人工智能培训

搜索

人工智能论文:学习定位 - 估计卷积神经网络在光学遥感图像中的建立检测(Learning Orientation-Estimation Convolutiona

[复制链接]
~Limi多多~ 发表于 2019-3-15 12:32:16 | 显示全部楼层 |阅读模式
~Limi多多~ 2019-3-15 12:32:16 491 0 显示全部楼层
人工智能论文:学习定位 - 估计卷积神经网络在光学遥感图像中的建立检测(Learning Orientation-Estimation Convolutional Neural Network for  Building Detection in Optical Remote Sensing Image)受益于计算机视觉深度学习的巨大成功,基于CNN的物体检测方法引起了人们的极大关注。已经提出了各种框架,其对于大范围的数据集显示出令人敬畏且稳健的性能。然而,对于远程传感图像中的建筑物检测,建筑物总是呈现出多样化的方向,这使其成为应用现成方法进行建筑物检测的挑战。在这项工作中,我们的目标是将方向回归整合到人口对齐的边界中。 -box检测方法来解决这个问题。为了使轴对齐的边界框适应任意定向的边界框,我们还开发了算法来估计任意两个任意方向的框之间的交叉点(IoU)重叠,这便于在图形处理单元(GPU)中实现以加速计算。建议将CNN用于稳健特征提取和旋转边界框回归。我们以端到端的方式展示我们的模型,使其轻松实现。该模型被制定和训练以预测方向,位置和范围,同时获得更紧密的边界框,因此,更高的平均平均精度(mAP)。对不同尺度的遥感图像的实验显示出比传统的遥感图像有希望的性能。
Benefiting from the great success of deep learning in computer vision,CNN-based object detection methods have drawn significant attentions.Variousframeworks have been proposed which show awesome and robust performance for alarge range of datasets.However, for building detection in remote sensingimages, buildings always pose a diversity of orientations which makes it achallenge for the application of off-the-shelf methods to building detection.In this work, we aim to integrate orientation regression into the popularaxis-aligned bounding-box detection method to tackle this problem.To adapt theaxis-aligned bounding boxes to arbitrarily orientated ones, we also develop analgorithm to estimate the Intersection over Union (IoU) overlap between any twoarbitrarily oriented boxes which is convenient to implement in GraphicsProcessing Unit (GPU) for accelerating computation.The proposed methodutilizes CNN for both robust feature extraction and rotated bounding boxregression.We present our modelin an end-to-end fashion making it easy totrain.The model is formulated and trained to predict orientation, location andextent simultaneously obtaining tighter bounding box and hence, higher meanaverage precision (mAP).Experiments on remote sensing images of differentscales shows a promising performance over the conventional one.人工智能论文:学习定位 - 估计卷积神经网络在光学遥感图像中的建立检测(Learning Orientation-Estimation Convolutional Neural Network for  Building Detection in Optical Remote Sensing Image) ce8Bitzz0z7N918e.jpg
URL地址:https://arxiv.org/abs/1903.05862     ----pdf下载地址:https://arxiv.org/pdf/1903.05862    ----人工智能论文:学习定位 - 估计卷积神经网络在光学遥感图像中的建立检测(Learning Orientation-Estimation Convolutional Neural Network for  Building Detection in Optical Remote Sensing Image)
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则 返回列表 发新帖

~Limi多多~当前离线
新手上路

查看:491 | 回复:0

快速回复 返回顶部 返回列表