DeepIndices: Remote Sensing Indices Based on Approximation of Functions through Deep-Learning, Application to Uncalibrated Vegetation Images
The form of a remote sensing index is generally empirically defined, whether by choosing specific reflectance bands, equation forms or its coefficients. These spectral indices are used as preprocessing stage before object detection/classification. But no study seems to search for the best form through function approximation in order to optimize the classification and/or segmentation. The objective of this study is to develop a method to find the optimal index, using a statistical approach by gradient descent on different forms of generic equations. From six wavebands images, five equations have been tested, namely: linear, linear ratio, polynomial, universal function approximator and dense morphological. Few techniques in signal processing and image analysis are also deployed within a deep-learning framework. Performances of standard indices and DeepIndices were evaluated using two metrics, the dice (similar to f1-score) and the mean intersection over union (mIoU) scores. The study focuses on a specific multispectral camera used in near-field acquisition of soil and vegetation surfaces. These DeepIndices are built and compared to 89 common vegetation indices using the same vegetation dataset and metrics. As an illustration the most used index for vegetation, NDVI (Normalized Difference Vegetation Indices) offers a mIoU score of 63.98% whereas our best models gives an analytic solution to reconstruct an index with a mIoU of 82.19%. This difference is significant enough to improve the segmentation and robustness of the index from various external factors, as well as the shape of detected elements.
PDF AbstractDatasets
Task | Dataset | Model | Metric Name | Metric Value | Global Rank | Benchmark |
---|---|---|---|---|---|---|
2D Semantic Segmentation | Deep Indices | 7x7 dense-morphological + ibf + sprb | mIoU | 82.19 | # 1 | |
2D Semantic Segmentation | Deep Indices | 7x7 universal-function + ibf + sprb | mIoU | 80.36 | # 9 | |
2D Semantic Segmentation | Deep Indices | 7x7 polynomial + ibf + sprb | mIoU | 81.21 | # 5 | |
2D Semantic Segmentation | Deep Indices | 7x7 linear-ratio + ibf + sprb | mIoU | 81.35 | # 4 | |
2D Semantic Segmentation | Deep Indices | 7x7 linear + ibf+sprb | mIoU | 81.49 | # 3 | |
2D Semantic Segmentation | Deep Indices | 5x5 dense-morphological + ibf + sprb | mIoU | 81.92 | # 2 | |
2D Semantic Segmentation | Deep Indices | 5x5 universal-function + ibf + sprb | mIoU | 80.63 | # 8 | |
2D Semantic Segmentation | Deep Indices | 5x5 polynomial + ibf + sprb | mIoU | 80.67 | # 7 | |
2D Semantic Segmentation | Deep Indices | 5x5 linear-ratio + ibf + sprb | mIoU | 80.08 | # 12 | |
2D Semantic Segmentation | Deep Indices | 3x3 dense-morphological + ibf + sprb | mIoU | 80.29 | # 10 | |
2D Semantic Segmentation | Deep Indices | 3x3 universal-function + ibf + sprb | mIoU | 81.08 | # 6 | |
2D Semantic Segmentation | Deep Indices | 1x1 polynomial + ibf | mIoU | 80.03 | # 13 | |
2D Semantic Segmentation | Deep Indices | 1x1 universal-function + ibf + sprb | mIoU | 80.15 | # 11 | |
2D Semantic Segmentation | Deep Indices | 1x1 dense-morphological + ibf + sprb | mIoU | 80.00 | # 14 | |
2D Semantic Segmentation | Deep Indices | Adjusted Transformed Soil Adjusted VI | mIoU | 64.96 | # 22 | |
2D Semantic Segmentation | Deep Indices | Global Environment Monitoring Index | mIoU | 65.04 | # 21 | |
2D Semantic Segmentation | Deep Indices | Enhanced Vegetation Index 3 | mIoU | 65.05 | # 20 | |
2D Semantic Segmentation | Deep Indices | Soil And Atmospherically Resistant VI 3 | mIoU | 65.86 | # 19 | |
2D Semantic Segmentation | Deep Indices | Soil Adjusted Vegetation Index | mIoU | 67.28 | # 18 | |
2D Semantic Segmentation | Deep Indices | Enhanced Vegetation Index 2 | mIoU | 67.94 | # 17 | |
2D Semantic Segmentation | Deep Indices | Modified Chlorophyll Absorption In Reflectance Index 1 | mIoU | 73.68 | # 16 | |
2D Semantic Segmentation | Deep Indices | Modified Triangular Vegetation Index 1 | mIoU | 73.71 | # 15 | |
2D Semantic Segmentation | Deep Indices | NDVI | mIoU | 63.98 | # 23 |