TY - JOUR

T1 - Investigations on feature similarity and the impact of training data for land cover classification

AU - Voelsen, M.

AU - Torres, D. Lobo

AU - Feitosa, R. Q.

AU - Rottensteiner, F.

AU - Heipke, C.

PY - 2021/6/17

Y1 - 2021/6/17

N2 - Fully convolutional neural networks (FCN) are successfully used for pixel-wise land cover classification - the task of identifying the physical material of the Earth's surface for every pixel in an image. The acquisition of large training datasets is challenging, especially in remote sensing, but necessary for a FCN to perform well. One way to circumvent manual labelling is the usage of existing databases, which usually contain a certain amount of label noise when combined with another data source. As a first part of this work, we investigate the impact of training data on a FCN. We experiment with different amounts of training data, varying w.r.t. the covered area, the available acquisition dates and the amount of label noise. We conclude that the more data is used for training, the better is the generalization performance of the model, and the FCN is able to mitigate the effect of label noise to a high degree. Another challenge is the imbalanced class distribution in most real-world datasets, which can cause the classifier to focus on the majority classes, leading to poor classification performance for minority classes. To tackle this problem, in this paper, we use the cosine similarity loss to force feature vectors of the same class to be close to each other in feature space. Our experiments show that the cosine loss helps to obtain more similar feature vectors, but the similarity of the cluster centers also increases.

AB - Fully convolutional neural networks (FCN) are successfully used for pixel-wise land cover classification - the task of identifying the physical material of the Earth's surface for every pixel in an image. The acquisition of large training datasets is challenging, especially in remote sensing, but necessary for a FCN to perform well. One way to circumvent manual labelling is the usage of existing databases, which usually contain a certain amount of label noise when combined with another data source. As a first part of this work, we investigate the impact of training data on a FCN. We experiment with different amounts of training data, varying w.r.t. the covered area, the available acquisition dates and the amount of label noise. We conclude that the more data is used for training, the better is the generalization performance of the model, and the FCN is able to mitigate the effect of label noise to a high degree. Another challenge is the imbalanced class distribution in most real-world datasets, which can cause the classifier to focus on the majority classes, leading to poor classification performance for minority classes. To tackle this problem, in this paper, we use the cosine similarity loss to force feature vectors of the same class to be close to each other in feature space. Our experiments show that the cosine loss helps to obtain more similar feature vectors, but the similarity of the cluster centers also increases.

KW - Cosine similarity loss

KW - FCN

KW - Land cover classification

KW - Remote sensing

UR - http://www.scopus.com/inward/record.url?scp=85113134867&partnerID=8YFLogxK

U2 - 10.5194/isprs-annals-V-3-2021-181-2021

DO - 10.5194/isprs-annals-V-3-2021-181-2021

M3 - Conference article

AN - SCOPUS:85113134867

VL - 5

SP - 181

EP - 189

JO - ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences

JF - ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences

SN - 2194-9042

IS - 3

T2 - 24th ISPRS Congress on Imaging today, foreseeing tomorrow, Commission III

Y2 - 5 July 2021 through 9 July 2021

ER -