This work is focused on detection of loss in nerve fibre layer on colour pictures of retina, witch are makes by fundus camera. It describe every simple objects of retina, optic nerve head, macula lutea and vascular bed. It detect optic nerve head and his near area, witch is general for detection of breakdownds. It use several metodes of picture adjusting for picture elaboration and objects detection (segmentation, thresholding, enhancement, hough transformation ). The detection of loss in nerve fibre layer is based on comparing of statistic parameters ( average, standart deviation, skewness coefficient and kurtosis coefficient histogram, entropy ) in choosed areas with and withou destruction of nerve layers. Vascular bed have badwatsh on results, cause of this we using hand choosing of essay.

Road safety is a serious problem in many countries and affects the lives of many people. Improving road safety starts with the drivers, and the best way to make them change their habits is to offer incentives for better, safer driving styles. This project aims to make that possible by offering a means to calculate a quantified indicator of how safe a driver's habits are. This is done by developing an on-board, visual road-sign recognition system that can be coupled with a vehicle tracking system to determine how often a driver violates the rules of the road. The system detects stop signs, red traffic lights and speed limit signs, and outputs this data in a format that can be read by a vehicle tracking system, where it can be combined with speed information and sent to a central database where the driver safety rating can be calculated. Input to the system comes from a simple, standard dashboard mounted camera within the vehicle, which generates a continuous stream of images of the scene directly in front of the vehicle. The images are subjected to a number of cascaded detection sub-systems to determine if any of the target objects (road signs) appear within that video frame. The detection system software had to be optimized for minimum false positive detections, since those will unfairly punish the driver, and it also had to be optimized for speed to run on small hardware that can be installed in the vehicle. The first stage of the cascaded system consists of an image detector that detects circles within the image, since traffic lights and speed signs are circular and a stop sign can be approximated by a circle when the image is blurred or the resolution is lowered. The second stage is a neural network that is trained to recognize the target road sign in order to determine which road sign was found, or to eliminate other circular objects found in the image frame. The output of the neural network is then sent through an iterative filter with a majority voted output to eliminate detection 'jitter' and the occasional incorrect classifier output. Object tracking is applied to the 'good' detection outputs and used as an additional input for the detection phase on the next frame. In this way the continuity and robustness of the image detector are improved, since the object tracker indicates to it where the target object is most likely to appear in the next frame, based on the track it has been following through previous frames. In the final stage the detection system output is written to the chosen pins of the hardware output port, from where the detection output can be indicated to the user and also used as an input to the vehicle tracking system. To find the best detection approach, some methods found in literature were studied and the most likely candidates compared. The scale invariant feature transform (SIFT) and speeded up robust features (SURF) algorithms are too slow compared to the cascaded approach to be used for real-time detection on an in-vehicle hardware platform. In the cascaded approach used, different detection stage algorithms are tested and compared. The Hough circle transform is measured against blob detection on stop signs and speed limit signs. On traffic light state detection two approaches are tested and compared, one based on colour information and the other on direct neural network classification. To run the software in the user's vehicle, an appropriate hardware platform is chosen. A number of promising hardware platforms were studied and their specifications compared before the best candidate was selected and purchased for the project. The developed software was tested on the selected hardware in a vehicle during real public road driving for extended periods and under various conditions. / Dissertation (MEng)--University of Pretoria, 2017. / Electrical, Electronic and Computer Engineering / MEng / Unrestricted

Autonome Roboter müssen sich in einer dynamischen Umwelt zurechtfinden können und zeitnah auf Ereignisse reagieren. Der echtzeitfähigen Verarbeitung visueller Information kommt dabei eine Schlüsselrolle zu. Am Beispiel zweier spezieller Roboter wird die Gewinnung steuerungsrelevanter Daten aus einer Bildsequenz demonstriert. Dazu werden in Form von Modellen umfassende Voraussetzungen aufgestellt von dem, was gesehen werden soll und kann. Es wird ein Erkennungssystem entworfen und implementiert, welches die Extraktion der gesuchten Daten aus den Bilddaten u.a. mittels Bildkorrelation im Frequenzraum und Hough-Transformation durchführt. Die gewonnenen Daten werden zur Steuerung eines der Roboter verwendet.

I modern industrimiljö finns fortfarande en stor mängd analoga visarinstrument. Det är önskvärt att övervaka dessa instrument digitalt vilket medför att kontroll av mätdata kan göras utan att personal behöver vara på plats. På marknaden finns idag ingen aplikation som är utvecklad för att uppfylla denna funktion.  Detta examensarbete har undersökt metoder för hur en analog visares värde ska läsas av digitalt och utvecklat en prototyp som kan utföra uppgiften.  Prototypen utvecklades med hjälp av datorseende algoritmer för att läsa av den analoga visarens värde. Algoritmerna för datorseende implementerades på en Raspberry Pi4 Model B och en kamera, Rasperry Pi Kameramodul V2. Prototypen som utvecklades uppfyller de funktioner som efterfrågades, och uppnåde en noggranhet på 0.97% +- 0.75 av det procentuella uppmätta värdet hos en analog visares fulla mätspann med en upplösning på 2.5%

No description available.

Real-time multi-circle detection has been a challenging problem in the field of biomedical image processing, due to the variable sizes and non-ideal shapes of cells in microscopic images. In this study, two new multi-circle detection algorithms are developed to facilitate an automatic bladder cancer diagnosis system: one is a modified circular Hough Transform algorithm integrated with edge gradient information; and the other one is a stochastic search approach based on real valued artificial immune systems. Computer simulation results show both algorithms outperform traditional methods such as the Hough Transform and the geometric feature based method, in terms of both precision and speed.

Feature extraction from a tensor based local image representation introduced by Knutsson in [37] is discussed. The tensor representation keeps statements of structure, certainty of statement and energy separate. Further processing for obtaining new features also having these three entities separate is achieved by the use of a new concept, tensor field filtering. Tensor filters for smoothing and for extraction of circular symmetries are presented and discussed in particular. These methods are used for corner detection and extraction of more global features such as lines in images. A novel method for grouping local orientation estimates into global line parameters is introduced. The method is based on a new parameter space, the Möbius Strip parameter space, which has similarities to the Hough transform. A local centroid clustering algorithm is used for classification in this space. The procedure automatically divides curves into line segments with appropriate lengths depending on the curvature. A linked list structure is built up for storing data in an efficient way. / <p>Ogiltigt nummer / annan version: I publ. nr 290:s ISBN: 91-7870-815-X.</p>

No description available.

No description available.

License plate recognition/identification (LPR/I) applies image processing and character recognition technology to identify vehicles by automatically reading their license plates. In the United States, however, each state has its own standard-issue plates, plus several optional styles, which are referred to as special license plates or varieties. There is a clear absence of standardization and multi-colored, complex backgrounds are becoming more frequent in license plates. Commercially available optical character recognition (OCR) systems generally fail when confronted with textured or poorly contrasted backgrounds, therefore creating the need for proper image segmentation prior to classification. The image segmentation problem in LPR is examined in two stages: license plate region detection and license plate character extraction from background. Three different approaches for license plate detection in a scene are presented: region distance from eigenspace, border location by edge detection and the Hough transform, and text detection by spectral analysis. The experiments for character segmentation involve the RGB, HSV/HSI and 1976 CIE L*a*b* color spaces as well as their Karhunen-Loéve transforms. The segmentation techniques applied include multivariate hierarchical agglomerative clustering and minimum-variance color quantization. The trade-off between accuracy and computational expense is used to select a final reliable algorithm for license plate detection and character segmentation. The spectral analysis approach together with the K-L L*a*b* transformed color quantization are found experimentally as the best alternatives for the two identified image segmentation stages for US license plate recognition. / Master of Science