Analysis of the shape and location of abrasive grain tips as well as their changes during the grinding process, is the basis for forecasting the machining process results. This paper presents a methodology of using the watershed segmentation in identifying abrasive grains on the abrasive tool active surface. Some abrasive grain tips were selected to minimize the errors of detecting many tips on a single abrasive grain. The abrasive grains, singled out as a result of the watershed segmentation, were then analyzed to determine their geometric parameters. Moreover, the statistical parameters describing their locations on the abrasive tool active surface and the parameters characterizing intergranular spaces were determined.
The article presents a method for 3D point cloud segmentation. The point cloud comes from a FARO LS scanner – the device creates a dense point cloud, where 3D points are organized in the 2D table. The input data set consists of millions of 3D points – it makes widely known RANSAC algorithms unusable. We add some modifi cations to use RANSAC for such big data sets.
A phoneme segmentation method based on the analysis of discrete wavelet transform spectra is described. The localization of phoneme boundaries is particularly useful in speech recognition. It enables one to use more accurate acoustic models since the length of phonemes provide more information for parametrization. Our method relies on the values of power envelopes and their first derivatives for six frequency subbands. Specific scenarios that are typical for phoneme boundaries are searched for. Discrete times with such events are noted and graded using a distribution-like event function, which represent the change of the energy distribution in the frequency domain. The exact definition of this method is described in the paper. The final decision on localization of boundaries is taken by analysis of the event function. Boundaries are, therefore, extracted using information from all subbands. The method was developed on a small set of Polish hand segmented words and tested on another large corpus containing 16 425 utterances. A recall and precision measure specifically designed to measure the quality of speech segmentation was adapted by using fuzzy sets. From this, results with F-score equal to 72.49% were obtained.
Laser triangulation is one of the machine vision measurement methods most commonly used in 3D quality control. However, considering its susceptibility to interference, it cannot be used in certain areas of industrial production e.g. very shiny surfaces. Thus, for the improvement of its applicability, a predictive algorithm of light profile segmentation was designed, where - as a result of using a'priori knowledge - the method becomes resistant to secondary reflexes.
The developed technique has been tested on selected parts with surfaces typical for the machine-building industry. The evaluation has been presented based on the surface representation (mapping) error analysis, using the difference between the obtained cloud of points and the nominal surface as processing data, as well as scatter of the discrete Gauss curvature.
The tendencies of modern industry are to increase the quality of manufactured products, simultaneously decreasing production time and cost. The hybrid system combines advantages of the high accuracy of contact CMM and the high measurement speed of non-contact structured light optical techniques. The article describes elements of a developed system together with the steps of the measurement process of the hybrid system, with emphasis on segmentation algorithms. Additionally, accuracy determination of such a system realized with the help of a specially designed ball-plate measurement standard is presented.
This paper presents the improved version of the classification system for supporting glaucoma diagnosis in ophthalmology. In this
paper we propose the new segmentation step based on the support vector clustering algorithm which enables better classification performance.
A vocal tract model based on a digital waveguide is presented in which the vocal tract has been decomposed into uniform cylindrical segments of variable lengths. We present a model for the real-time numerical solution of the digital waveguide equations in a uniform tube with the temporally varying cross section. In the current work, the uniform cylindrical segments of the vocal tract may have their different lengths, the time taken by the sound wave to propagate through a cylindrical segment in an axial direction may not be an integer multiple of each other. In such a case, the delay in an axial direction is necessarily a fractional delay. For the approximation of fractional-delay filters, Lagrange interpolation is used in the current model. Variable length of the individual segment of the vocal tract enables the model to produce realistic results. These results are validated with accurate benchmark model. The proposed model has been devised to elongate or shorten any arbitrary cylindrical segment by a suitable scaling factor. This model has a single algorithm and there is no need to make section of segments for elongation or shortening of the intermediate segments. The proposed model is about 23% more efficient than the previous model.
Image segmentation is a typical operation in many image analysis and computer vision applications. However, hyperspectral image segmentation is a field which have not been fully investigated. In this study an analogue- digital image segmentation technique is presented. The system uses an acousto-optic tuneable filter, and a CCD camera to capture hyperspectral images that are stored in a digital grey scale format. The data set was built considering several objects with remarkable differences in the reflectance and brightness components. In addition, the work presents a semi-supervised segmentation technique to deal with the complex problem of hyperspectral image segmentation, with its corresponding quantitative and qualitative evaluation. Particularly, the developed acousto-optic system is capable to acquire 120 frames through the whole visible light spectrum. Moreover, the analysis of the spectral images of a given object enables its segmentation using a simple subtraction operation. Experimental results showed that it is possible to segment any region of interest with a good performance rate by using the proposed analogue-digital segmentation technique.
This paper presents a deep learning-based image texture recognition system. The methodology taken in this solution is formed in a bottom-up manner. It means we swipe a moving window through the image in order to categorize if a given region belongs to one of the classes seen in the training process. This categorization is done based on the Deep Neural Network (DNN) of fixed architecture. The training process is fully automated regarding the training data preparation, investigation of the best training algorithm, and its hyper-parameters. The only human input to the system is the definition of the categories for further recognition and generation of the samples (region markings) in the external application chosen by the user. The system is tested on road surface images where its task is to categorize image regions to a different road category (e.g. curb, road surface damage, etc.) and is featured with 90% and above accuracy.
This paper presents signal processing aspects for automatic segmentation of retinal layers of the human eye. The paper draws attention to the problems that occur during the computer image processing of images obtained with the use of the Spectral Domain Optical Coherence Tomography (SD OCT). Accuracy of the retinal layer segmentation for a set of typical 3D scans with a rather low quality was shown. Some possible ways to improve quality of the final results are pointed out. The experimental studies were performed using the so-called B-scans obtained with the OCT Copernicus HR device.
Minimally invasive procedures for the kidney tumour removal require a 3D visualization of topological relations between kidney, cancer, the pelvicalyceal system and the renal vascular tree. In this paper, a novel methodology of the pelvicalyceal system segmentation is presented. It consists of four following steps: ROI designation, automatic threshold calculation for binarization (approximation of the histogram image data with three exponential functions), automatic extraction of the pelvicalyceal system parts and segmentation by the Locally Adaptive Region Growing algorithm. The proposed method was applied successfully on the Computed Tomography database consisting of 48 kidneys both healthy and cancer affected. The quantitative evaluation (comparison to manual segmentation) and visual assessment proved its effectiveness. The Dice Coefficient of Similarity is equal to 0.871 ± 0.060 and the average Hausdorff distance 0.46 ± 0.36 mm. Additionally, to provide a reliable assessment of the proposed method, it was compared with three other methods. The proposed method is robust regardless of the image acquisition mode, spatial resolution and range of image values. The same framework may be applied to further medical applications beyond preoperative planning for partial nephrectomy enabling to visually assess and to measure the pelvicalyceal system by medical doctors.
Cardiovascular system diseases are the major causes of mortality in the world. The most important and widely used tool for assessing the heart state is echocardiography (also abbreviated as ECHO). ECHO images are used e.g. for location of any damage of heart tissues, in calculation of cardiac tissue displacement at any arbitrary point and to derive useful heart parameters like size and shape, cardiac output, ejection fraction, pumping capacity. In this paper, a robust algorithm for heart shape estimation (segmentation) in ECHO images is proposed. It is based on the recently introduced variant of the level set method called level set without edges. This variant takes advantage of the intensity value of area information instead of module of gradient which is typically used. Such approach guarantees stability and correctness of algorithm working on the border between object and background with small absolute value of image gradient. To reassure meaningful results, the image segmentation is proceeded with automatic Region of Interest (ROI) calculation. The main idea of ROI calculations is to receive a triangle-like part of the acquired ECHO image, using linear Hough transform, thresholding and simple mathematics. Additionally, in order to improve the images quality, an anisotropic diffusion filter, before ROI calculation, was used. The proposed method has been tested on real echocardiographic image sequences. Derived results confirm the effectiveness of the presented method.
With development of medical diagnostic and imaging techniques the sparing surgeries are facilitated. Renal cancer is one of examples. In order to minimize the amount of healthy kidney removed during the treatment procedure, it is essential to design a system that provides three-dimensional visualization prior to the surgery. The information about location of crucial structures (e.g. kidney, renal ureter and arteries) and their mutual spatial arrangement should be delivered to the operator. The introduction of such a system meets both the requirements and expectations of oncological surgeons. In this paper, we present one of the most important steps towards building such a system: a new approach to kidney segmentation from Computed Tomography data. The segmentation is based on the Active Contour Method using the Level Set (LS) framework. During the segmentation process the energy functional describing an image is the subject to minimize. The functional proposed in this paper consists of four terms. In contrast to the original approach containing solely the region and boundary terms, the ellipsoidal shape constraint was also introduced. This additional limitation imposed on evolution of the function prevents from leakage to undesired regions. The proposed methodology was tested on 10 Computed Tomography scans from patients diagnosed with renal cancer. The database contained the results of studies performed in several medical centers and on different devices. The average effectiveness of the proposed solution regarding the Dice Coefficient and average Hausdorff distance was equal to 0.862 and 2.37 mm, respectively. Both the qualitative and quantitative evaluations confirm effectiveness of the proposed solution.
The study of the different engineering materials according to their mechanical and dynamic characteristics has become an area of research interest in recent years. Several studies have verified that the mechanical properties of the material are directly affected by the distribution and size of the particles that compose it. Such is the case of asphalt mixtures. For this reason, different digital tools have been developed in order to be able to detect the structural components of the elements in a precise, clear and efficient manner. In this work, a segmentation model is developed for different types of dense-graded asphalt mixtures with grain sizes from 9.5 mm to 0.0075 mm, using sieve size reconstruction of the laboratory production curve. The laboratory curve is used to validate the particles detection model that uses morphological operations for elements separation. All this with the objective of developing a versatile tool for the analysis and study of pavement structures in a non-destructive test. The results show that the model presented in this work is able to segment elements with an area greater than 0.0324 mm2 and reproduce the sieve size curves of the mixtures with a high percentage of precision.
The mechanical characteristics of the railway superstructure are related to the properties of the ballast, and especially to the particle size distribution of its grains. Under the constant stress-strain of carriages, the ballast can deteriorate over time, and consequently it should properly be monitored for safety reasons. The equipment which currently monitors the railway superstructure (like the Italian diagnostic train Archimede) do not make any “quantitative” evaluation of the ballast. The aim of this paper is therefore to propose a new methodology for extracting railway ballast particle size distribution by means of the image processing technique. The procedure has been tested on a regularly operating Italian railway line and the results have been compared with those obtained from laboratory experiments, thus assessing how effective is the methodology which could potentially be implemented also in diagnostic trains in the near future.
In this paper methods and their examination results for automatic segmentation and parameterization of vessels based on spectral domain optical coherence tomography (SD-OCT) of the retina are presented. We present three strategies for morphologic image processing of a fundus image reconstructed from OCT scans. A specificity of initial image processing for fundus reconstruction is analysed. Then, the parameterization step is performed based on the vessels segmented with the proposed algorithm. The influence of various methods on the vessel segmentation and fully automatic vessel measurement is analysed. Experiments were carried out with a set of 3D OCT scans obtained from 24 eyes (12 healthy volunteers) with the use of an Avanti RTvue OCT device. The results of automatic vessel segmentation were numerically compared with those prepared manually by the medical doctor experts.