A Soft Computing Approach for Selecting and Combining Spectral Bands

Albarracín, Juan Felipe Hernández; Oliveira, Rafael S.; Hirota, Marina; Santos, Jefersson A. dos; Torres, Ricardo da S.

doi:10.3390/rs12142267

Cited by 5 publications

(3 citation statements)

References 56 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The scores obtained by this function are used to sort the vertices aiming to minimize the number of vertices explored to discover the true nearest neighbors. GP has been shown to perform well in optimization scenarios like this [2,10,30,43]. In the context of the GP approach, an individual from the population corresponds to a candidate scoring function.…”

Section: Introductionmentioning

confidence: 99%

A genetic programming approach for searching on nearest neighbors graphs

Muñoz

Dias

Torres

2022

Multimed Tools Appl

Self Cite

View full text Add to dashboard Cite

Nearest neighbors graphs have gained a lot of attention from the information retrieval community since they were demonstrated to outperform classical approaches in the task of approximate nearest neighbor search. These approaches, firstly, index feature vectors by using a graph-based data structure. Then, for a given query, the search is performed by traversing the graph in a greedy-way, moving in each step towards the neighbor of the current vertex that is closer to the query (based on a distance function). However, local topological properties of vertices could be also considered at the moment of deciding the next vertex to be explored. In this work, we introduce a Genetic Programming framework that combines topological properties along with the distance to the query, aiming to improve the selection of the next vertex in each step of graph traversal and, therefore, reduce the number of vertices explored (scan rate) to find the true nearest neighbors. Experimental results, conducted over three large collections of feature vectors and four different graph-based techniques, show significant gains of the proposed approach over classic graph-based search algorithms.

show abstract

Section: Introductionmentioning

confidence: 99%

A genetic programming approach for searching on nearest neighbors graphs

Muñoz

Dias

Torres

2022

Multimed Tools Appl

Self Cite

View full text Add to dashboard Cite

show abstract

“…UAVs have shown an immense potential to close the gap between field observations and satellite imagery [19], since its low elevation flights produce very high-resolution images (0.02-0.03 m) and can cover areas of several tens to hundreds of hectares. The very high-resolution data have renewed interest in the use of a wide variety of data processing techniques, which include computer vision, soft computing, and deep learning [10,14,[20][21][22] among others. Furthermore, pixel counting in satellite images to estimate classes in a given area [23] can be used at a micro level for counting pixels to estimate defoliation in a given tree canopy area from very-high resolution images.…”

Section: Introductionmentioning

confidence: 99%

Classifying the Degree of Bark Beetle-Induced Damage on Fir (Abies mariesii) Forests, from UAV-Acquired RGB Images

et al. 2022

View full text Add to dashboard Cite

Bark beetle outbreaks are responsible for the loss of large areas of forests and in recent years they appear to be increasing in frequency and magnitude as a result of climate change. The aim of this study is to develop a new standardized methodology for the automatic detection of the degree of damage on single fir trees caused by bark beetle attacks using a simple GIS-based model. The classification approach is based on the degree of tree canopy defoliation observed (white pixels) in the UAV-acquired very high resolution RGB orthophotos. We defined six degrees (categories) of damage (healthy, four infested levels and dead) based on the ratio of white pixel to the total number of pixels of a given tree canopy. Category 1: <2.5% (no defoliation); Category 2: 2.5–10% (very low defoliation); Category 3: 10–25% (low defoliation); Category 4: 25–50% (medium defoliation); Category 5: 50–75% (high defoliation), and finally Category 6: >75% (dead). The definition of “white pixel” is crucial, since light conditions during image acquisition drastically affect pixel values. Thus, whiteness was defined as the ratio of red pixel value to the blue pixel value of every single pixel in relation to the ratio of the mean red and mean blue value of the whole orthomosaic. The results show that in an area of 4 ha, out of the 1376 trees, 277 were healthy, 948 were infested (Cat 2, 628; Cat 3, 244; Cat 4, 64; Cat 5, 12), and 151 were dead (Cat 6). The validation led to an average precision of 62%, with Cat 1 and Cat 6 reaching a precision of 73% and 94%, respectively.

show abstract

“…Another study has proposed to optimize the weights in an NDVI equation form based on a genetic algorithm [25] but does not optimize the equation forms. An other approach has been proposed to automatically construct a vegetation index using a genetic algorithm [26]. They optimize the equation forms by building a set of arithmetic graphs with mutations, crossovers and replications to change the shape of each equation during learning but it does not take into account the weights, since it's use calibrated data.…”

Section: Introductionmentioning

confidence: 99%

DeepIndices: Remote Sensing Indices Based on Approximation of Functions through Deep-Learning, Application to Uncalibrated Vegetation Images

et al. 2021

View full text Add to dashboard Cite

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.

show abstract

A Soft Computing Approach for Selecting and Combining Spectral Bands

Cited by 5 publications

References 56 publications

A genetic programming approach for searching on nearest neighbors graphs

A genetic programming approach for searching on nearest neighbors graphs

Classifying the Degree of Bark Beetle-Induced Damage on Fir (Abies mariesii) Forests, from UAV-Acquired RGB Images

DeepIndices: Remote Sensing Indices Based on Approximation of Functions through Deep-Learning, Application to Uncalibrated Vegetation Images

Contact Info

Product

Resources

About