After an attack the by pine wood nematode, pine tree needles turn red. Using convolutional neural networks (CNNs) based object detection methods, machines can detect red-attacked trees. However, most deep learning object detection algorithms (such as Faster R-CNN and YOLO among others) often require a large number of labelled training datasets, where in each image every object must be given a bounding box label. To increase the cost-effectiveness of this process, we propose a simple yet efficient weakly supervised processing pipeline, based on class activation maps to locate the target. Unlike object detection methods that require bounding-box-labelled data for training, the proposed pipeline only needs image-level-labelled data. Using the proposed pipeline, we could achieve an average precision (AP) of 91.82% on test dataset. Comparing with sliding window-based method which achieves an average precision (AP) of 89.95%, our method not only gets a better AP but also runs faster than sliding window-based pipeline. This result not only indicates that the pipeline is a highly effective one but also demonstrates that image-level-labelled aerial images can be used for the detection of red-attacked tree. The proposed method should also find use in other object detection applications in the field of remote sensing.

松木线虫侵袭后，松树针变成红色。使用基于卷积神经网络（CNN）的对象检测方法，机器可以检测红色攻击的树木。但是，大多数深度学习对象检测算法（例如Faster R-CNN和YOLO等）通常需要大量带标签的训练数据集，其中在每个图像中，每个对象都必须被赋予边界框标签。为了提高此过程的成本效益，我们基于类激活图来定位目标，提出了一个简单而有效的弱监督处理管道。与需要边界框标记的数据进行训练的对象检测方法不同，建议的管道仅需要图像级标记的数据。使用建议的管道，我们可以在测试数据集上达到91.82％的平均精度（AP）。与可达到89.95％的平均精度（AP）的基于滑动窗口的方法相比，我们的方法不仅获得了更好的AP，而且比基于滑动窗口的管道运行速度更快。这一结果不仅表明该管道是一种高效的管道，而且还证明了图像级标记的航拍图像可用于检测红色袭击的树木。所提出的方法还应该在遥感领域的其他物体检测应用中找到用途。