BIRD DETECTION NEAR WIND TURBINES FROM HIGH-RESOLUTION VIDEO USING LSTM NETWORKS-猫眼课题宝

DOI：

发表时间：

2016

期刊：

影响因子：

通讯作者：

中科院分区：

其他

文献类型：

作者：

T. Trinh;Ryota Yoshihashi;Rei Kawakami;M. Iida;T. Naemura

风力涡轮机已成为导致野生鸟类死亡的重大风险之一。为了评估这种生态影响，一种可以自动检测鸟类的系统引起了业界越来越多的关注。我们提出了一种结合卷积神经网络（CNN）和长短期记忆网络（LSTM）的鸟类检测方法，以利用从CNN中提取的丰富特征以及在后续时间帧中记忆鸟类连续外观变化的能力。使用风力涡轮机周围捕获的高分辨率视频进行的实验表明，只要正确跟踪鸟类，LSTM与CNN相结合的识别鸟类的性能优于单独使用CNN。

Wind turbines have become one of significant risks causing mortality of wild birds. In order to evaluate this ecological impact, a system that can automatically detect birds draws increased attention from the industry. We propose a bird detection method combining Convolutional Neural Networks (CNNs) and Long Short-term Memory Networks (LSTMs) to leverage rich features extracted from CNNs and the ability of memorizing continuous appearance change of birds in subsequent time frames. Experiments using highresolution videos captured around wind turbines show that LSTMs combined with CNNs outperform solely using CNNs for recognizing birds, as long as birds are correctly tracked.

BIRD DETECTION NEAR WIND TURBINES FROM HIGH-RESOLUTION VIDEO USING LSTM NETWORKS

文献摘要

相似文献