A Novel Intelligent Classification Method for Urban Green Space Based on High-Resolution Remote Sensing Images

Abstract

The real-time, accurate, and refined monitoring of urban green space status information is of great significance in the construction of urban ecological environment and the improvement of urban ecological benefits. The high-resolution technology can provide abundant information of ground objects, which makes the information of urban green surface more complicated. The existing classification methods are challenging to meet the classification accuracy and automation requirements of high-resolution images. This paper proposed a deep learning classification method for urban green space based on phenological features constraints in order to make full use of the spectral and spatial information of green space provided by high-resolution remote sensing images (GaoFen-2) in different periods. The vegetation phenological features were added as auxiliary bands to the deep learning network for training and classification. We used the HRNet (High-Resolution Network) as our model and introduced the Focal Tversky Loss function to solve the sample imbalance problem. The experimental results show that the introduction of phenological features into HRNet model training can effectively improve urban green space classification accuracy by solving the problem of misclassification of evergreen and deciduous trees. The improvement rate of F1-Score of deciduous trees, evergreen trees, and grassland were 0.48%, 4.77%, and 3.93%, respectively, which proved that the combination of vegetation phenology and high-resolution remote sensing image can improve the results of deep learning urban green space classification.

Mapping of Urban Vegetation with High-Resolution Remote Sensing: A Review

“…For optical imagery, both spectral and spatial resolution are important. The mapping of functional green types or green infrastructure does not always require a very high spatial resolution since single plants are grouped together, creating larger spatial units that can be mapped from imagery with a resolution that is coarser than 3 m [36,73,74]. However, for the mapping of individual plants, the use of higher-resolution imagery is required to limit the complex background contamination that occurs in an urban environment [53].…”

“…Capturing the phenological differences between vegetation types or species through the use of multi-temporal imagery is a well-known mapping approach in global vegetation studies and in monitoring vegetation in rural environments. However, it also proves to be a promising method for improving vegetation differentiation and for monitoring vegetation species in urban settings [66,73,79]. The challenge here lies in selecting a data source (or combining different sources) offering an adequate temporal resolution to capture imagery at multiple times throughout the year while, at the same time, having a spatial and spectral resolution that is sufficiently high.…”

“…A promising way to improve vegetation mapping is by making use of multi-temporal information such that the phenological characteristics of a plant species can be taken into consideration [57,66,70,73,79,106]. This has become possible with the launch of satellites with a short revisit time in combination with an adequate spatial and spectral resolution, such as RapidEye or PlanetScope.…”

“…The strength of deep learning techniques is based on the capability of a network to attain a feature representation that maximizes the separability between the different classes. Thus far, only a limited number of studies have applied deep learning techniques for the classification of urban green [55,73,98,108,121]. Overall, various groups of deep learning algorithms can be distinguished, of which three have been used in the papers identified in this literature review for mapping urban vegetation: (a) a Boltzman machine, (b) an MLP with more than one hidden layer and (c) a convolutional neural network architecture (CNN).…”

“…Xu et al [73] used a CNN architecture (HRNet) to discriminate between basic vegetation classes (decidious trees, evergreen trees, grassland) in an urban environment using multi-temporal GaoFen-2 imagery, reaching an overall accuracy of 93%, which was 6% points higher than the performance of an SVM on the same dataset. Hartling et al [55] adopted a DenseNet architecture [117] to model eight tree species in a city park environment using WorldView-2/3 data.…”