Object detection using hybridization of static and dynamic feature spaces and its exploitation by ensemble classification

This paper presents a learning mechanism based on hybridization of static and dynamic learning. Realizing the detection performances offered by the state-of-the-art deep learning techniques and the competitive performances offered by the conventional static learning techniques, we propose the idea of exploitation of the concatenated (parallel) hybridization of the static and dynamic learning-based feature spaces. This is contrary to the cascaded (series) hybridization topology in which the initial feature space (provided by the conventional, static, and handcrafted feature extraction technique) is explored using deep, dynamic, and automated learning technique. Consequently, the characteristics already suppressed by the conventional representation cannot be explored by the dynamic learning technique. Instead, the proposed technique combines the conventional static and deep dynamic representation in concatenated (parallel) topology to generate an information-rich hybrid feature space. Thus, this hybrid feature space may aggregate the good characteristics of both conventional and deep representations, which are then explored using an appropriate classification technique. We also hypothesize that ensemble classification may better exploit this parallel hybrid perspective of the feature spaces. For this purpose, pyramid histogram of oriented gradients-based static learning has been incorporated in conjunction with convolution neural network-based deep learning to produce concatenated hybrid feature space. This hybrid space is then explored with various state-of-the-art ensemble classification techniques. We have considered the publicly available INRIA person and Caltech pedestrian standard image datasets to assess the performance of the proposed hybrid learning system. Furthermore, McNemar’s test has been used to statistically validate the outperformance of the proposed technique over various contemporary techniques. The validated experimental results show that the employment of the proposed hybrid representation results in effective detection performance (an AUC of 0.9996 for INRIA person and 0.9985 for Caltech pedestrian datasets) as compared to the individual static and dynamic representations.

     

An Efficient Mechanism for Product Data Extraction from E-Commerce Websites

A large amount of data is present on the web which can be used for useful purposes like a product recommendation, price comparison and demand forecasting for a particular product. Websites are designed for human understanding and not for machines. Therefore, to make data machine-readable, it requires techniques to grab data from web pages. Researchers have addressed the problem using two approaches, i.e., knowledge engineering and machine learning. State of the art knowledge engineering approaches use the structure of documents, visual cues, clustering of attributes of data records and text processing techniques to identify data records on a web page. Machine learning approaches use annotated pages to learn rules. These rules are used to extract data from unseen web pages. The structure of web documents is continuously evolving. Therefore, new techniques are needed to handle the emerging requirements of web data extraction. In this paper, we have presented a novel, simple and efficient technique to extract data from web pages using visual styles and structure of documents. The proposed technique detects Rich Data Region (RDR) using query and correlative words of the query. RDR is then divided into data records using style similarity. Noisy elements are removed using a Common Tag Sequence (CTS) and formatting entropy. The system is implemented using JAVA and runs on the dataset of real-world working websites. The effectiveness of results is evaluated using precision, recall, and F-measure and compared with five existing systems. A comparison of the proposed technique to existing systems has shown encouraging results.     

A new rectangular window based image cropping method for generalization of brain neoplasm classification systems

Classification of brain neoplasm images is one of the most challenging research areas in the field of medical image processing. The main objective of this study is to design a brain neoplasm classification system that can be trained using multiple various sized MR images of different institutions. The proposed method is a generalized classification system; it can be used in a single institute or in a number of institutions at the same time, without any restriction on image size. The generalization and unbiased capability of the proposed method can bring researchers on a single platform to work on some standard forms of computer aided diagnosis system with more efficient diagnostic capabilities. In this study, a suitable size of moveable rectangular window is used between segmentation and feature extraction stages. A semiautomatic, localized region based active contour method is used for segmentation of brain neoplasm region. Discrete wavelet transform for feature extraction, principal component analysis for feature selection and support vector machine is used as classifier. For the first time MR images of 2 sizes and from different institutions are used in training and testing of brain neoplasm classifier. Three glioma grades were classified using 92 MR images. The proposed method achieved the highest accuracy of 88.26%, the highest sensitivity of 92.23% and the maximum specificity of 93.93%. In addition, the proposed method is computationally less complex, requires shorter processing time and is more efficient in terms of storage capacity.     

On increasing the sensitivity of moving average control chart using auxiliary variable

In the statistical process control, the most useful tool to monitor the manufacturing processes in the industries is the control chart. Quality practitioners always desire the charting structure that identifies sustainable changes in the monitoring processes. The sensitivity of the control chart is improved when additional correlated auxiliary information about the study variable is introduced. The regression estimate in the form of auxiliary and supporting variables presents an unbiased and efficient statistic of the mean of the process variable. In this study, auxiliary information-based moving average (AB-MA) control chart is designed for efficient monitoring of shifts in the process location parameter. The performance of the AB-MA control chart is evaluated and compared with existing charts using average run length and other run length characteristics. The comparison reveals that the AB-MA control chart outperforms the competitors in detecting the small and medium changes in the process location parameter. The application of the proposal is also provided to implement it in real situation.     

Adaptive CUSUM and EWMA charts with auxiliary information and variable sampling intervals for monitoring the process mean

The variable sampling interval (VSI) feature enhances the sensitivity of a control chart that is based on fixed sampling interval (FSI). In this paper, we enhance the sensitivities of the auxiliary information-based (AIB) adaptive Crosier cumulative sum (CUSUM) (AIB-ACC) and adaptive exponentially weighted moving average (EWMA) (AIB-AE) charts using the VSI feature when monitoring a mean shift which is expected to lie within a given interval. The Monte Carlo simulations are used to compute zero-state and steady-state run length properties of these control charts. It is found that the AIB-ACC and AIB-AE charts with VSI feature are uniformly more sensitive than those based on FSI feature. Real datasets are also considered to demonstrate the implementation of these control charts.     

An enhanced approach for the progressive mean control charts

To maintain and improve the quality of the processes, control charts play an important role for reduction of variation. To detect large shifts in the process parameters, Shewhart control charts are commonly applied but for small shifts, exponentially weighted moving averages (EWMA), cumulative sum (CUSUM), double exponentially weighted moving average (DEWMA), double CUSUM, moving average (MA), double moving average (DMA), and progressive mean (PM) control charts, are used. This study proposes double progressive mean (DPM) and optimal DPM control charts to enhance the performance of the PM chart. As the proposed DPM control charts use information sequentially, hence their performance is compared with natural competitors EWMA, CUSUM, DEWMA, double CUSUM, MA, DMA, and PM control charts. Run length and its different properties are evaluated to compare the performance of the proposed charts and counterparts. Results reveal that proposed optimal DPM outperforms the other charts. An example related to voltage on fixed capacitance level is also provided to illustrate the proposed charts.     

Novel analysis of the size and orientation dependent resonance phenomenon of the microstrip line coupled complementary split‐ring resonator

The purpose of this article is to provide a comprehensive investigation on the resonance phenomenon of microstrip line coupled complementary split‐ring resonator (CSRR) with different orientation and relative size. It is shown that when the relative size of the CSRR is smaller than the host line, the CSRR with its slit oriented orthogonal to the line axis will not excite effectively and show weak resonance behavior. However, when the slit is positioned along the line axis, the cross‐polarization effect comes into play, which excites the CSRR through the mixed coupling. To ensure the correctness, several numerical simulations are carried out for different substrate height and relative permittivity. Finally, a prototype is fabricated and measured for the experimental validation.