The advantages that a side-scan sonar can offer include large-scale survey areas and high-resolution imagery which can provide the detection and positioning of underwater targets effectively. The purpose of image analysis, classification and positioning in this research was presented by the development of an automated recognition and classification system based on sonographs collected off Kenting area. Major components of the system include gray level co-occurrence matrix method, Baysian classification and cluster analysis. The sonograph classified by the automated recognition and classification system was split into two stages. The first stage divided the seafloor into three categories: (1) Rocky seafloor. (2) Sandy seafloor. (3) Acoustic shadow seafloor. Based on the characteristics of the rocky seafloor, the rocky seafloor was subdivided into five types in the second stage: (1) Flank reef and small independent reef. (2) Smooth reef. (3) Small coral on reef. (4) Coral on independent reef. (5) Large coral on reef. Analysis and proof of the system was conducted by underwater photographs collected off Kenting area in August 4, to 6, 2004. The identification accuracy of the first stage can reach 93% in Shiniuzai area. The characteristic features selected in this research (i.e., entropy and homogeneity) for the classification of various coral reef seafloors was proved adequate and the results was described in map within a Geographic Information System in the second stage. The results of this research illustrated that the rocky area identified in Shiniuzai was 98,863 m2. Due to image resolution restrictions, only 62,199 m2 of the total rocky area could be defined and classified properly. Among them, the flank reef and small independent reef covered an area of 15,954 m2 (26.3%); the smooth reef covered 3,133 m2 (5.0%); the small coral on reef covered 8,021 m2 (12.8%); the coral on independent reef covered 25,504 m2 (40.7%) and the large coral on reef covered 9,587 m2 (15.3%). Key words:side scan sonar,coral reef,gray level co-occurrence matrix

Gray level co-occurrence matrix has proven to be a powerful basis for use in texture classification. Various textural parameters calculated from the gray level co-occurrence matrix help understand the details about the overall image content. The aim of this research is to investigate the use of the gray level co-occurrence matrix technique as an absolute image quality metric. The underlying hypothesis is that image quality can be determined by a comparative process in which a sequence of images is compared to each other to determine the point of diminishing returns. An attempt is made to study whether the curve of image textural features versus image memory sizes can be used to decide the optimal image size. The approach used digitized images that were stored at several levels of compression. GLCM proves to be a good discriminator in studying different images however no such claim can be made for image quality. Hence the search for the best image quality metric continues.

Abstract The interpretation and identification of information from the side-scan sonar imagery are mainly depended on visual observation and personal experiences. Recent studies tended to increase the identification efficiency by using numerical analysis methods. This can reduce the error that cause by the differences of observer¡¦s experience as well as by extended time observation. The position around the center line of the slant range corrected side-scan sonar imagery might result in the degradation of the ability of numerical methods to successfully detect artificial objects. Theoretically, this problem could be solved by using a specific characteristic function to identify the existence of concrete reefs, and then filtering the noise of the central line area with a threshold value. This study was intended to develop fully automatic sonar imagery processing system for the identification of cubic concrete and cross-type protective artificial reefs in Taiwan offshore area. The procedures of the automatic sonar imagery processing system are as follows: (1) Image Acquisition¡G500kHz with slant range of 75m. (2) Feature Extraction¡Ggrey level co-occurrence matrix (i.e., Entropy, Homogeneity and Mean) (3) Classification¡Gunsupervised Bayesian classifier. (4) Object Identification¡Gby characteristic feature (i.e., Entropy). (5) Object¡¦s Status Analysis¡Gobject¡¦s circumference¡Barea¡Bcenter of mass and quantity. This study used the sonar images collected at Chey-Ding artificial reef site in Kaohsiung City as a case study, aiming to verify the automatic sonar imagery processing system and find out the optimum window size. The image characteristic functions include one set of first order parameter (i.e., mean) and two sets of second order parameter (i.e., entropy and homogeneity). Eight(8) sonar images with 1-8 sets of cubic concrete and cross-type protective artificial reefs where used in this step. The identification efficiency of the system, in terms of the produce¡¦s accuracy, is 79.41%. The results illustrated that there were 16~28 sets of artificial reefs being detected in this case which is comparable with the actual amount of 17 sets. Based on this investigation, the optimum window size was concluded to be 12¡Ñ12 pixels with sliding size of 4 pixel. Imagery collected at Fang-Liau artificial reef site of Pingtung County was tested. For the purpose of applicability, the original imagery (2048¡Ñ2800 pixels) was divided into 8 consecutive smaller sized imagery frames(2048¡Ñ350 pixels). The influence of using a two-fold classification procedure and a central filtering method to reduce the noise that caused by slant range correction were discussed. The results showed that central line filtering method is applicable. The results of object¡¦s status analysis showed that there are 156-236 sets of reefs existed. Automatic determination of the target using the characteristic function of entropy is feasible. If the value is larger than 1.45, it represents positive identification of concrete artificial reefs. It can be classified as muddy sand seabed type if the value is smaller than 1.35. If the value is between 1.35~1.45, it illustrates the existence of a transition zone where objects of smaller in dimensions might exist. To achieve the purpose of automatic operation, firstly, we have to identify the existence of the concrete reefs by using the specific characteristic function. Based on the result of existing concrete reefs, suture line filtering method will hence be used to filter the noise from the image information. For that all of the procedures are automatically operated without human intervention. Key word: side-scan sonar ; characteristic function ; gray level co-occurrence matrix ; Bayesian classification ;entropy ; homogeneity ; mean

The normal procedure for the detection of underwater stationary targets is mainly by the application of side-scan sonar. In addition, the identification of targets within the side-scan sonar imagery is primarily based on the visual observation of the operator. Due to its complexity and poor effectiveness, the visual observation procedure was gradually been substituted by numerical analysis procedures and programs. The purpose of the current investigation was dedicated to the development of an automatic image analysis program for the detection and identification of cubic concrete artificial reefs (2 m x 2 m x 2m) in the south-western coastal area off Taiwan. The major components and methodologies of the program include: (1)Image acquisition; side-scan sonar at 500 kHz and slant range at 75 m. (2)Feature extraction; grey level co-occurrence matrix. (3)Feature Classification; unsupervised Bayesian classifier. (4)Target identification; cluster analysis. (5)Target properties analysis, includes circumference, area, central coordinates and quantity of the targets. Program verification and optimal parameters determination were conducted with a sonograph (650 ¡Ñ 650 pixels) acquired at the Chey-Ding artificial reef site off Kaohsiung County. Feature functions employed in this program include entropy, homogeneity, and mean value. The identification accuracy can reach 93% at the most. In addition, the number of artificial reefs estimated by the program was within 9 to 20, while the actual number is 15. A realistic evaluation of this program was conducted with a sonograph (2048 ¡Ñ 6050 pixels) acquired at Fang-Liau artificial reef site off Pyngdong County. In addition to the cubic reefs, the targets at this site include cross-shaped artificial reefs with dimensions less than the cubic reefs. The sonograph was divided into smaller blocks with dimensions of 2048 x 550 pixels during evaluation. The results showed that each block can be evaluated based on the value of the seed point obtained by cluster analysis. The seed point which fells between 20.6 and 24.4 indicates that there are cubic reefs existed. Between 15.3 and 17.4 indicates that there are targets with smaller dimensions (i.e., crossed reefs) existed which can not be identified properly. Between 10.1 and 10.9, there is no target existed on the seafloor. The results indicated that the number of targets identified is between 122 and 240. According to the results of this investigation, the automatic image analysis program can improve the detection and identification of stationary targets within side-scan sonar imagery.

Confocal microendoscopy is a promising new diagnostic imaging technique that is minimally invasive and provides in-vivo cellular-level images of tissue. In this study, we developed various image analysis techniques for ovarian cancer detection using the confocal microendoscope system. Firstly, we developed a technique for automatic classification of images based on focus, to prune out the out-of-focus images from the ovarian dataset. Secondly, we modified the texture analysis technique developed earlier to improve the stability of the textural features. The modified technique gives stable features and more consistent performance for ovarian cancer detection. Although confocal microendoscopy provides cellular-level resolution, it is limited by a small field of view. We present a fast technique for stitching the individual frames of the tissue to form a large mosaic. Such a mosaic will aid the physician in diagnosis, and also makes quantitative and statistical analysis possible on a larger field of view.

In this dissertation, new digital image processing methods for hyperdimensional imagery are developed and experimentally tested on remotely sensed Earth observations and medical imagery. The high dimensionality of the imagery is either inherent due to the type of measurements forming the image, as with imagery obtained with hyperspectral sensors, or the result of preprocessing and feature extraction, as with synthetic aperture radar imagery and digital mammography. In the first study, two omni-directional adaptations of gray level co-occurrence matrix analysis are developed and experimentally evaluated. The adaptations are based on a previously developed rubber band straightening transform that has been used for analysis of segmented masses in digital mammograms. The new methods are beneficial because they can be applied to imagery where the region of interest is either poorly segmented or not segmented. The methods are based on the concept of extracting circular windows s around each pixel in the image which are radially resampled to derive rectangular images. The images derived from the resampling are then suitable for standard GLCM techniques. The methods were applied to both remotely sensed synthetic aperture radar imagery, for the purpose of automated detection of landslides on earthen levees, and to digital mammograms, for the purpose of automated classification of masses as either benign or malignant. Experimental results show the newly developed methods to be valuable for texture feature extraction and classification of un-segmented objects. In the second study, a new technique of using spatial information in spectral band grouping for remotely sensed hyperspectral imagery is developed and experimentally tested. The technique involves clustering the spectral bands based on similarity of spatial features extracted from each band. The newly developed technique is evaluated in automated classification systems that utilize a single classifier and systems that utilize multiple classifiers combined with decision fusion. The systems are experimentally tested on remotely sensed imagery for agricultural applications. The spatial-spectral band grouping approach is compared to uniform band windowing and spectral only band grouping. The results show that the spatial-spectral band grouping method significantly outperforms both of the comparison methods, particularly when using multiple classifiers with decision fusion.

As technology advances in all areas of society and industry, the technology used to produce one of life's essentials - food - is also improving. The majority of agriculture production in developed countries has gone from family farms to industrial operations. With the advent of large-scale farming, the automation of basic farming operations has increasingly made practical and economic sense. Broccoli, which is still harvested by hand, is one of the most expensive crops to produce. Investing in sensing technology that can provide detailed information about the location, maturity and viability of broccoli heads has the potential to produce great commercial benefits. This technology is also a prerequisite for developing an autonomous harvester that could select and harvest mature heads of broccoli. This thesis details the work done to develop a computer vision algorithm that has the ability to locate the broccoli head within an image of an entire broccoli plant and to distinguish between mature and immature broccoli heads. Locating the head involves the use of a Hough transform to find the leaf stems and, once the stems are found, the location and extent of the broccoli head can be ascertained with the use of contrast texture analysis at the intersection of the stems. A co-occurrence matrix is then produced of the head and statistical texture analysis is performed to determine the maturity of the broccoli head. The conceptual design of a selective autonomous broccoli harvester, as well as suggestions for further research, is also presented. / Master of Science

Thesis (MScEng)--Stellenbosch University, 2014. / ENGLISH ABSTRACT: Hydrocyclones are stationary separating machines that separate materials based on centrifugal separation and are widely used in chemical engineering and mineral processing industries. Their design and operation, compact structure, low running costs and versatility all contribute to their applications in liquid clarification, slurry thickening, solid washing and classification. With any of these operations, the overall profitability of the process relies on the effective control of the process equipment. However, in practice, hydrocyclones are difficult to monitor and control, owing to the complexity and difficulty in measuring internal flows in the equipment. Several studies have indicated that hydrocyclone underflow images can be used to monitor process conditions. The research described in this thesis considers the use of image analysis to monitor particle size and solids concentration in the underflow discharge of a hydrocyclone. The experimental work consisted of laboratory and industrial-based case studies. The laboratory cyclone used was a 76 mm general laboratory cyclone. A Canon EOS 400D digital camera was used for the underflow imaging. Image features such as pixel intensity values, underflow discharge width and grey level co-occurrence matrix (GLCM) were extracted from the images using MATLAB Toolbox software. Linear discriminant analysis (LDA) and neural network (NN) classification models were used to discriminate between different PGM ore types based on features extracted from the underflow of the hydrocyclone. Likewise, multiple linear regression and neural network models were used to estimate the underflow solids content and mean particle size in the hydrocyclone underflow. The LDA model could predict the PGM ore types with 61% reliability, while the NN model could do so with a statistically similar 62% reliability. The multiple linear regression models could explain 56% and 40% of variance in the mean particle size and solids content respectively. In contrast, the neural network model could explain 67% and 45% of the variance of the mean particle size and solids content respectively. For the industrial system, a 100% correct classification was achieved with all methods. However, these results are regarded as unreliable, owing to the insufficient data used in the models. / AFRIKAANSE OPSOMMING: Hidrosiklone is stasionêre skeidingsmasjiene wat materiale skei op grond van sentrifugale skeiding en word algemeen gebruik in die chemiese ingenieurswese en mineraalprosessering industrieë. Hul ontwerp en werking, kompakte struktuur, lae bedryfskoste en veelsydigheid dra by tot hul gebruik vir toepassings in vloeistofsuiwering, slykverdikking, vastestof wassing en klassifikasie. In enige van hierdie prosesse hang die oorhoofse winsgewendheid van die proses af van die effektiewe beheer van die prosestoerusting. In die praktyk is hidrosiklone egter moeilik om te monitor en beheer weens die kompleksiteit en moeilikheidsgraad daarvan om die interne vloei in die apparaat te meet. Verskeie studies het aangedui dat hidrosikloon ondervloeibeelde gebruik kan word om die proseskondisies te monitor. Die navorsing beskryf in hierdie tesis maak gebruik van beeldanalise moniteringstegnieke om die ertstipes en grootte- verspreidingsgebiede/ klasse van die ondervloei afvoerpartikels te bepaal. Sodoende word ‘n grondslag gelê vir verbeterde sikloon monitering en beheer. Die eksperimentele werk het bestaan uit beide laboratorium en industrieel-gebaseerde studies. Die laboratorium sikloon wat gebruik is, was ‘n 76 mm algemene laboratorium sikloon. ‘n Canon EOS 400D digitale kamera is gebruik om die hidrosikloon ondervloei beelde vas te vang. Beeldeienskappe soos beeldelement intensiteitswaardes, ondervloei afvoerwydte en grysvlak mede-voorkoms matriks is onttrek uit die beelde deur gebruik te maak van MATLAB Toolbox sagteware. Lineêre diskriminantanalise (LDA) en neural netwerk (NN) klassifikasiemodelle is gebou om te onderskei tussen die verskillende PGM ertse en gebaseer op veranderlikes wat afgelei is uit beelde van die ondervloei van die sikloon. Net so is daar ook gebruik gemaak van lineêre regressie- en neural netwerkmodelle om die vasestofkonsentrasie en gemiddelde partikelgrootte in die ondervloei van die sikloon te beraam. Die LDA model kon die PGM ertstipes met 61% betroubaarheid voorspel, terwyl die neural netwerkmodel dit kon doen met statisties dieselfde betroubaarheid van 62%. Die lineêre regressiemodelle kon onderskeidelik 56% en 40% van die variansie in die gemiddelde partikelgrootte en vastestofkonsentrasie verduidelik. In teenstelling iermee, kon die neurale netwerkmodel 67% en 45% van die variansie in die gemiddelde partikelgrootte en vastestofkonsentrasie verduidelik. In die nywerheidstelsel kon beide tipe modelle perfekte onderskeid tref tussen die partikelgroottes wat gemeet is op opeenvolgende dae van die bedryf van die siklone. Hierdie resultate is egter nie betroubaar nie, a.g.v. die beperkte hoeveelheid data wat beskikbaar was vir modellering.

Mapping ice and open water in ocean bodies is important for numerous purposes including environmental analysis and ship navigation. The Canadian Ice Service (CIS) currently has several expert ice analysts manually generate ice maps on a daily basis. The CIS would like to augment their current process with an automated ice-water discrimination algorithm capable of operating on dual-pol synthetic aperture radar (SAR) images produced by RADARSAT-2. Automated methods can provide mappings in larger volumes, with more consistency, and in finer resolutions that are otherwise impractical to generate. We have developed such an automated ice-water discrimination system called MAGIC. The algorithm first classifies the HV scene using the glocal method, a hierarchical region-based classification method. The glocal method incorporates spatial context information into the classification model using a modified watershed segmentation and a previously developed MRF classification algorithm called IRGS. Second, a pixel-based support vector machine (SVM) using a nonlinear RBF kernel classification is performed exploiting SAR grey-level co-occurrence matrix (GLCM) texture and backscatter features. Finally, the IRGS and SVM classification results are combined using the IRGS approach but with a modified energy function to accommodate the SVM pixel-based information. The combined classifier was tested on 61 ground truthed dual-pol RADARSAT-2 scenes of the Beaufort Sea containing a variety of ice types and water patterns across melt, summer, and freeze-up periods. The average leave-one-out classification accuracy with respect to these ground truths is 95.8% and MAGIC attains an accuracy of 90% or above on 88% of the scenes. The MAGIC system is now under consideration by CIS for operational use.

Η χρήση των δεδομένων Ραντάρ Συνθετικού Ανοίγματος (SAR) σε εφαρμογές απομακρυσμένης παρακολούθησης της Γης έχει ήδη αρχίσει να πρωταγωνιστεί τις τελευταίες δεκαετίες. Τα συστήματα SAR με δυνατότητες μεταξύ άλλων συνεχούς λειτουργίας παντός καιρού, ημέρα και νύχτα, προσφέροντας μεγάλη κάλυψη εδάφους και με δυνατότητα λήψης απεικονίσεων πολλαπλών πολώσεων, έχουν αποτελέσει πηγή πολύτιμων πληροφοριών τηλεπισκόπησης. Έτσι, η χρήση των SAR δεδομένων για την ταξινόμηση κάλυψης γης προσελκύει όλο και περισσότερο την προσοχή των ερευνητών και φαίνεται να είναι πολλά υποσχόμενη. Η παρούσα ειδική επιστημονική εργασία έχει στόχο τη μελέτη και ερμηνεία των δεδομένων SAR μέσω επιβλεπόμενης ταξινόμησης, με τη χρήση νευρωνικών δικτύων (Neural Networks). Αφού πρώτα γίνεται εκτενής αναφορά στη τεχνολογία και τα συστήματα SAR, παρουσιάζεται αναλυτικά η πειραματική διαδικασία ταξινόμησης τριών βασικών δομών κάλυψης γης. Τα δεδομένα προέρχονται από το Προηγμένο Ραντάρ Συνθετικού Ανοίγματος (ASAR) του δορυφόρου ENVISAT από τον Ευρωπαϊκό Οργανισμό Διαστήματος και αφορούν στην ευρύτερη περιοχή του Άμστερνταμ. Πριν την διεξαγωγή της ταξινόμησης, τα δεδομένα δέχθηκαν τις απαραίτητες διαδικασίες προ-επεξεργασίας (ραδιομετρική βαθμονόμηση, γεωαναφορά, φιλτράρισμα θορύβου, συμπροσαρμογή). Όσον αφορά τη διαδικασία της ταξινόμησης, εξετάζεται η συμπεριφορά του ταξινομητή του νευρωνικού δικτύου για μεταβολές ποικίλων παραμέτρων, όπως η επιλογή δεδομένων διαφόρων πολώσεων, το πλήθος των νευρώνων κ.α. και ήδη από τα πρώτα πειράματα λαμβάνονται ικανοποιητικά αποτελέσματα. Στη συνέχεια εφαρμόζονται τεχνικές σύνθεσης πληροφορίας (average rule, majority rule) βελτιώνοντας τις επιδόσεις ταξινόμησης. Τέλος, ένα σημαντικό βήμα που εφαρμόζεται στη διαδικασία ταξινόμησης αποτελεί η εξαγωγή χαρακτηριστικών υφής από τις μήτρες συνεμφάνισης φωτεινοτήτων (Gray Level Co-occurrence Matrix-GLCM) και μήκους διαδρομής φωτεινότητας (Gray Level Run Length Matrix-GLRLM). Η χρήση των χαρακτηριστικών αυτών βελτιστοποιεί το σύστημα ταξινόμησης, δίνοντας εξαιρετικά αποτελέσματα. / The use of Synthetic Aperture Radar (SAR) data in remote sensing applications has become a cutting edge technology during the past few decades. The SAR systems have several capabilities, like day & night and all weather operation and they offer large ground coverage with the ability of multi-polarized imagery; therefore, they have proved to be a valuable source of remote sensing data. As a result, the use of SAR data for land cover classification increasingly attracts the attention of researchers and seems to be highly promising. Goal of this master thesis is the study and interpretation of SAR data through supervised classification, with the use of Neural Networks method. First, there is an extensive presentation of SAR systems and technology and then follows the detailed presentation of the experimental classification process for three basic land cover structures. The available data are from the Advanced SAR (ASAR) radar of the ESA ENVISAT satellite and correspond to the Amsterdam city and suburbs. Prior to the classification process, the data have been appropriately pre-processed (radiometric calibration, geocoding, speckle filtering, co-registration). Regarding the classification process, the response of the neural network classifier with the variation of several parameters (e.g. data polarization and number of neurons) is studied and from the initial test already the results were quite satisfactory. Further on, ensemble classifying methods (average rule, majority rule) are applied to improve the classification performance. Finally, as an essential step applied in the classification process is the textural feature extraction from Gray Level Co-occurrence Matrix (GLCM) and Gray Level Run Length Matrix (GLRLM). The use of these texture features optimizes the classification system, resulting to an exceptional performance.