International Journal of Image, Graphics and Signal Processing @ijigsp
Статьи журнала - International Journal of Image, Graphics and Signal Processing
Все статьи: 1128
Automated Pre-Seizure Detection for Epileptic Patients Using Machine Learning Methods
Статья научная
Epilepsy is a neurological disorder resulting from unusual electrochemical discharge of nerve cells in the brain, and EEG (Electroencephalography) signals are commonly used today to diagnose the disorder that occurs in these signals. In this study, it was aimed to use EEG signals to automatically detect pre-epileptic seizure with machine learning techniques. EEG data from two epileptic patients were used in the study. EEG data is passed through the preprocessing stage and then subjected to feature extraction in time and frequency domain. In the feature extraction step 26 features are obtain to determine the seizure time. When the feature vector is analyzed, it is observed that the characteristics of the pre-seizure and non-seizure period are unevenly distributed. A systematic sampling method has been applied for this imbalance. For the balanced data, two test sets with and without Eta correlation are established. Finally, the classification process is performed using the k-Nearest Neighbor classification method. The obtained data are evaluated in terms of Eta-correlated and uncorrelated accuracy, error rate, precision, sensitivity and F-criterion for each channel.
Бесплатно
Automated Quality Inspection of PCB Assembly Using Image Processing
Статья научная
Quality inspection of PCB is a crucial stage in the assembly line as it provides an insight on whether the board works correctly or not. When the inspection is done manually, it is susceptible to human errors and is time consuming. The boards should thus be inspected at every stage of the assembly line and the process should be dynamic. This is achieved in this work through three crucial stages in the assembly line and by replacing the conventional manual inspection by using image processing to obtain a faster and more precise quality inspection. The solder paste inspection consists of pre-processing using blue plane conversion, comparing with the unsoldered board in blue color plane and post processing using overlay. The X-ray inspection basically consists of pre- processing the captured image by RGB to gray conversion with thresholding, comparing with the expected image and post processing using overlay to show the shorts that has occurred along the assembly. The conformal coating inspection uses conversion of the blue intensity emitted off the board under UV light to RGB scale. Each of the algorithms were tested using 48 actual in-production boards from Vinyas IT Pvt Ltd, a PCB assembly company based in Mysore. The processing time of the algorithms were found to be less than 2 seconds with an accuracy of 85.7%. The system was also found to be cost effective over existing systems available in the market.
Бесплатно
Automated paddy variety recognition from color-related plant agro-morphological characteristics
Статья научная
The paper presents an image-based paddy plant variety recognition system to recognize 15 different paddy plant varieties using 18 color-related agro-morphological characteristics. The k-means color clustering method has been used to segment the target regions in the paddy plant images. The RGB, HSI and YCbCr color models have been employed to construct color feature vectors from the segmented images and the feature vectors are reduced using Principal Component Analysis (PCA) technique. The reduced color feature vectors are used as input to back propagation neural network (BPNN) and support vector machine (SVM). The set of six combined agro-morphological characteristics recorded during maturity growth stage has given the highest average paddy plant variety recognition accuracies of 91.20% and 86.33% using the BPNN and SVM classifiers respectively. The work finds application in developing a tool for assisting botanists, Rice scientists, plant breeders, and certification agencies.
Бесплатно
Automatic Detection of Surface Defects on Citrus Fruit based on Computer Vision Techniques
Статья научная
In this paper, we present computer vision based technique to detect surface defects of citrus fruits. The method begins with background removal using k-means clustering technique. Mean shift segmentation is used for fruit region segmentation. The candidate defects are detected using threshold based segmentation. In this stage, it is very difficult to differentiate stem-end from actual defects due to similarity in appearance. Therefore, we proposed a novel technique to differentiate stem-end from actual defects based on the shape features. We conducted experiments on our citrus data set captured in controlled environment. The experiment results demonstrate that our technique outperforms the existing techniques.
Бесплатно
Статья научная
This paper describes automatic detection and classification of visual symptoms affected by fungal disease. Algorithms are developed to acquire and process color images of fungal disease affected on commercial crops like chili, cotton and sugarcane. The developed algorithms are used to preprocess, segment, extract and reduce features from fungal affected parts of a crop. The feature extraction is done with discrete wavelet transform (DWT) and features are further reduced by using Principal component analysis (PCA). Reduced features are then used as inputs to classifiers and tests are performed to classify image samples. We have used statistical based Mahalanobis distance and Probabilistic neural network (PNN) classifiers. The average classification accuracies using Mahalanobis distance classifier are 83.17% and using PNN classifier are 86.48%
Бесплатно
Automatic Image Segmentation Base on Human Color Perceptions
Статья научная
In this paper we propose a color image segmentation algorithm based on perceptual color vision model. First, the original image is divide into image blocks which are not overlapped; then, the mean and variance of every image back was calculated in CIEL*a*b* color space, and the image blocks were divided into homogeneous color blocks and texture blocks by the variance of it. The initial seed regions are automatically selected depending on calculating the homogeneous color blocks' color difference in CIEL*a*b* color space and spatial information. The color contrast gradient of the texture blocks need to calculate and the edge information are stored for regional growing. The fuzzy region growing algorithm and coloredge detection to obtain a final segmentation map. The experimental segmentation results hold favorable consistency in terms of human perception, and confirm effectiveness of the algorithm.
Бесплатно
Automatic Recognition and Counterfeit Detection of Ethiopian Paper Currency
Статья научная
Currency recognition is a technology used to identify currencies of various countries. The use of automatic methods of currency recognition has been increasing due its importance in many sectors such as vending machine, railway ticket counter, banking system, shopping mall, currency exchange service, etc. This paper describes the design of automatic recognition of Ethiopian currency. In this work, we propose hardware and software solutions which take images of an Ethiopian currency from a scanner and camera as an input. We combined characteristic features of currency and local feature descriptors to design a four level classifier. The design has a categorization component, which is responsible to denominate the currency notes into their respective denomination and verification component which is responsible to validate whether the currency is genuine or not. The system is tested using genuine Ethiopian currencies, counterfeit Ethiopian currencies and other countries' currencies. The denomination accuracy for genuine Ethiopian currency, counterfeit currencies and other countries' currencies is found to be 90.42%, 83.3% and 100% respectively. The verification accuracy of our system is 96.13%.
Бесплатно
Automatic Speech Segmentation Based On Audio and Optical Flow Visual Classification
Статья научная
Automatic speech segmentation as an important part of speech recognition system (ASR) is highly noise dependent. Noise is made by changes in the communication channel, background, level of speaking etc. In recent years, many researchers have proposed noise cancelation techniques and have added visual features from speaker’s face to reduce the effect of noise on ASR systems. Removing noise from audio signals depends on the type of the noise; so it cannot be used as a general solution. Adding visual features improve this lack of efficiency, but advanced methods of this type need manual extraction of visual features. In this paper we propose a completely automatic system which uses optical flow vectors from speaker’s image sequence to obtain visual features. Then, Hidden Markov Models are trained to segment audio signals from image sequences and audio features based on extracted optical flow. The developed segmentation system based on such method acts totally automatic and become more robust to noise.
Бесплатно
Automatic dead zone detection in 2-D leaf image using clustering and segmentation technique
Статья научная
Plant is a gift of almighty to the living being in the earth. Leaf is an essential component for any types of plant including crops, fruit and vegetables. Before the scheduled decay of the leaf due to deficiency there are patches of dead zone spot or sections generally visible. This paper introduces a novel image based analysis to identify patches of dead zone spot or sections generally visible due to deficiency. Clustering, colour object based segmentation and colour transformation techniques using significant salient features identification are applied over 12 plant leaves collected naturally from gardens and crop fields. Hue, saturation and Value based and L*a*b* colour model based object analysis is being applied over diseased leaf and portion of leaf to identify the dead zone automatically. Derivative based edge analysis is being applied to identify the outline edge and dead zone segmentation in leaf image. K-means clustering has played an important role to cluster dead zone using colour based object area segmentation.
Бесплатно
Статья научная
In this paper we propose a highly accurate method of automatically estimation of the Gaussian noise level in digital images, which is based on image filtering and analysis of the region of interest. Noise level is an important parameter to many digital image processing applications, for example, when removing noise. As the noise level its standard deviation is calculated. The selection of the noise component in an image is performed by high-pass filtration, where the Laplacian difference is used as the filter kernel. Based on the noise component of the image, regions of interest with homogeneous areas of the image are calculated. Region of interest are selected by the iterative method using low-pass filtration, where Gaussian two-dimensional function is used as the filter kernel. The noise level is calculated only in the regions of interest that contain almost no edges and textures, because edges and textures cause errors during the noise level estimation. In order to improve the accuracy of the method, edges of images are detected and out of region of interest. The high accuracy of the proposed method provides the use of high-pass and low-pass filtrations, iterative selection of region of interest and analysis of image edges. The accuracy of the developed method has been tested on the processing of 100 test images with different levels of software added Gaussian noise, as well as the processing of real photos with noise. The proposed method for the noise level estimation can be used for optimal automatic image filtering and for assessing the quality of photosensitive sensors.
Бесплатно
Статья научная
Medical image segmentation is a fundamental task in the medical imaging field. Optimal segmentation is required for the accurate judgment or appropriate clinical diagnosis. In this paper, we proposed automatically gradient threshold estimator of anisotropic diffusion for Meyer’s Watershed algorithm based optimal segmentation. The Meyer’s Watershed algorithm is the most significant for a large number of regions separations but the over segmentation is the major drawback of the Meyer’s Watershed algorithm. We are able to remove over segmentation after using anisotropic diffusion as a preprocessing step of segmentation in the Meyer’s Watershed algorithm. We used a fixed window size for dynamically gradient threshold estimation. The gradient threshold is the most important parameter of the anisotropic diffusion for image smoothing. The proposed method is able to segment medical image accurately because of obtaining the enhancement image. The introducing method demonstrates better performance without loss of any clinical information while preserving edges. Our investigated method is more efficient and effective in order to segment the region of interests in the medical images indeed.
Бесплатно
Automating the Process of Work-Piece Recognition and Location for a Pick-and-Place Robot in a SFMS
Статья научная
This paper reports the development of a vision system to automatically classify work-pieces with respect to their shape and color together with determining their location for manipulation by an in-house developed pick-and-place robot from its work-plane. The vision-based pick-and-place robot has been developed as part of a smart flexible manufacturing system for unloading work-pieces for drilling operations at a drilling workstation from an automatic guided vehicle designed to transport the work-pieces in the manufacturing work-cell. Work-pieces with three different shapes and five different colors are scattered on the work-plane of the robot and manipulated based on the shape and color specification by the user through a graphical user interface. The number of corners and the hue, saturation, and value of the colors are used for shape and color recognition respectively in this work. Due to the distinct nature of the feature vectors for the fifteen work-piece classes, all work-pieces were successfully classified using minimum distance classification during repeated experimentations with work-pieces scattered randomly on the work-plane.
Бесплатно
Autonomous Image Segmentation using Density-Adaptive Dendritic Cell Algorithm
Статья научная
Contemporary image processing based applications like medical diagnosis automation and analysis of satellite imagery include autonomous image segmentation as inevitable facility. The research done shows the efficiency of an adaptive evolutionary algorithm based on immune system dynamics for the task of autonomous image segmentation. The recognition dynamics of immune-kernels modeled with infinite Gaussian mixture models exhibit the capability to automatically determine appropriate number of segments in presence of noise. In addition, the model using representative density-kernel-parameters processes the information with much reduced space requirements. Experiments conducted with synthetic images as well as real images recorded assured convergence and optimal autonomous model estimation. The segmentation results tested in terms of PBM-index values have been found comparable to those of the Fuzzy C-Means (FCM) for the same number of segments as generated by our algorithm.
Бесплатно
Autonomous Multiple Gesture Recognition System for Disabled People
Статья научная
The paper presents an intelligent multi gesture spotting system that can be used by disabled people to easily communicate with machines resulting into easement in day-to-day works. The system makes use of pose estimation for 10 signs used by hearing impaired people to communicate. Pose is extracted on the basis of silhouettes using timed motion history (tMHI) followed by gesture recognition with Hu-Moments. Signs involving motion are recognized with the help of optical flow. Based on the recognized gestures, particular instructions are sent to the robot connected to system resulting into an appropriate action/movement by the robot. The system is unique as it can act as a assisting device and can communicate in local as well as wide area to assist the disabled person.
Бесплатно
Baker Map's Itinerary Based Image Scrambling Method and Its Watermarking Application in DWT Domain
Статья научная
An image scrambling method based on the itinerary of the improved 3D Baker map is proposed in this paper. The standard 3D Baker map is improved by the tent map, so that the itinerary becomes more complicated and can be used to encode the image pixel positions to scramble the image. The scrambling method is applied to the preprocessing in watermarking. The watermark bits are embedded in the discrete wavelet transform (DWT) spectral domain based on the scrambling of watermark, the odd-even adjustment rule and the neighbor mean value. The watermark bits are embedded in medium coefficients in DWT domain of the host image. Experimental results show that the watermarked image looks visually identical to the original one and the watermark can be effectively extracted upon image processing attacks, which demonstrates strong robustness against a variety of attacks.
Бесплатно
Статья научная
Low poly image abstraction is an art form that has gained popularity in recent years, particularly in the digital art community. The process involves simplifying an image by reducing the number of polygons used to represent it while preserving its overall appearance and details. This paper proposes a new approach to low poly image abstraction that utilizes edge-preserved seed points to preserve important details while reducing triangle count. The proposed approach involves six steps. First, the input image is smoothed using an anisotropic diffusion filter. Second, the entropy of each pixel in the smoothed image is computed and stored in an entropy map. Third, seed points for Delaunay triangulation are selected by identifying pixels with maximum entropy values in the entropy map. Fourth, the Delaunay triangulation is generated using the seed points as input. Fifth, colors are assigned to the triangles in the Delaunay triangulation using a color selection module. Finally, the final low poly image is generated by rendering the colored Delaunay triangulation. The effectiveness of the proposed method was evaluated through qualitative and quantitative experiments, comparing its results with other comprehensive methods using a diverse range of images from a benchmark dataset. The results showed that the proposed method outperformed other methods in preserving image details while maintaining low polygon count. Additionally, the proposed method was found to be efficient and capable of producing visually appealing results.
Бесплатно
Bangla Handwritten Character Recognition using Convolutional Neural Network
Статья научная
Handwritten character recognition complexity varies among different languages due to distinct shapes, strokes and number of characters. Numerous works in handwritten character recognition are available for English with respect to other major languages such as Bangla. Existing methods use distinct feature extraction techniques and various classification tools in their recognition schemes. Recently, Convolutional Neural Network (CNN) is found efficient for English handwritten character recognition. In this paper, a CNN based Bangla handwritten character recognition is investigated. The proposed method normalizes the written character images and then employ CNN to classify individual characters. It does not employ any feature extraction method like other related works. 20000 handwritten characters with different shapes and variations are used in this study. The proposed method is shown satisfactory recognition accuracy and outperformed some other prominent exiting methods.
Бесплатно
Bearing Health Assessment Using Time Domain Analysis of Vibration Signal
Статья научная
Objective: Bearing defects are the most frequently occurring fault in any electrical machine. In this perspective, this manuscript proposed a novel statistical time-domain approach utilizing the vibration signal to detect incipient faults of rolling-element bearing used in three-phase induction motor. Methodology: To detect bearing defects, six time-domain features (TDFs) namely Mean Value (µ), Peak, Root Mean Square (RMS), Crest Factor (CRF), Skewness (SKW) and Kurtosis (K) were extracted from the standard database of the vibration signal. The standard databases of vibration signals were taken from the publicly available datacenter website of Case Western Reserve University (CWRU) relating to healthy, inner raceway and ball defects of bearing. Initially, the mean and standard deviation analysis of each considered TDFs of vibration signals were performed to discriminate the health conditions of bearing. Then, the box or whisker plot method was applied to visualize the variation in each TDF in terms of median and interquartile range (IQR) value for better analysis of bearing defects. Finally, a new index parameter termed as bearing fault index (BFIT) was also computed and this parameter predicts the bearing defects based on the mean of all considered TDFs. Results: The results of the “mean±σ” analysis have depicted that all considered TDFs except µ feature are almost independent to operating loads, and have discerning potential to diagnose bearing defects. The computations of these TDFs are mathematically very simple. The box plot representation of TDFs of vibration databases have shown that peak, RMS, and skewness features outperforms to demarcate bearing health conditions in terms of median and IQR value. The results of quantitative analysis of BFIT parameter have shown that if the magnitude of this parameter is higher than 1.8 then bearing is supposed to be faulty at all operating loads of machine. Thus, the BFIT analysis of TDFs is more simple and reliable to discriminate the health conditions of bearing. As most of the available techniques rely on the multi-processing of vibration data that requires large processing time and complicated mathematical model, so the proposed method prove to be simple and reliable in identifying the incipient bearing defects.
Бесплатно
Bearing fault detection using logarithmic wavelet packet transform and support vector machine
Статья научная
Objective: This paper presents an automated approach that combines Fisher ranking and dimensional reduction method as kernel principal component analysis (KPCA) with support vector machine (SVM) to accurately classify the defects of rolling element bearing used in induction motor. Methodology: In this perspective, vibration signal produced by rolling element bearing was decomposed to four levels using wavelet packet decomposition (WPD) method. Thirty one Logarithmic Root Mean Square Features (LRMSF) were extracted from four level decomposed vibration signals. Initially, thirty one features were rank by Fisher score and top ten rank features were selected. For effective detection, top ten features were reduced to a new feature using dimension reduction methods as KPCA and generalized discriminant analysis (GDA). After this, the new feature applied to SVM for binary classification of bearing defects. For analysis of this thirty six standard vibration datasets taken from online available bearing data center website of Case Western Reserve University on bearing conditions like healthy (NF), inner race defect (IR) and ball bearing (BB) defects at different loads. Results: The simulated numerical results show that proposed method KPCA with SVM classifier using Gaussian Kernel achieved an accuracy (AC) of 100, Sensitivity (SE) of 100%, Specificity (SP) of 99.3% and Positive prediction value (PPV) of 99.3% for NF_IRB dataset, and an AC of 100, SE of 99.8%, SP of 100% and PPV of 100% for NF_BBB dataset.
Бесплатно
Binary Segmentation Dataset Distances for Transfer Learning
Статья научная
This work is devoted to developing a novel transfer learning approach for solving binary semantic segmentation problems that often arise on short samples in the medical (segmentation of nodules in lungs, tumors, polyps, etc.) and other domains. The goal is to optimally select the most suitable dataset from a different subject area with similar feature space and distribution to the target data. Examples show that a severe disadvantage of transfer learning is the difficulty of selecting an initial training sample for pre-training a neural network. In this paper, we propose metrics for calculating the distance between binary segmentation datasets, allowing us to select the optimal initial training set for transfer learning. These metrics are based on the geometric distances estimation of the dataset using optimal transport, Wasserstein distance for Gaussian mixture models, clustering, and their hybrids. Experiments on datasets of medical segmentation Decathlon, LIDC, and a private dataset of tuberculomas in the lungs are presented, proving a statistically strict correlation of these metrics with a relative increase in segmentation accuracy during transfer learning.
Бесплатно
