融合深度学习与无人机图像采集的野生鸟类细粒度识别与监测方法研究

Fine-Grained Recognition and Monitoring of Wild Birds Based on Deep Learning and UAV Image Acquisition

摘要: 野生鸟类的细粒度识别与监测对于生态研究和保护工作具有重要意义，但传统的识别方法在复杂自然环境中面临着泛化能力有限等挑战。为提高野生鸟类细粒度识别的准确性和实时性，研究结合无人机图像采集技术构建并优化了一种基于改进卷积神经网络(Convolutional Neural Network, CNN)的识别模型，通过引入残差连接和注意力机制，结合数据增强和正则化技术，显著提升了模型的性能。实验结果中，改进CNN模型在Birdsnap数据集上的物种识别时间为15毫秒，准确率为92.3%；在CUB-200-2011数据集上的物种识别时间为14毫秒，准确率为94.5%。结果表明，改进CNN模型在识别速度和准确性方面均优于对照模型，其具有更强的鲁棒性和更高的物种识别多样性。研究结果不仅为野生鸟类的细粒度识别提供了新的技术路径，也为生态监测和保护工作提供了更高效、更准确的技术支持，有助于推动鸟类生态学研究和保护工作的深入开展。

Abstract: Fine-grained identification and monitoring of wild birds are of great significance for ecological research and protection. However, traditional identification methods face challenges such as limited generalization ability in complex natural environments. To improve the accuracy and real-time performance of fine-grained recognition of wild birds, a recognition model based on the improved Convolutional Neural Network (CNN) was constructed and optimized by combining unmanned aerial vehicle (UAV) image acquisition technology. By introducing residual connection and attention mechanism, combined with data augmentation and regularization techniques, the performance of the model has been significantly improved. In the experimental results, the species recognition time of the improved CNN model on the Birdsnap dataset was 15 milliseconds, and the accuracy rate was 92.3%. The species identification time on the CUB-200-2011 dataset was 14 milliseconds and the accuracy rate was 94.5%. The results show that the improved CNN model is superior to the control model in both recognition speed and accuracy, and it has stronger robustness and higher species recognition diversity. The research results not only provide a new technical path for fine-grained identification of wild birds, but also offer more efficient and accurate technical support for ecological monitoring and protection work, which is conducive to promoting the in-depth development of bird ecology research and protection work.