Search

Optical colonoscopy is an invasive procedure used to examine surface lining of the colon, by inserting a flexible tube with a light and camera into the body. Virtual colonoscopy is emerging as a non-invasive alternative to optical colonoscopy. This paper describes 3D reconstruction of colon structure from patient specific CT images, interactive visualization and navigation through the reconstructed colon, automated computation of the navigation path and automated polyp detection with 90.91% sensitivity. This enables a doctor to perform a fast diagnosis through virtual exploration of the colon and resort to invasive colonoscopy procedures only if suspicious polyps are detected.

The raw data acquired from MR scanners are susceptible to noises resulting from truncation, inhomogeneity and patient motion. Truncation or Gibb's artifact is one of the most commonly seen artifact in MR images. Since the raw MR data are samples of Fourier transform, reconstructing images from this data will generate truncation artifacts. The application of apodization filters can reduce truncation artifacts in MR images. In this paper, we discuss and analyze the image quality improvements on applying two popular apodization filters called Fermi and Hamming in the image reconstruction pipeline.

Vehicle tracking systems are conventionally used for applications like fleet management, logistics, online taxi service, asset tracking, fuel monitoring etc. Vehicle tracking system comprises of Vehicle Tracking Device (VTD) mounted on the vehicles, which sends current data like location, speed and alerts to a backend software system. The software system processes the data from device to generate reports, plot the vehicle locations on a GIS map etc. Apart from the normal applications of vehicle tracking, this technology can be used for the purpose of traffic enforcement and safety by the Transport or Police department of a state or country. In India, the state of Kerala has an upcoming implementation of vehicle tracking system to assist enforcement and safety. The proposed system mandates public transport vehicles to be fitted with VTD and the vehicles shall be monitored from the control room. The backend software processes the data to generate reports of offences such as over speeding, entry to restricted area etc. for each vehicle connected to the system. The offence report lists out the details of offences including location, time, offence and a snapshot of the location with a marker at the point of offence. The snapshot of map is obtained from the static map functionality of the GIS platform using center geo-coordinate, width, height and zoom level as inputs. Depending on the zoom level of the GIS map, the image generated may or may not contain any text content as a reference to identify the location of offence. This paper proposes to use image processing mechanisms to identify presence of text content on the GIS image which helps the offence processing system to select GIS image at the appropriate zoom level, so as to identify text content for reference.

In India a countless number of children are reported missing every year. Among the missing child cases a large percentage of children remain untraced. This paper presents a novel use of deep learning methodology for identifying the reported missing child from the photos of multitude of children available, with the help of face recognition. The public can upload photographs of suspicious child into a common portal with landmarks and remarks. The photo will be automatically compared with the registered photos of the missing child from the repository. Classification of the input child image is performed and photo with best match will be selected from the database of missing children. For this, a deep learning model is trained to correctly identify the missing child from the missing child image database provided, using the facial image uploaded by the public. The Convolutional Neural Network (CNN), a highly effective deep learning technique for image based applications is adopted here for face recognition. Face descriptors are extracted from the images using a pre-trained CNN model VGG-Face deep architecture. Compared with normal deep learning applications, our algorithm uses convolution network only as a high level feature extractor and the child recognition is done by the trained SVM classifier. Choosing the best performing CNN model for face recognition, VGG-Face and proper training of it results in a deep learning model invariant to noise, illumination, contrast, occlusion, image pose and age of the child and it outperforms earlier methods in face recognition based missing child identification. The classification performance achieved for child identification system is 99.41%. It was evaluated on 43 Child cases.

Automated cervical cancer detection system is primarily based on delineating the cell nuclei and analyzing their textural and morphometric features for malignant characteristics. The presence of cell clusters in the slides have diagnostic value, since malignant cells have a greater tendency to stick together forming clusters than normal cells. However, cell clusters pose difficulty in delineating nucleus and extracting features reliably for malignancy detection in comparison to free lying cells. LBC slide preparation techniques remove biological artifacts and clustering to some extent but not completely. Hence cluster detection in automated cervical cancer screening becomes significant. In this work, a graph theoretical technique is adopted which can identify and compute quantitative metrics for this purpose. This method constructs a cell graph of the image in accordance with the Waxman model, using the positional coordinates of cells. The computed graph metrics from the cell graphs are used as the feature set for the classifier to deal with cell clusters. It is a preliminary exploration of using the topological analysis of the cellgraph to cytological images and the accuracy of classification using SVM showed that the results are well suited for cluster detection.

This paper presents a method for induction motor fault diagnosis based on rotor current signal analysis using Support Vector Machine. A dynamic model of induction motor developed using SIMULINK/MATLAB environment is used for simulation testing. A rotor fault is incorporated into the developed dynamic model which is mathematically complaint. The simulated model gives rotor currents, the multi-resolution analysis of which is conducted in the wavelet domain for the detection of broken bars. The analyzed data itself is indicative of the incipient faults, but mere human inspection can sometimes lead to unexpected faults. Hence, a classification scheme using Support Vector Machine is adopted. Finally, the results of Support Vector classification is compared against that of Artificial Neural Networks.