In recent years, change detection (CD) in remote sensing images has achieved a huge success by using deep learning, and it is essentially a subtask of semantic segmentation. Both of them are tightly related. However, the complementarity of them has not been fully explored. In this paper, we propose a new change detection converter (CDC), which can be easily inserted into the existing semantic segmentation networks and be transformed into new networks suitable for change detection task. In addition, a new task-specific data augmentation method named Crossover is proposed, which can enhance the feature representation ability in sequence-form features without any additional cost. Extensive experiments demonstrate that the proposed method obtains significantly better performance and more efficiency than previous methods on two public datasets, and can achieve continuous performance improvement than other complex-designed solutions.

In this paper, we propose a novel probabilistic graphical model to address the off-line signature verification problem. Different from previous work, our approach introduces the concept of feature roles according to their distribution in genuine and forgery signatures, with all these features represented by a unique graphical model. And we propose several new techniques to improve the performance of the new signature verification system. Results based on 200 persons' signatures (16000 signature samples) indicate that the proposed method outperforms other popular techniques for off-line signature verification with a great improvement.

The facility location problem has been studied in many industries including banking network, chain stores, and wireless network. Maximal covering location problem (MCLP) is a general model for this type of problems. Motivated by a real-world banking facility optimization project, we propose an enhanced MCLP model which captures the important features of this practical problem, namely, varied costs and revenues, multitype facilities, and flexible coverage functions. To solve this practical problem, we apply an existing hybrid nested partitions algorithm to the large-scale situation. We further use heuristic-based extensions to generate feasible solutions more efficiently. In addition, the upper bound of this problem is introduced to study the quality of solutions. Numerical results demonstrate the effectiveness and efficiency of our approach.

It is critical for retail enterprises to select good sites or locations to open their stores, especially in current competitive retail market. However, evaluating the goodness of sites in real business applications is a complex problem. That is, how to judge whether the market around a store site is good? We don't know the exact mechanism of how a site can be good and it is hard to have correct site goodness values as supervised labels. The Retail Outlet Site Evaluation (ROSE) tool is designed to learn the site evaluation model by integrating city geographic & demographic data and two kinds of expert knowledge: sample preference and feature preference. The feature preference information can help greatly reduce the required number of sample preferences. It enables our application practicable because it is almost impossible to give such amount of sample preference pairs manually by experts when ranking hundreds of data points. In the experiment and case study part, we show that the ROSE tool can achieve good results and useful for users to do site evaluation work in real cases.

A Hidden Markov Model (HMM) approach to off-line signature verification is presented. First, each of the signature images is represented as a landmark point set, which includes turning points, isolated points, trifurcate points, intersection points and termination points on signature skeleton. Then we propose a novel deformable grid partition technique. Based on landmark point matching, we build the matching relations between planar regions to get the deformable grids, and then extract grid features from them. By using HMM in signature modeling, the deformable grid partition method shows remarkable improvements over traditional grid partition methods in discriminative ability.

We address the problem of feature weight learning for image clustering. In practice, before clustering data, we generally normalize all data features between 0 and 1, because we cannot determine which features are more important. In this paper, we provide a feature weight learning framework for clustering which can obtain the feature weights and cluster labels simultaneously. An alternative optimization algorithm is adopted to solve this problem. Empirical studies on the toy data and real image data demonstrate our algorithm's effectiveness in improving the clustering performance.

This paper describes a new approach to emotion recognition based on pressure sensor keyboards. The pressure sensor keyboard is a new product that occurs in the market recently, which produces a pressure sequence when keystroke occurs. The analysis of the pressure sequence should be a novel research area. It has been used for identity verification in our previous research. In this paper, we use the pressure sequence for emotion recognition. Three methods (global features of pressure sequences, dynamic time warping and traditional keystroke dynamics) are proposed for the emotion recognition task; then we combined the three methods together using a classifier fusion technique. Several experiments were performed on a database containing 3000 samples (from 50 individuals, including six emotions: neutral, anger, fear, happiness, sadness and surprise) and the best result were achieved utilizing all the method, obtaining an overall accuracy of 93.4%. Our technique of emotion recognition has been used for intelligent game controlling and several other applications.

This paper proposes a semi-supervised distance metric learning algorithm for the ranking problem. Instead of giving the computer what are the important factors that affect the final rank value, we only give several most certainly ranked points which implicitly contain the knowledge of the ranking factors. Then the computer can automatically use the most certain points and plenty of unlabeded data to learn an informative metric for ranking. This metric not only can help to regress an order in the observed data, but also can be used to retrieve the data by querying new test points. Moreover, the lower-rank distance metric can be used to visualize high-dimensional data. We also present an application to the housing potential estimation problem. It is shown that the algorithm is efficient to help consultants to refine their consulting work.

Genetic programming (GP) has been rarely applied to scheduling problems. In this paper the use of GP to learn single-machine predictive scheduling (PS) heuristics with stochastic breakdowns is investigated, where both tardiness and stability objectives in face of machine failures are considered. The proposed bi-tree structured representation scheme makes it possible to search sequencing and idle time inserting programs integratedly. Empirical results in different uncertain environments show that GP can evolve high quality PS heuristics effectively. The roles of inserted idle time are then analysed with respect to various weighting objectives. Finally some guides are supplied for PS design based on GP-evolved heuristics.