Finding logos in real-world images with point-context representation-based region search

Finding logos in the real-world images is a challenging task due to their small size, simple shape, less texture and clutter background. In this paper, through visual logo analysis with different types of features, we propose a novel framework for finding visual logos in the real-world images. First, we exploit the contextual shape and patch information around feature points, merge them into a combined feature representation (point-context). Considering the characteristics of logos, this kind of fusion is an effective enhancement for the discriminability of single point features. Second, to eliminate the interference of the complex and noisy background, we transfer the logo recognition to a region-to-image search problem by segmenting real-world images into region trees. A weak geometric constraint based on regions is encoded into an inverted file structure to accelerate the search process. Third, we apply global features to refine initial results in the re-ranking stage. Finally, we combine each region score both in max-response and accumulate-response mode to obtain the final results. Performances of the proposed approach are evaluated on both our CASIA-LOGO dataset and the standard Flickr logos 27 dataset. Experiments and comparisons show that our approach is superior to the state-of-the-art approaches.     

Nonlocal variational image segmentation models on graphs using the Split Bregman

Variational functionals such as Mumford-Shah and Chan-Vese methods have a major impact on various areas of image processing. After over 10 years of investigation, they are still in widespread use today. These formulations optimize contours by evolution through gradient descent, which is known for its overdependence on initialization and the tendency to produce undesirable local minima. In this paper, we propose an image segmentation model in a variational nonlocal means framework based on a weighted graph. The advantages of this model are twofold. First, the convexity global minimum (optimum) information is taken into account to achieve better segmentation results. Second, the proposed global convex energy functionals combine nonlocal regularization and local intensity fitting terms. The nonlocal total variational regularization term based on the graph is able to preserve the detailed structure of target objects. At the same time, the modified local binary fitting term introduced in the model as the local fitting term can efficiently deal with intensity inhomogeneity in images. Finally, we apply the Split Bregman method to minimize the proposed energy functional efficiently. The proposed model has been applied to segmentation of real medical and remote sensing images. Compared with other methods, the proposed model is superior in terms of both accuracy and efficient.     

Analytical study on cantilever resonance type magnet-integrated sensor device for micro-magnetic field detection

A beam-shaped cantilever resonance type magnetic sensor device has been proposed with a micro magnet. Two structural designs, named as design 1 and design 2, have been comparatively analyzed using ANSYS in order to obtain larger frequency shifts (higher magnetism sensitivity) due to the applied exterior magnetic field. The analytical results show that, in the range of 0–10 mT, the frequency shifts are small, while under 100 mT, a relatively larger frequency shift of about 30 Hz can be theoretically obtained. The power consumption of the proposed devices has been further theoretically studied for preliminary understanding. Using the well-known displacement equations, the estimated power consumption is around 0.21 μW, which is very lower than that of the reported magnetic field sensors. This implies that it is possible to fabricate higher sensitive magnetic field sensor with lower power consumption.     

自构建小波神经网络的内模控制在深度脱硅中的应用

为了稳定氧化铝深度脱硅过程的热工制度和降低能耗,采用小波神经网络辨识的内模控制方法进行氧化铝深度脱硅工艺过程控制。根据小波基函数的激励强度和衰减程度可以添加或者删除小波神经网络隐含层神经元,从而优化小波神经网络隐含层结构。再用自构建小波神经网络辨识内模控制系统的正模型和逆模型,从而改进神经网络内模控制技术。实验结果表明,所提出的控制方法比传统方法在鲁棒性和抗扰性方面具有更好的性能表现,各项指标均优于传统控制方法,实现了氧化铝深度脱硅工艺优化。     

基于动态约简的增量贝叶斯分类算法的研究

朴素贝叶斯由于条件独立性假设使其分类效果不明显,同时在处理海量数据时缺乏灵活性。针对以上情况,提出一种基于动态约简的增量贝叶斯分类算法。算法首先利用(F-λ)广义动态约简计算出数据集的核属性,然后根据训练集的先验信息构造分类器对测试实例进行分类,最后利用类置信度进行选择性增量学习,增强处理增量数据的能力。实验结果表明,该算法在处理属性少的小量数据时,分类效果有一定的改善,在处理多属性大量数据时,分类效果明显提高。