Jinwen Tian

Underwater Object Detection Based on Gravity Gradient

A novel method of underwater object detection based on gravity gradient is presented, which can be used on autonomous underwater vehicles (AUVs) to detect abnormal objects underwater. Gravity gradient anomalies of partial area, which are caused by the object, can be measured by a gravity gradiometer on an AUV. Then, anomalies can be inversed with a gravity gradient inversion algorithm, so the mass and barycenter of an object can be estimated. Simulation results show that approximate information of an object can be provided by the proposed method.

Flower algorithm for star pattern recognition in space surveillance with star trackers

Using star tracker to perform space surveillance is a focal point of research in aerospace engineering. However, autonomous attitude determination with star trackers in missions is a challenging task, because of spacecraft attitude dynamics and false stars. We present a novel star pattern recognition algorithm to resolve these problems. The algorithm defines a star pattern, called a flower code, composed of angular distances and circular angles. Then, a three-step strategy is adopted to find the correspondence of the sensor pattern and the catalog pattern, including initial lookup table match, cyclic dynamic match, and validation. A number of experiments are carried out on simulated and real star images. The simulation results show that the proposed method provides improved performance, especially on robustness against false stars. Also, the results for real star images demonstrate the reliability of the method for ground-based measurements.

Automated gravity gradient tensor inversion for underwater object detection

Underwater abnormal object detection is a current need for the navigation security of autonomous underwater vehicles (AUVs). In this paper, an automated gravity gradient tensor inversion algorithm is proposed for the purpose of passive underwater object detection. Full-tensor gravity gradient anomalies induced by an object in the partial area can be measured with the technique of gravity gradiometry on an AUV. Then the automated algorithm utilizes the anomalies, using the inverse method to estimate the mass and barycentre location of the arbitrary-shaped object. A few tests on simple synthetic models will be illustrated, in order to evaluate the feasibility and accuracy of the new algorithm. Moreover, the method is applied to a complicated model of an abnormal object with gradiometer and AUV noise, and interference from a neighbouring illusive smaller object. In all cases tested, the estimated mass and barycentre location parameters are found to be in good agreement with the actual values.

Underwater Obstacle Detection Based on the Change of Gravity Gradient

In view of the difficulty in obtaining gravity gradient reference maps at present, this paper presents a new method for detecting the barrier which is underwater. The method is autonomous, independent, and needing no gravity gradient reference map.This method Firstly needs record gravity gradiometers three readings in the submarine current route. Secondly subtract the first reading with the second reading, and subtract the second reading with the third reading, then we can get two differences. Do division of the two differences , finally we can get the inversion formula for the location of the obstacle. Through the Trust-Region-Reflective method to solve this equation, we will get the location information of the obstacle, then get its mass information. Experiments showed that using the gravity gradiometer whose precision is 10-5E, the method can detect not only the reef whose mass is 109kg magnitude within 500 meters but also the wreck whose mass is 107kg magnitude within 600 meters with no need of the gravity gradient reference map.Comparing with the traditional method of detecting obstacles, the new method meets the stringent requirements of safety and concealment very well. Whats more, the new method detects obstacles with no need of the data of the backgrounds gravity or gravity gradient. Therefore, the method which our paper presents is of great value in practical applications.

A New Electronic Image Stabilization Technology Based on Random Ferns for Fixed Scene

The paper propose a new electronic image stabilization based on Random Ferns algorithm for the fixed scene. We make an offline training after getting a view of the fixed scene to receive random ferns classifier for image matching and estimating motion. At last we remove the unintentional motion with kalman filter(KF) and make corrections to image sequence to get stable frames. This method is effective with scale, rotation, and shift. What’s more, the method is real-time and with high-precision.

Boundary extraction using supervised edgelet classification

Traditional learning-based boundary extraction algorithms clas- sify each pixel edge separately and then get boundaries from the local decisions of a classifier. However, we propose a supervised learning method for boundary extraction by using edgelets as boundary elements. First, we extract edgelets by clustering probabilities of boundary. Second, we use features of edgelets to train a classifier that determines whether an edgelet belongs to a boundary. The classifier is trained by utilizing edgelet features, including local appearance, multiscale features, and global scene features such as saliency maps. Finally, we use the classifier to decide the probability that the edgelet belongs to the boundary. The experimental results in the Berkeley Segmentation Dataset demonstrate that our algorithm can improve the performance of boundary extraction.

A Methodology for Ground Targets Detection in Complex Scene Based on Airborne LiDAR

In this paper, an approach to detecting ground targets using LiDAR point data is proposed. First, outliers are weeded out and point cloud is divided into ground points and non-ground points. Second, the ground surface plane is fitted by ground points and then the relative elevations of all non-ground points are estimated. If the relative elevations of non-ground points exceed a predefined threshold, they will be removed. Subsequently, a 3D region growing algorithm based on the normal vector consistency is employed to generate potential ground targets. Geometric information is used for further filtration of these potential targets on the object level. Finally, the detection performance of the algorithm is analyzed. The experimental results show that the method proposed is effective.