Jiaheng Cao

A new watermark mechanism for relational data

This paper studies digital watermark technology of numeric attributes in relational database for datas copyright protection. It analyzes current researches on this issue and proposes a new watermark mechanism (NWM) using classifying and twice majority voting method, which pays much attention to datas usability through relational structure and semantic constraints, marks each attribute separately, and makes marked bits unobtrusive and robust to various attacks. We implement the watermarking algorithm and give the experimental evaluation.

Geo-WDBMS: an improved DBMS with the function of watermarking geographical data

This paper focuses on the issue of geographical datas copyrights protection. A Geo-WDBMS has been built by embedding the watermarking functions into the inner code of the open source DBMS PostgreSQL. And its core watermarking mechanism is to insert and detect mark bits in the coordinates of the vertices in geographical objects using the methods of classifying and twice majority-voting. Further more, error correcting mechanism is used to enhance the resilience of the system and blind watermark is realized. Experiments on watermarking digital maps showed that the marked maps are inconspicuous and robust to various attacks.

Data mining based on objects in video flow with dynamic background

This paper presents a model OMDB for mining the region information of non-rigid foreground object in video flow with dynamic background. The model constructs RDM algorithm and optimize the strategy of region matching using Q-learning to obtain better motion information of regions. Moreover, OMDB utilizes NEA algorithm to detect and merge gradually object regions of foreground based on the characteristics that there is motion difference between foreground and background and the regions of an object maintain integrality during moving. Experimental results on extracting region information of foreground object and tracking the object are presented to demonstrate the efficacy of the proposed model.

A novel 3D video trajectory tracking method

In this paper an automatic trajectory tracking method is presented which combines with nodes in 3D space to structure object trajectory. The nodes are extracted frequently from two analogue cameras. A 3D orientation model is brought forward to ascertain the orientation of object centroid and predict the probabilistic location some time ahead. A 3D trajectory model is used to log object motion information, recognize object behavior pattern and classify object trajectories based on the concept of deputy object. The method is efficient in many domains, such as especial video guard against theft, video surveillance for objects in air, etc.

Indexing partial history trajectory and future position of moving objects using HTPR*-Tree

Currently, most indexing methods of moving objects are focused on the past position, or the present and future one. In this paper, we propose a novel indexing method, called History TPR*-tree(HTPR*-tree), which not only supports predictive queries but also partial history ones involved from the most recent update instant of each object to the last update time of all objects. Based on the TPR*-tree, our Basic HTPR*-tree adds creation or update time of moving objects to leaf node entries. In order to improve the update performance, we present a bottom-up update strategy for the HTPR*-tree by supplementing a hash table, a bit vector and a direct access table. Experimental results show that the update performance of the HTPR*-tree is better than that of the Basic HTPR*-and TPR*-tree. In addition to support partial history queries, the update and predictive query performance of the HTPR*-tree are greatly improved compared with those of the RPPF-tree.

Content based image retrieval using Gaussian Lie group spatiogram similarity

In this paper, we apply the spatiogram features to the image retrieval problem using a recent proposed Lie group based similarity measure. The spatiogram features are extracted at HSV space. For each channel of the HSV space, we extract the corresponding spatiogram features separately. Such kind of feature extraction method produce much lower dimension features than extracting features directly on the 3-D HSV space. Then, we compare the images using a recently proposed spatiogram distance measure, which is based on the Lie group theory. We test our algorithm on the corel image retrieval benchmark dataset. Experiments show better performance than the previous proposed spatiogram similarity measure.

Kernel Based Spatiogram Tracking Using Improved Similarity Measure

Spatiogram were generalization of histograms, which can harvest spatial information of images. In this paper, we address the object tracking problem using spatiogram as feature descriptor. We use an improved spatiogram imilaritymeasure which is recently proposed. Based on the measure, we derive a kernel tracking algorithm utilizing Mean Shift procedure. We test our tracking algorithm on several datasets. Experiment show better tracking result compared with the previously proposed kernel based spatiogram tracking algorithm.

A Novel Cross-Media Layered Semantic Mining Model

This paper presents a cross-media semantic mining model (CSMM) based on object semantic. This model obtains object-level semantic information in terms of maximum probability principle. Then semantic templates are trained and constructed with STTS (Semantic Template Training System), which are taken as the bridge to realize the transition from various low-level media feature to object semantic. Furthermore, we put forward a kind of double layers metadata structure to efficaciously store and manage mined low-level feature and high-level semantic. This model has broad application in lots of domains such as intelligent retrieval engine, medical diagnoses, multimedia design and so on.

Efficient Indexing of the Past, Present and Future Positions of Moving Objects on Road Network

Aim at moving objects on road network, we propose a novel indexing named PPFN*-tree to store past trajectories, present positions, and predict near future positions of moving objects. PPFN*-tree is a hybrid indexing structure which consists of a 2D R*-tree managing the road networks, a set of TB*-tree indexing objects’ movement history trajectory along the polylines, and a set of basic HTPR*-tree indexing the position of moving objects after recent update. PPFN*-tree can not only support past trajectory query and present position query, but also support future predictive query. According to the range query time, query in PPFN*-tree can be implemented only in the TB*-tree, or only in the HTPR*-tree, or both of them. Experimental results show that the update performance of the PPFN*-tree is better than that of the PPFI and the RPPF-tree. The query performance of the PPFN*-tree is better than that of the MON-Tree and the PPFI.