Sheng Weiguo

A steganalysis method in the DCT domain

In this paper, we propose a steganalysis method based on the principle of additive operator, which chooses non-zero AC coefficients as carriers, with secret information independent of the carrier information flow. In the proposed method, AC coefficient statistical and energy features are initially extracted and used to construct a 3D feature vector. By employing the principle of Fisher linear discriminate analysis, a flexible discriminate classifier suitable for the extracted features is designed to improve detection performance. We infer and confirm theory of change in the statistical and energetic characteristics of the AC coefficient before and after additive steganography. The effectiveness of the proposed method is proven by the experiments. Moreover, the proposed method consistently outperforms related methods.

A 3D occlusion tracking Model of the underwater fish targets

Multi-objective shining positioning and tracking technology is essential to behavior analysis in fish tracking. To achieve multi-target tracking, this paper proposes an occlusion-based method of fish tracking based on water quality testing platform. We systematically derive a theoretical model of real fish occlusion tracking, imaginary fish occlusion tracking, and real and imaginary fish concurrent occlusion tracing based on the following methods: the plan mirror imaging principle, and basic techniques of image processing such as target segmentation. Using this model, we aim to achieve 3D fish tracking without occlusion, fulfilling the requirements beforehand for fish target behavior analysis. The experimental results show the effectiveness of the proposed 3D fish target occlusion tracking Model.