Convolutional neural networks (CNNs) have demonstrated their ability object detection of very high resolution remote sensing images. However, CNNs have obvious limitations for modeling geometric variations in remote sensing targets. In this paper, we introduced a CNN structure, namely deformable ConvNet, to address geometric modeling in object recognition. By adding offsets to the convolution layers, feature mapping of CNN can be applied to unfixed locations, enhancing CNNs' visual appearance understanding. In our work, a deformable region-based fully convolutional networks (R-FCN) was constructed by substituting the regular convolution layer with a deformable convolution layer. To efficiently use this deformable convolutional neural network (ConvNet), a training mechanism is developed in our work. We first set the pre-trained R-FCN natural image model as the default network parameters in deformable R-FCN. Then, this deformable ConvNet was fine-tuned on very high resolution (VHR) remote sensing images. To remedy the increase in lines like false region proposals, we developed aspect ratio constrained non maximum suppression (arcNMS). The precision of deformable ConvNet for detecting objects was then improved. An end-to-end approach was then developed by combining deformable R-FCN, a smart fine-tuning strategy and aspect ratio constrained NMS. The developed method was better than a state-of-the-art benchmark in object detection without data augmentation.

No abstract

Retrieval of relevant vectors produced by representation learning critically influences the efficiency in natural language processing (NLP) tasks. In this paper we demonstrate an efficient method for searching vectors via a typical nonmetric matching function: inner product. Our method, which constructs an approximate Inner Product Delaunay Graph (IPDG) for top-1 Maximum Inner Product Search (MIPS), transforms retrieving the most suitable latent vectors into a graph search problem with great benefits of efficiency. Experiments on data representations learned for different machine learning tasks verify the outperforming effectiveness and efficiency of the proposed IPDG.

Unmanned aerial vehicles (UAVs) have been widely used in urban traffic supervision in recent years. However, the detection, tracking, and geolocation of moving vehicle based on the airborne platform suffer from small object sizes, complex scenes, and low-accuracy sensors. To address these problems, this paper develops a framework for moving vehicle detecting, tracking, and geolocating based on a monocular camera, a GPS receiver, and inertial measurement units (IMUs) sensors. First, the method based on YOLOv3 was employed for vehicle detection due to its effectiveness and efficiency for small object detection in complex scenes. Then, a visual tracking method based on correlation filters is introduced, and a passive geolocation method is presented to calculate the GPS coordinates of the moving vehicle. Finally, a flight control method in terms of the previous image processing results is introduced to lead the UAV that is following the interesting moving vehicle. The proposed scheme has been built on a DJI M100 platform on which a monocular camera and a microcomputer Jetson TX1 are added. The experimental results show that this scheme is capable of detecting, tracking, and geolocating the interesting moving vehicle with high precision. The framework demonstrates its capacity in automatic supervision on target vehicles in real-world experiments, which suggests its potential applications in urban traffic, logistics, and security. INDEX TERMS Unmanned aerial vehicle, YOLOv3, object geolocation, moving vehicle tracking.

No abstract

Wireless sensor networks are proved to be effective in long-time localized torrential rain monitoring. However, the existing widely used architecture of wireless sensor networks for rain monitoring relies on network transportation and back-end calculation, which causes delay in response to heavy rain in localized areas. Our work improves the architecture by applying logistic regression and support vector machine classification to an intelligent wireless sensor node which is created by Raspberry Pi. The sensor nodes in front-end not only obtain data from sensors, but also can analyze the probabilities of upcoming heavy rain independently and give early warnings to local clients in time. When the sensor nodes send the probability to back-end server, the burdens of network transport are released. We demonstrate by simulation results that our sensor system architecture has potentiality to increase the local response to heavy rain. The monitoring capacity is also raised.

No abstract

Landslides endanger regular industrial production and human safety. Displacement trend analysis gives us an explicit way to observe and forecast landslides. Although satellite-borne remote sensing methods such as synthetic aperture radar have gradually replaced manual measurement in detecting deformation trends, they fail to observe displacement in a north-south direction. Wireless low-cost GPS sensors have been developed to assist remote sensing methods in north-south deformation monitoring because of their high temporal resolution and wide usage. In our paper, a DLM-LSTM framework is developed to extract and predict north-south land deformation trends from meter accuracy GPS receivers. A dynamic linear model is introduced to model the relation between measurement and the state vector, including the trend, periodic variation, and autoregressive factors in a discontinuous low-cost latitude time series. The deformation trend with submeter-level accuracy is extracted by a Kalman filter and smoother. With validated input as in previous work, the power of an LSTM network is also shown in its ability to predict deformation trends in submeter-level accuracy. A submeter-level deformation trend is detected from wireless low-cost GPS sensors with meter-level navigation error. The framework will have broad application prospects in geological disaster monitoring.Coal mining in the Fushun Western Open-Pit Coal Mine (FWOCM), which is located at 41°50′38.33″ north latitude and 123°53′21.77″ east longitude, started in the early 20th century. Production of coal used to be a pillar of the Fushun economy. Decades of mining has caused frequent surface deformation and landslides that threaten the city, Fushun. An open pit with 6600 m east-to-west length and 2200 m north-to-south width stands among the residential districts [6,7]. Therefore, a landslide is a serious threat to miners and local people. According to our field investigation, deformation in the north-south direction is much worse than in the east-west direction. The velocity of displacement on the north and south sides is less than 3 m/year. A fault zone