Search

The automatic extraction of small objects such as roadside milestones, small traffic signs, and other urban furniture remains a technical challenge. This study focuses on methods of deep learning to detect small urban elements in mobile mapping system (MMS) images. Based on images obtained by an MMS in urban areas, we create an urban element detection (UED) data set containing several kinds of small objects found in a city. A simple feature extraction convolution neural network (CNN) called SlimNet is proposed and combined with an optimized faster R-CNN framework. The resulting deep learning method can automatically extract small objects commonly found in cities, including manhole covers, milestones, and license plates. Experiments on the UED data set show that SlimNet has the highest accuracy compared with other popular networks, including VGG, MobileNet, ResNet, and YOLOv3. The SlimNet model can achieve a mean average precision (AP) that is up to 12.3% higher than that of the lowest ResNet-152 network and can accelerate both training and detection owing to its relative simplicity. Moreover, $k$ -means clustering is used to choose the dimensions of the anchor box for detection. We ran $k$ -means clustering for different numbers of clusters, and the results show that at least four clusters are needed for detection using a small data set such as the UED. We also propose a method to use templates of different scales for anchors to further improve small object detection; this approach improved the AP by 3%–4% in our experiments. [ABSTRACT FROM AUTHOR]

The pointing bias of the laser altimeter would change because of the launch vibration, variations in the space environment or other factors, which is one of the most important factors affecting the geometric accuracy of a laser altimeter. To calibrate pointing bias and improve the measuring accuracy of spaceborne laser altimeter, this paper proposed an in-orbit calibration method based on terrain matching with pyramid-search for this spaceborne laser altimeter. First, we used published digital terrain data as a reference to match with stripes of a point cloud obtained by the geolocation model of the laser altimeter, and then the optimal matching terrain could be determined by a pyramid search. Finally, the pointing bias of the laser altimeter was calibrated. The proposed method was applied in calibration experiment of ZY3-02 laser altimeter. Several tracks of ZY3-02 laser data and advanced Land Observation Satellite Digital Surface Model (AW3D30) with a 30-m grid size were deployed for calibration and validation. After calibration, the systematic error was effectively eliminated. The elevation accuracy of the laser altimeter was improved from more than 100 to 3 m approximately, and the algorithm efficiency with pyramid search was improved by 10 times at least. The experimental result demonstrates that the proposed method is an effective means to calibrate the current spaceborne laser altimeters. [ABSTRACT FROM AUTHOR]