A dynamical system approach for pattern recognition and image analysis in biometrics and phytopathology

The research presented in this thesis concerns the efficient application of the positive dynamical systems theory to problems arising in pattern recognition and image analysis, specifically, in the biometric security and plant pathology areas, providing both theoretical and experimental results. Thus, a novel approach to this kind of problems has been investigated. With this in mind, the principle contributions of this thesis can be summarised within the context of the above overlapping lines of research. In the first part, an introduction to the field of biometrics is given in order to present the concepts and primitives of performance metrics due to their impact on secure biometric systems. Secondly, it has been investigated the feasibility of the proposed approach in biometrics. This study has led to the definition of a unified method for line-like feature matching that relies on a recursive algorithm based on a monotone dynamical system whose output converges either to zero, when a deep mismatch exists between the samples to be compared, or to a high value, when a good matching is observed, thus allowing the system to be employed in several applications, including all possible vascular-based biometric security systems based on blood vessel pattern matching. Thirdly, to consolidate the theoretical results, two examples of biometric security systems have been developed. In particular, it has been considered the case of hand palm-based human recognition first using the samples acquired in the visible spectrum and then those acquired in the near-infrared spectrum. In the second part, at first is given an introduction to the field of phytopathology oriented to image-based diagnosis of plant disease symptoms in order to present the concepts and primitives of performance metrics due to their impact on such systems. After that, it has been investigated the feasibility of the proposed approach in plant pathology. Hence, to detect potential plant pathogens as quickly as possible in order to reduce the likelihood of an infection spreading, it has been proposed a unified method based on the positive dynamical systems theory that allows the detection and severity estimation of grape diseases regardless of disease type. Lastly, to consolidate the theoretical results, an example of grape leaf disease detection and severity estimation has been developed. In particular, it has been considered the case of a specific disease-causing agent due to biotic factors (i.e., those caused by living components such as pathogens). In both the proposed unified methods, the main advantage rely in the robustness when dealing with low-resolution and noisy images. Indeed, an essential issue related to digital image processing is to effectively reduce noise from an image whilst keeping its features intact. The impact of noise (e.g., signal independent and uncorrelated noise) is effectively reduced and does not affect the final result allowing the proposed systems to ensure a high accuracy and reliability.