形特征参数 R_{\mathrm{fi}{\mathrm{t}}_1} 、K₁'和 \overline{K_1} ;将本研究提取的波形特征参数与其他波形特征参数相结合,建立波形特征参数组合;对样本数据波形特征参数进行指标归一化和奇异点检测处理,剔除样本数据中的奇异点样本;结合模糊模式识别算法,计算不同森林类型分类精度。【结果】针叶林和阔叶林森林类型分类总精度为96.30%,其中,针叶林和阔叶林森林类型分类精度分别为92.86%和97.50%;针叶林、阔叶林和混交林森林类型分类总精度为84.51%,其中,针叶林、阔叶林和混交林森林类型分类精度分别为85.71%、97.50%和52.94%。【结论】模糊模式识别算法在森林类型分类方面具有一定优势,尤其在针叶林和阔叶林森林类型识别方面,识别精度较高。

Identification of forest types based on ICESat-GLAS data and fuzzy pattern recognition algorithm

【Objective】 Using space-borne LiDAR waveform data and the fuzzy pattern recognition algorithm, combined with the waveform characteristic parameters proposed in this paper, the forest type was identified and studied to improve the accuracy of forest type classification. 【Method】 First, using the differences in GLAS-received waveforms of the canopies of different forest types, the waveform characteristic parameters R_{\mathrm{fi}{\mathrm{t}}_1} , K_1 \text{'} and \overline{K_1} were acquired. Second, the extracted waveform feature parameters were combined with the other waveform feature parameters to establish a combination of waveform feature parameters. Thereafter, the waveform feature parameters of the sample data were subjected to index normalization and singularity detection processing to unify the dimensions of different waveform feature parameters and remove the sample data. Finally, combined with the fuzzy pattern recognition algorithm, the classification accuracy of different forest types was calculated. 【Result】 The total classification accuracies of coniferous and broadleaved forest types were 96.30%, of which the classification accuracies of coniferous and broadleaved forest types were 92.86% and 97.50%, respectively. Furthermore, the total accuracy was 84.51%, of which the classification accuracies of coniferous forest, broadleaved forest, and mixed forest types were 85.71%, 97.50% and 52.94%, respectively. 【Conclusion】 The fuzzy pattern recognition algorithm has certain advantages for the forest-type classification. This is especially so in the classification of coniferous and broadleaved forests, where the classification accuracy was higher.