鸟类全景观测系统中的人工智能识别技术

Artificial Intelligence Recognition in Bird Panoramic Observation System

摘要: 随着数字化基础设施建设的推进，采用人工智能技术对鸟类观测影像进行实时识别，以期为湿地监控与生物多样性保护提供新的手段。2020 年，在长兴仙山湖国家湿地公园，根据鸟类监测、识别场景的特点，利用级联分类器和卷积神经网络人工智能识别技术对鸟类进行监测、识别。结果表明，该鸟类检测和分类模型在验证集中分类精度达到87.75%以上。该分类检测方法适用于广角、远景监测下的鸟类观测识别，可以在实际的鸟类观测中达到自动化监测目标鸟类、人工智能辅助发现未知鸟类的效果，未来随着数据的积累，可以持续优化模型并提高检测精度。

Abstract: In 2020, according to the characteristics of bird monitoring and recognition scenario, cascade classifier and convolution neural network artificial intelligence recognition technology was used to monitor and recognize birds in Changxing Xianshanhu National Wetland Park, Zhejiang province. The results showed that the accuracy of the model was over 87.75% in the validation set.