FPGA-based object detection using classification circuits

Dans l'apprentissage machine, la classification est le processus d’assigner une nouvelle observation à une certaine catégorie. Les classifieurs qui mettent en œuvre des algorithmes de classification ont été largement étudié au cours des dernières décennies. Les classifieurs traditionnels sont basés sur des algorithmes tels que le SVM et les réseaux de neurones, et sont généralement exécutés par des logiciels sur CPUs qui fait que le système souffre d’un manque de performance et d’une forte consommation d'énergie. Bien que les GPUs puissent être utilisés pour accélérer le calcul de certains classifieurs, leur grande consommation de puissance empêche la technologie d'être mise en œuvre sur des appareils portables tels que les systèmes embarqués. Pour rendre le système de classification plus léger, les classifieurs devraient être capable de fonctionner sur un système matériel plus compact au lieu d'un groupe de CPUs ou GPUs, et les classifieurs eux-mêmes devraient être optimisés pour ce matériel. Dans ce mémoire, nous explorons la mise en œuvre d'un classifieur novateur sur une plate-forme matérielle à base de FPGA. Le classifieur, conçu par Alain Tapp (Université de Montréal), est basé sur une grande quantité de tables de recherche qui forment des circuits arborescents qui effectuent les tâches de classification. Le FPGA semble être un élément fait sur mesure pour mettre en œuvre ce classifieur avec ses riches ressources de tables de recherche et l'architecture à parallélisme élevé. Notre travail montre que les FPGAs peuvent implémenter plusieurs classifieurs et faire les classification sur des images haute définition à une vitesse très élevée.

Object and Scene Classification: what does a Supervised Approach Provide us?

Given a set of images of scenes containing different object categories (e.g. grass, roads) our objective is to discover these objects in each image, and to use this object occurrences to perform a scene classification (e.g. beach scene, mountain scene). We achieve this by using a supervised learning algorithm able to learn with few images to facilitate the user task. We use a probabilistic model to recognise the objects and further we classify the scene based on their object occurrences. Experimental results are shown and evaluated to prove the validity of our proposal. Object recognition performance is compared to the approaches of He et al. (2004) and Marti et al. (2001) using their own datasets. Furthermore an unsupervised method is implemented in order to evaluate the advantages and disadvantages of our supervised classification approach versus an unsupervised one

Image classification for a large number of object categories

L'increment de bases de dades que cada vegada contenen imatges més difícils i amb un nombre més elevat de categories, està forçant el desenvolupament de tècniques de representació d'imatges que siguin discriminatives quan es vol treballar amb múltiples classes i d'algorismes que siguin eficients en l'aprenentatge i classificació. Aquesta tesi explora el problema de classificar les imatges segons l'objecte que contenen quan es disposa d'un gran nombre de categories. Primerament s'investiga com un sistema híbrid format per un model generatiu i un model discriminatiu pot beneficiar la tasca de classificació d'imatges on el nivell d'anotació humà sigui mínim. Per aquesta tasca introduïm un nou vocabulari utilitzant una representació densa de descriptors color-SIFT, i desprès s'investiga com els diferents paràmetres afecten la classificació final. Tot seguit es proposa un mètode par tal d'incorporar informació espacial amb el sistema híbrid, mostrant que la informació de context es de gran ajuda per la classificació d'imatges. Desprès introduïm un nou descriptor de forma que representa la imatge segons la seva forma local i la seva forma espacial, tot junt amb un kernel que incorpora aquesta informació espacial en forma piramidal. La forma es representada per un vector compacte obtenint un descriptor molt adequat per ésser utilitzat amb algorismes d'aprenentatge amb kernels. Els experiments realitzats postren que aquesta informació de forma te uns resultats semblants (i a vegades millors) als descriptors basats en aparença. També s'investiga com diferents característiques es poden combinar per ésser utilitzades en la classificació d'imatges i es mostra com el descriptor de forma proposat juntament amb un descriptor d'aparença millora substancialment la classificació. Finalment es descriu un algoritme que detecta les regions d'interès automàticament durant l'entrenament i la classificació. Això proporciona un mètode per inhibir el fons de la imatge i afegeix invariança a la posició dels objectes dins les imatges. S'ensenya que la forma i l'aparença sobre aquesta regió d'interès i utilitzant els classificadors random forests millora la classificació i el temps computacional. Es comparen els postres resultats amb resultats de la literatura utilitzant les mateixes bases de dades que els autors Aixa com els mateixos protocols d'aprenentatge i classificació. Es veu com totes les innovacions introduïdes incrementen la classificació final de les imatges.

A comparison of pixel and object-based land cover classification. A case study of the Asmara region, Eritrea

In this paper we compare the performance of two image classification paradigms (object- and pixel-based) for creating a land cover map of Asmara, the capital of Eritrea and its surrounding areas using a Landsat ETM+ imagery acquired in January 2000. The image classification methods used were maximum likelihood for the pixel-based approach and Bhattacharyya distance for the object-oriented approach available in, respectively, ArcGIS and SPRING software packages. Advantages and limitations of both approaches are presented and discussed. Classifications outputs were assessed using overall accuracy and Kappa indices. Pixel- and object-based classification methods result in an overall accuracy of 78% and 85%, respectively. The Kappa coefficient for pixel- and object-based approaches was 0.74 and 0.82, respectively. Although pixel-based approach is the most commonly used method, assessment and visual interpretation of the results clearly reveal that the object-oriented approach has advantages for this specific case-study.

Multiscale object-based classification of satellite images merging multispectral information with panchromatic textural features

Once admitted the advantages of object-based classification compared to pixel-based classification; the need of simple and affordable methods to define and characterize objects to be classified, appears. This paper presents a new methodology for the identification and characterization of objects at different scales, through the integration of spectral information provided by the multispectral image, and textural information from the corresponding panchromatic image. In this way, it has defined a set of objects that yields a simplified representation of the information contained in the two source images. These objects can be characterized by different attributes that allow discriminating between different spectral&textural patterns. This methodology facilitates information processing, from a conceptual and computational point of view. Thus the vectors of attributes defined can be used directly as training pattern input for certain classifiers, as for example artificial neural networks. Growing Cell Structures have been used to classify the merged information.

Habitat map of Danajon Bank, Philippines, derived from a high-spatial-resolution multi-spectral satellite image and georeferenced point intercept transect and spot-check survey field data, using an object-based image classification method

A mosaic of two WorldView-2 high resolution multispectral images (Acquisition dates: October 2010 and April 2012), in conjunction with field survey data, was used to create a habitat map of the Danajon Bank, Philippines (10°15'0'' N, 124°08'0'' E) using an object-based approach. To create the habitat map, we conducted benthic cover (seafloor) field surveys using two methods. Firstly, we undertook georeferenced point intercept transects (English et al., 1997). For ten sites we recorded habitat cover types at 1 m intervals on 10 m long transects (n= 2,070 points). Second, we conducted geo-referenced spot check surveys, by placing a viewing bucket in the water to estimate the percent cover benthic cover types (n = 2,357 points). Survey locations were chosen to cover a diverse and representative subset of habitats found in the Danajon Bank. The combination of methods was a compromise between the higher accuracy of point intercept transects and the larger sample area achievable through spot check surveys (Roelfsema and Phinn, 2008, doi:10.1117/12.804806). Object-based image analysis, using the field data as calibration data, was used to classify the image mosaic at each of the reef, geomorphic and benthic community levels. The benthic community level segregated the image into a total of 17 pure and mixed benthic classes.

An approach to document classification using verb-object pairs

The aim of this thesis is to present a new approach to document classification using verb-object pairs. We explore one possible strategy that uses the presence of relevant verb-object pairs in documents as features and a Naive Bayes classifier as a classifier on which the model is trained. Then, we assess the results from the case study which uses a software based on the strategy and make conclusions.

Multilabeled classification approach to find a plaint source for terpenoids

Recently, we have built a classification model that is capable of assigning a given sesquiterpene lactone (STL) into exactly one tribe of the plant family Asteraceae from which the STL has been isolated. Although many plant species are able to biosynthesize a set of peculiar compounds, the occurrence of the same secondary metabolites in more than one tribe of Asteraceae is frequent. Building on our previous work, in this paper, we explore the possibility of assigning an STL to more than one tribe (class) simultaneously. When an object may belong to more than one class simultaneously, it is called multilabeled. In this work, we present a general overview of the techniques available to examine multilabeled data. The problem of evaluating the performance of a multilabeled classifier is discussed. Two particular multilabeled classification methods-cross-training with support vector machines (ct-SVM) and multilabeled k-nearest neighbors (M-L-kNN)were applied to the classification of the STLs into seven tribes from the plant family Asteraceae. The results are compared to a single-label classification and are analyzed from a chemotaxonomic point of view. The multilabeled approach allowed us to (1) model the reality as closely as possible, (2) improve our understanding of the relationship between the secondary metabolite profiles of different Asteraceae tribes, and (3) significantly decrease the number of plant sources to be considered for finding a certain STL. The presented classification models are useful for the targeted collection of plants with the objective of finding plant sources of natural compounds that are biologically active or possess other specific properties of interest.

Intercepting a moving target: effects of temporal precision constraints and movement amplitude

The effects of temporal precision constraints and movement amplitude on performance of an interceptive aiming task were examined. Participants were required to strike a moving target object with a 'bat' by moving the bat along a straight path (constrained by a linear slide) perpendicular to the path of the target. Temporal precision constraints were defined in terms of the time period (or window) within which contact with the target was possible. Three time windows were used (approx. 35, 50 and 65 ms) and these were achieved either by manipulating the size of the bat (experiment 1a), the size of the target (experiment 1b) or the speed of the target (experiment 2). In all experiments, movement time (MT) increased in proportion to movement amplitude but was only affected by differences in the temporal precision constraint if this was achieved by variation in the target's speed. In this case the MT was approximately inversely proportional to target speed. Peak movement speed was affected by temporal accuracy constraints in all three experiments: participants reached higher speeds when the temporal precision required was greater. These results are discussed with reference to the speed-accuracy trade-off observed for temporally constrained aiming movements. It is suggested that the MT and speed of interceptive aiming movements may be understood as responses to the spatiotemporal constraints of the task.

The Edinburgh-Cape Blue Object Survey - I. Description of the survey

The Edinburgh-Cape Blue Object Survey is a major survey to discover blue stellar objects brighter than B similar to 18 in the southern sky. It is planned to cover an area of sky of 10 000 deg(2) with \b\ > 30 degrees and delta < 0 degrees. The blue stellar objects are selected by automatic techniques from U and B pairs of UK Schmidt Telescope plates scanned with the COSMOS measuring machine. Follow-up photometry and spectroscopy are being obtained with the SAAO telescopes to classify objects brighter than B = 16.5. This paper describes the survey, the techniques used to extract the blue stellar objects, the photometric methods and accuracy, the spectroscopic classification, and the limits and completeness of the survey.

Using different satellite imagery and classification techniques to assess the contribution of trees outside forests in the municipality of Maringá, Brazil

Forest cover of the Maringá municipality, located in northern Parana State, was mapped in this study. Mapping was carried out by using high-resolution HRC sensor imagery and medium resolution CCD sensor imagery from the CBERS satellite. Images were georeferenced and forest vegetation patches (TOFs - trees outside forests) were classified using two methods of digital classification: reflectance-based or the digital number of each pixel, and object-oriented. The areas of each polygon were calculated, which allowed each polygon to be segregated into size classes. Thematic maps were built from the resulting polygon size classes and summary statistics generated from each size class for each area. It was found that most forest fragments in Maringá were smaller than 500 m². There was also a difference of 58.44% in the amount of vegetation between the high-resolution imagery and medium resolution imagery due to the distinct spatial resolution of the sensors. It was concluded that high-resolution geotechnology is essential to provide reliable information on urban greens and forest cover under highly human-perturbed landscapes.

Land use / land cover change detection: an object oriented approach, Münster, Germany

Dissertation submitted in partial fulfilment of the requirements for the Degree of Master of Science in Geospatial Technologies

Velocity estimation of moving ships using C-band SLC SAR data

A new algorithm for the velocity vector estimation of moving ships using Single Look Complex (SLC) SAR data in strip map acquisition mode is proposed. The algorithm exploits both amplitude and phase information of the Doppler decompressed data spectrum, with the aim to estimate both the azimuth antenna pattern and the backscattering coefficient as function of the look angle. The antenna pattern estimation provides information about the target velocity; the backscattering coefficient can be used for vessel classification. The range velocity is retrieved in the slow time frequency domain by estimating the antenna pattern effects induced by the target motion, while the azimuth velocity is calculated by the estimated range velocity and the ship orientation. Finally, the algorithm is tested on simulated SAR SLC data.

Multi-scale vegetation classification using earth observation data of the Sundarban mangrove forest, Bangladesh.

Dissertation submitted in partial fulfilment of the requirements for the Degree of Master of Science in Geospatial Technologies.

Amorphous silicon e 3D sensors applied to object detection

Nowadays, existing 3D scanning cameras and microscopes in the market use digital or discrete sensors, such as CCDs or CMOS for object detection applications. However, these combined systems are not fast enough for some application scenarios since they require large data processing resources and can be cumbersome. Thereby, there is a clear interest in exploring the possibilities and performances of analogue sensors such as arrays of position sensitive detectors with the final goal of integrating them in 3D scanning cameras or microscopes for object detection purposes. The work performed in this thesis deals with the implementation of prototype systems in order to explore the application of object detection using amorphous silicon position sensors of 32 and 128 lines which were produced in the clean room at CENIMAT-CEMOP. During the first phase of this work, the fabrication and the study of the static and dynamic specifications of the sensors as well as their conditioning in relation to the existing scientific and technological knowledge became a starting point. Subsequently, relevant data acquisition and suitable signal processing electronics were assembled. Various prototypes were developed for the 32 and 128 array PSD sensors. Appropriate optical solutions were integrated to work together with the constructed prototypes, allowing the required experiments to be carried out and allowing the achievement of the results presented in this thesis. All control, data acquisition and 3D rendering platform software was implemented for the existing systems. All these components were combined together to form several integrated systems for the 32 and 128 line PSD 3D sensors. The performance of the 32 PSD array sensor and system was evaluated for machine vision applications such as for example 3D object rendering as well as for microscopy applications such as for example micro object movement detection. Trials were also performed involving the 128 array PSD sensor systems. Sensor channel non-linearities of approximately 4 to 7% were obtained. Overall results obtained show the possibility of using a linear array of 32/128 1D line sensors based on the amorphous silicon technology to render 3D profiles of objects. The system and setup presented allows 3D rendering at high speeds and at high frame rates. The minimum detail or gap that can be detected by the sensor system is approximately 350 μm when using this current setup. It is also possible to render an object in 3D within a scanning angle range of 15º to 85º and identify its real height as a function of the scanning angle and the image displacement distance on the sensor. Simple and not so simple objects, such as a rubber and a plastic fork, can be rendered in 3D properly and accurately also at high resolution, using this sensor and system platform. The nip structure sensor system can detect primary and even derived colors of objects by a proper adjustment of the integration time of the system and by combining white, red, green and blue (RGB) light sources. A mean colorimetric error of 25.7 was obtained. It is also possible to detect the movement of micrometer objects using the 32 PSD sensor system. This kind of setup offers the possibility to detect if a micro object is moving, what are its dimensions and what is its position in two dimensions, even at high speeds. Results show a non-linearity of about 3% and a spatial resolution of < 2µm.