In Japan, many urban park trees are becoming large and aged, increasing the risk of structural failures caused by extreme weather events and biological deterioration. Effective management therefore requires reliable risk assessment, for which accurate tree species identification is essential. However, species identification still depends heavily on visual assessment by skilled professionals, creating challenges in efficiency and objectivity. This issue is particularly significant for broad-leaved trees because of their high species diversity and morphological variability. Labor shortages have also increased the demand for automated classification techniques. This study proposes a tree species classification method for broad-leaved trees using ground-level images captured with an RGB camera and deep learning. The method extracts small local patches containing species-specific visual features, such as leaf shape and bark texture, commonly used by professional arborists. These local features are independently evaluated using deep learning models, enabling effective use of fine-scale visual characteristics. To improve robustness against outdoor imaging variations, including illumination changes, shadows, and moss attachment, multiple patches are classified independently and integrated through majority voting. Experiments were conducted on seven tree species commonly found in Japanese urban parks: cherry, ginkgo, zelkova, konara oak, sawtooth oak, plane tree, and flowering dogwood. Results showed a maximum classification accuracy of approximately 95% under real-world conditions, demonstrating the effectiveness of the proposed method for practical urban tree management.