Toward a Yearly Country-Scale CORINE Land-Cover Map without Using Images: A Map Translation Approach

Abstract

CORINE Land-Cover (CLC) and its by-products are considered as a reference baseline for land-cover mapping over Europe and subsequent applications. CLC is currently tediously produced each six years from both the visual interpretation and the automatic analysis of a large amount of remote sensing images. Observing that various European countries regularly produce in parallel their own land-cover country-scaled maps with their own specifications, we propose to directly infer CORINE Land-Cover from an existing map, therefore steadily decreasing the updating time-frame. No additional remote sensing image is required. In this paper, we focus more specifically on translating a country-scale remote sensed map, OSO (France), into CORINE Land Cover, in a supervised way. OSO and CLC not only differ in nomenclature but also in spatial resolution. We jointly harmonize both dimensions using a contextual and asymmetrical Convolution Neural Network with positional encoding. We show for various use cases that our method achieves a superior performance than the traditional semantic-based translation approach, achieving an 81% accuracy over all of France, close to the targeted 85% accuracy of CLC.

A Scalable High-Performance Unsupervised System for Producing Large-Scale HR Land Cover Maps: The Italian Country Case Study

“…As an example, in [12] Sentinel-1 images acquired in 2014 were resampled to 100 m resolution and used to reproduce the CLC mapping. To increase the temporal resolution of the CLC map, in [13], Baudoux et al propose a method to translate an annually updated french national LC product into the CLC map. The method was able to achieve an accuracy of 81% for the entire French Country.…”

“…Although the availability of Sentinel-1 and Sentinel-2 data has significantly increased the possibility of frequently ac-quiring EO data, producing LC maps is expensive and timeconsuming, especially when performed at country, continental or global level [13]. In particular, the most challenging part is the collection of constantly updated annotated samples, which is extremely demanding on a large-scale through ground surveys or photo-interpretation [26].…”

Deep-Learning Based Multiple Land-Cover Map Translation

“…The chosen labels and resolutions highly impact the land-cover map potential usage and subsequent manipulations such as map fusion, harmonisation, comparison or update. The challenge in map-to-map translation lies in the difficult interleaved association of semantic and spatial resolutions of both maps [1]. Usually, two different LULC maps establish complex relationships between their classes, and direct one-to-one association is most of the time infeasible.…”

Improving the Accuracy of Land Cover Mapping by Distributing Training Samples

“…High-quality land cover maps are the basis for monitoring the status and dynamics of the earth's surface and one of the crucial parameters to understand the processes of a region [1,2]. They have been widely used in land resource management [3], disaster monitoring [4], and environmental assessment [5].…”

“…In this paper, we aim at developing a training sample distribution method to improve the representativeness and diversity of samples. Two specific objectives include (1) proposing an object-oriented sampling approach by segmenting image blocks expanded from systematically distributed seeds, and (2) in terms of classification accuracy and sample diversity, quantitatively comparing the proposed method with traditional probability sampling, stratified sampling, and manual sampling.…”