Visual place recognition and localization are critical components of autonomous vehicle navigation and mapping, particularly in global navigation satellite system (GNSS)-denied environments. Recognizing previously visited or mapped areas is essential for ensuring the reliability of autonomous systems. Recent works on visual place recognition and localization are mainly focused on ground or close to ground applications such as self-driving cars and indoor-scenarios. Compared with autonomous cars and mobile robots, however, visual place recognition for Unmanned Aerial Vehicle (UAV) usually faces more severe challenges, including scale and viewpoint variations, lighting, weather, and seasonal changes. To tackle these challenges, we propose a novel multi-scale feature fusion network for UAV visual place recognition. This network integrates a multi-scale image feature extraction module and a global feature extraction module. Our approach aims to enhance the robustness and accuracy of visual place recognition in UAV applications, and we verify the effectiveness of our proposed network on two UAV visual place recognition datasets.