Classification of Different Vegetation Types Combining Two Information Sources Through a Probabilistic Segmentation Approach

Oliva, Francisco E.; Dalmau, Oscar; Alarcón, Teresa E.; De-la-Torre, Miguel

doi:10.1007/978-3-319-27101-9_29

Cited by 2 publications

(20 citation statements)

References 2 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We note that the results of the first row of Table 6 and Table 7 differ from those presented in Ref. [28,29]. This is due to, in this results, we only consider sites in the image that correspond to crops in the five categories of interest, i.e., only pixels in the region of interest.…”

Section: Experiments and Discussionmentioning

confidence: 82%

“…In this case, we use the matlab built-in function for SVM. In the comparison study we also consider the original version of the probabilistic segmentation approach described in [28,29]. Additionally, for a fair comparison, we include the performance analysis of these algorithms using the best results reached in the feature space study in this work, see Table 6 and Table 7, denoted as MICAI

2014^{*}

and MICAI

2015^{*}

in Table 9.…”

Section: Experiments and Discussionmentioning

confidence: 99%

“…From these two feature spaces they constructed histograms, and from these histograms they estimated the corresponding likelihood. Different from [29], authors in [28] introduced a modification in data term of the GMMF approach in order to select at each pixel r the likelihood with the least entropy:

\begin{matrix} p^{*} & = & arg \underset{bold-italicp}{trueprefixmin} \sum_{r \in L} \sum_{k \in K} \sum_{i = 1}^{2} ω_{i} (r) {(p_{k} (r) - v_{k}^{i} (r))}^{2} + λ \sum_{s \in {scriptN}_{r}} {(p_{k} (r) - p_{k} (s))}^{2}, \end{matrix}

where

v^{i} (r)

is the likelihood that comes from the i -th source and the weight function

ω_{i} (r) \in {0, 1}

is given by Equation (5)

\begin{matrix} ω_{i} (r) & = & ({, \begin{matrix} 1 & if E (v^{i} (r)) < E (v^{3 - i} (r)) \\ 0 & otherwise, \end{matrix}) \end{matrix}

i \in {1, 2}

and

scriptE (\cdot)

is an entropy measure [38]. The authors in [28] used Gini impurity index, i.e.,

\begin{matrix} E (f) & = & 1 - {bold-italicf}^{T} \end{matrix}

…”

Section: Previous Workmentioning

confidence: 99%

“…The previous idea [28] contributed to diminish the misclassification in the recognition of different crops. The inclusion of a new feature space improved the classification process with respect to the results obtained in [29].…”

Section: Previous Workmentioning

confidence: 99%

“…Equations (10)–(12) indicate the modifications.

\begin{matrix} p^{*} & = & arg \underset{bold-italicp}{trueprefixmin} \sum_{r \in L} \sum_{k \in K} \sum_{f \in F} ω_{f} (r) {(p_{k} (r) - v_{k}^{f} (r))}^{2} + λ \sum_{s \in {scriptN}_{r}} ω_{r s} {(p_{k} (r) - p_{k} (s))}^{2}, \end{matrix}

ω_{f} (r) = \frac{ω_{f}^{s} (r)}{\sum_{f \in scriptF} ω_{f}^{s} (r)}

ω_{f}^{s} (r) = \frac{μ}{μ + U_{f} (r)}

where

U_{f} (r) \in [0, 1]

is an uncertainty measure of the information source f at pixel r , for example, the measure of entropy (5) used in [28];

λ > 0

is the regularization parameter,

μ > 0

controls the relative importance of the likelihood for different sources. When

μ

is very large, the contribution of the likelihood for all sources tends to be the same and when it is close to zero, the functional in (10) tends to select the likelihood corresponding to the lowest uncertainty, i.e., this tends to the solution proposed in [28] when using the entropy as uncertainty measure. The weight

ω_{r s} \in [0, 1]

function allows to control the edges between classes, here we use…”

Section: The Proposalmentioning

confidence: 99%

See 4 more Smart Citations

Crop Classification in Satellite Images through Probabilistic Segmentation Based on Multiple Sources

Dalmau

Alarcón

Oliva

2017

Sensors

View full text Add to dashboard Cite

Classification methods based on Gaussian Markov Measure Field Models and other probabilistic approaches have to face the problem of construction of the likelihood. Typically, in these methods, the likelihood is computed from 1D or 3D histograms. However, when the number of information sources grows, as in the case of satellite images, the histogram construction becomes more difficult due to the high dimensionality of the feature space. In this work, we propose a generalization of Gaussian Markov Measure Field Models and provide a probabilistic segmentation scheme, which fuses multiple information sources for image segmentation. In particular, we apply the general model to classify types of crops in satellite images. The proposed method allows us to combine several feature spaces. For this purpose, the method requires prior information for building a 3D histogram for each considered feature space. Based on previous histograms, we can compute the likelihood of each site of the image to belong to a class. The computed likelihoods are the main input of the proposed algorithm and are combined in the proposed model using a contrast criteria. Different feature spaces are analyzed, among them are 6 spectral bands from LANDSAT 5 TM, 3 principal components from PCA on 6 spectral bands and 3 principal components from PCA applied on 10 vegetation indices. The proposed algorithm was applied to a real image and obtained excellent results in comparison to different classification algorithms used in crop classification.

show abstract

Section: Experiments and Discussionmentioning

confidence: 82%