Shijin Li

An effective feature selection method for hyperspectral image classification based on genetic algorithm and support vector machine

With the development and popularization of the remote-sensing imaging technology, there are more and more applications of hyperspectral image classification tasks, such as target detection and land cover investigation. It is a very challenging issue of urgent importance to select a minimal and effective subset from those mass of bands. This paper proposed a hybrid feature selection strategy based on genetic algorithm and support vector machine (GA-SVM), which formed a wrapper to search for the best combination of bands with higher classification accuracy. In addition, band grouping based on conditional mutual information between adjacent bands was utilized to counter for the high correlation between the bands and further reduced the computational cost of the genetic algorithm. During the post-processing phase, the branch and bound algorithm was employed to filter out those irrelevant band groups. Experimental results on two benchmark data sets have shown that the proposed approach is very competitive and effective.

Time Series Outlier Detection Based on Sliding Window Prediction

In order to detect outliers in hydrological time series data for improving data quality and decision-making quality related to design, operation, and management of water resources, this research develops a time series outlier detection method for hydrologic data that can be used to identify data that deviate from historical patterns. The method first built a forecasting model on the history data and then used it to predict future values. Anomalies are assumed to take place if the observed values fall outside a given prediction confidence interval (PCI), which can be calculated by the predicted value and confidence coefficient. The use of PCI as threshold is mainly on the fact that it considers the uncertainty in the data series parameters in the forecasting model to address the suitable threshold selection problem. The method performs fast, incremental evaluation of data as it becomes available, scales to large quantities of data, and requires no preclassification of anomalies. Experiments with different hydrologic real-world time series showed that the proposed methods are fast and correctly identify abnormal data and can be used for hydrologic time series analysis.

A Hybrid Evolutionary Approach to Band Selection for Hyperspectral Image Classification

With the development of the remote-sensing imaging technology, there are more and more applications of hyperspectral image classification tasks, in which to select a minimal and effective subset from a mass of bands is the key issue. This paper put forward a novel band selection strategy based on conditional mutual information between adjacent bands and branch and bound algorithm for the high correlation between the bands. In addition, genetic algorithm and support vector machine are employed to search for the best band combination. Experimental results on two benchmark data set have shown that this approach is competitive and robust.

A Novel Approach to Band Selection for Hyperspectral Image Classification

To select a minimal and effective subset from a mass of bands is the key issue of the study on hyperspectral image classification. This paper put forwards a novel band selection algorithm, which combines mutual information-based grouping method and genetic algorithm. The proposed algorithm reduces the computation cost significantly, as well as keeps a better precision. In addition, resampling based on sequential clustering is employed to tackle the imbalanced data issue and improve the classification accuracy of minority classes. Experimental results on the Washington DC Mall data set validate the effectiveness and efficiency of the proposed algorithm.

Study on Highlights Detection in Soccer Video Based on the Location of Slow Motion Replay and Goal Net Recognition

The problem of highlight detection and recognition in soccer video has been a hot research topic in many fields, such as image processing, pattern recognition, and machine learning. This paper puts forward a novel algorithm to detect soccer highlights, which is made up of three phases. In the first stage, slow motion replay segments are located through the analysis of visual rhythm and the structure tensor histogram; In the second stage, we concentrate on the detection of the goal net according to its edge projection in the vertical direction; In the end, a set of heuristic rules are employed to identify the goal shots or near-miss shots from the soccer video based on the recognition result of the previous stage. Experimental results show that our method is effective and efficient.

Region-based retrieval of remote sensing image patches with adaptive image segmentation

Over the past four decades, the satellite imaging sensors have acquired huge quantities of Earth- observation data. Content-based image retrieval allows for fast and effective queries of remote sensing images. Here, we take the following two issues into consideration. Firstly, different features and their combination should be chosen for different land covers. Secondly, for the block dividing strategy and the complexities of the remote sensing images, it can not effectively retrieve some small target areas scattered in multiple nontarget blocks. Aiming at the above two issues, a new region-based retrieval method with adaptive image segmentation is proposed. In order to improve the accuracy of remote sensing image segmentation, feature selection and weighing is performed by two-stage clustering, and image segmentation is accomplished based on the chosen features and mean shift procedure. Meanwhile, for the homogeneous characteristics of remote sensing land covers, a new regional representation and matching scheme are adopted to perform image retrieval. Experimental results on retrieving various land covers show that the method can avoid the impact of traditional blocking strategies, and can achieve an average percentage of 19% higher precision with the same level of recall rate, than the relevance feedback method for small target areas.

Clustering-based feature selection for content based remote sensing image retrieval

During the last three decades, the imaging satellite sensors have acquired huge quantities of remote sensing data. Content-based image retrieval is one of the effective and efficient techniques for utilizing those Earth observation data sources. In this paper, a novel remote sensing image retrieval approach, which is based on feature selection and semi-supervised learning, is proposed. The new method includes four steps. Firstly, clustering is employed to select features and the number of clusters is determined automatically by using the MDL criterion; Secondly, according to an improved clustering validity index, we select the optimal features which can describe the retrieval objectives efficiently; Thirdly, the weights of the selected features are dynamically determined; and finally, an appropriate semi-supervised learning scheme is adaptively selected and image retrieval is thus conducted. Experimental results show that, the proposed approach can achieve comparable performance to the relevance feedback method, while ours need no human interaction.

Soil Erosion Remote Sensing Image Retrieval Based on Semi-Supervised Learning

Soil erosion is one of the most typical natural disasters in China. However, due to the limitation of current technology, the investigation of soil erosion through remote sensing images is currently by human beings manually which depends on human interpretation and interactive selection. The work burden is so heavy that errors are usually inevitably unavoidable. This paper proposes the technique of content-based image retrieval to tackle this problem. Due to the large amount of computation in co-training retrieval based on multiple classifier systems, and for the purpose of improving efficiency, an improved approach using co-training in two classifier systems is proposed in this paper. Prior to retrieving, we firstly select the optimal color feature and texture feature respectively, and then use the corresponding color classifier and texture classifier for co-training. By this approach, the time of co-training is reduced greatly, meanwhile, the selected optimal features can represent color and texture features better for remote sensing image, resulting in better retrieval accuracy. Experimental results show that the improved approach using co-training in two classifier systems needs less amount of computation and less retrieval time, while it can lead to better retrieval results.