Urban green spaces constitute a vital component of the ecosystem. This study focused on urban green spaces located within the Fifth Ring Road of Beijing, using Gaofen 7 (GF-7) as the primary data source for analysis. The main objective was to develop a system for extracting and classifying urban green spaces in Beijing by applying deep learning and machine learning algorithms, and further, the results were validated with ground survey samples. This study provides detailed extraction and classification of urban green space coverage by creating a comprehensive evaluation system. The primary findings indicate that the deep learning algorithm enhances the precision of green space information extraction by 10.68% compared to conventional machine learning techniques, effectively suppresses “pretzel noise”, and eventually aids in extracting green space information with complete edges. The thorough assessment of green spaces within the study area indicated favorable outcomes showing the high service capacity of park green spaces. The overall classification accuracy of the final extraction results was 94.31%. Nonetheless, challenges, such as unequal distribution of green zones and a significant fragmentation level throughout the study area, were still encountered. Consequently, the use of GF-7 high-resolution imagery, in conjunction with the collaborative application of deep learning and machine learning techniques, enabled the acquisition of highly accurate information regarding urban green zone coverage. According to the established grading standards of evaluation indices, the landscape pattern of urban green spaces within the study area was comprehensively assessed. This evaluation offers essential data support for monitoring urban green spaces and planning landscape patterns, thereby contributing to the achievement of sustainable development objectives related to urban greening and ecological conservation.
Real-time, accurate and detailed monitoring of urban green space is of great significance for constructing the urban ecological environment and maximizing ecological benefits. Although high-resolution remote sensing technology provides rich ground object information, it also makes the surface information of urban green spaces more complex. Existing classification methods often struggle to meet the requirements of classification accuracy and the automation demands of high-resolution images. This study utilized GF-7 remote sensing imagery to construct an urban green space classification method for Beijing. The study used the YOLO v8 model as the framework to conduct a fine classification of urban green spaces within the Fifth Ring Road of Beijing, distinguishing between evergreen trees, deciduous trees, shrubs and grasslands. The aims were to address the limitations of insufficient model fit and coarse-grained classifications in existing studies, and to improve vegetation extraction accuracy for green spaces in northern temperate cities (with Beijing as a typical example). The results show that the overall classification accuracy of the trained YOLO v8 model is 89.60%, which is 25.3% and 28.8% higher than that of traditional machine learning methods such as Maximum Likelihood and Support Vector Machine, respectively. The model achieved extraction accuracies of 92.92%, 93.40%, 87.67%, and 93.34% for evergreen trees, deciduous trees, shrubs, and grasslands, respectively. This result confirms that the combination of deep learning and high-resolution remote sensing images can effectively enhance the classification extraction of urban green space vegetation, providing technical support and data guarantees for the refined management of green spaces and “garden cities” in megacities such as Beijing.
Urban green spaces critically govern urban carbon cycling, though fragmentation and vegetation differences hinder precise carbon stock calculation. Targeting Beijing’s Fifth Ring Road area (666.5 km2), this study combined GF-7 sub-meter remote sensing imagery and field data from 750 sampling plots to build four machine learning models (RF, XGBoost, GBDT, BP neural network), adopting SHAP analysis to interpret feature importance and formulate carbon improvement schemes. XGBoost with the Boruta-selected feature set achieved robust performance (R2 = 0.83, RMSE = 1.79 kg/m2, MAE = 1.13 kg/m2), with its R2 elevated by 6.4%--38.7% compared to other algorithms. Canopy height dominated carbon storage variation (SHAP contribution: 0.321, 31.7% explanatory rate), and removing it cut model R2 by 47.98%. The region’s total above-ground vegetation carbon stock reached 9.04 × 105 Mg at a mean above-ground carbon density of 4.80 kg/m2; deciduous trees contributed the largest share (80.61%) of total above-ground carbon stock) and presented a low-inner, high-outer ring spatial distribution. Raising canopy height by 1 m could increase above-ground carbon storage by 21.38% within tree-covered areas, equivalent to a 4.10% increase in the total study-area carbon stock. The combined implementation of canopy height enhancement, a 10% replacement of shrub/grassland areas, and low-carbon sink area renovation could increase the total carbon storage by 8.82 × 104 Mg (with the low-carbon zone renovation alone contributing 1.76 × 104 Mg), representing a 9.76% enhancement potential over the current above-ground baseline.