Diseño e implementación de un prototipo de iluminación automático en base a visión artificial orientado a mejorar la adquisición de imágenes en entornos no controlados

The ability of a computer vision system to capture images is largely affected by the lighting conditions in the scene. This research highlights the importance of an automatic lighting system in computer vision applications, and proposes a solution to improve image acquisition in a workspace. The development of the work began with the selection of relevant features that describe the qualities of an image, such as luminance, sharpness, colorfulness, and information entropy, which are directly related to the amount of illumination of a scene. Algorithms were designed and implemented to obtain a numerical metric for each of the features mentioned above. The application of these algorithms on a dataset of images with different illumination levels provided a subspace of attributes for training of a supervised learning classification model, based on Support Vector Machine (SVM) with a polynomial kernel. The classification model categorized the images into 'dark', 'normal', and 'bright', results that were used to control an external device which adjusts the lighting intensity of a scene accordingly. Once the system was designed, a prototype was developed to perform practical tests of its performance, delivering satisfactory results when using the proposed work. Keywords: computer vision, automatic illumination, image processing, SVM.

• Autor principal: Rivera Merchán, David Andrés (author)
• Formato: bachelorThesis
• Lenguaje: spa
• Publicado: 2023
• Materias: <VISIÓN POR COMPUTADOR> < ILUMINACIÓN AUTOMÁTICA> < PROCESAMIENTO DE IMÁGENES> < SVM>
• Acceso en línea: https://dspace.unl.edu.ec/jspui/handle/123456789/26421