Main contributions and contents: Based on the theory of probability, statistics, and robotics control theory, we construct eye-hand co-working characterizing models. We also adopt from multi-view points to use valid view-angle cameras to extract users' figure pointing gesture multi-modal characteristics. Through the pattern classification of multi-modal visual characteristics of various users, we determine the target pointing objects of different users, to achieve free multi-view point, non-wearable multi-people figure pointing gesture visual recognition.

Main contributions and contents: Based on finite video sequence, the characteristics of behaviors automatically labeled are used for adaptive training of unsupervised learning models, an then employed for the anomaly detection of the behaviors of the moving subjects in the scene, to achieve automatic anomaly recognition of moving subjects in intelligent video surveillance

Main contributions and contents: In our developed system, the human hands are directly used as the input control devices, and their moving ranges are not limited. Via the understanding and inference of the figure pointing gesture of the targeting users, we achieve free multi-view point based non-wearable intelligent figure pointing gesture visual recognition.

Main contributions and contents: With integrated utilization of various geo-physics sensing methods, we analysis the observed data using intelligent data processing systems, in order to automatically determine the level of security and exact positions of any dangerous parts of the sea levee; we also establish synchronous databases to manage the sensed data systematically, overcoming the drawbacks of using the traditional paper-based materials and information management.