Procesamiento digital de imágenes en detección de material contaminante inorgánico en orillas de lagunas.

Abstract

The objective of this work was to apply digital image processing in the detection of inorganic contaminant material on the shores of lagoons, for which an object detection system capable of identifying plastic bottles present in the ecosystem was carried out by calibrating a detector. In the development of the research, we worked with theoretical-empirical methods of quantitative and qualitative approaches. For the calibrating of the detector, different detection algorithms were used based on the following descriptors: Local Binary Patterns (LBP), Histogram of oriented gradients (HOG) and Aggregated Channel Characteristics (ACF) which were evaluated based on their sensitivity and precision. As a result of the evaluation, it was determined that the cascade detection system with a Histogram descriptor of oriented gradients has a greater precision in the recognition of plastic bottles with a detection percentage of 70%. The detection system was designed with two modules, the first, the image acquisition module, in charge of capturing, selecting, and storing the data for which an Unmanned Aerial Vehicle (UAV) is used together with the software of mapping PIX4D Capture. In the second module, the image processing module, the data is standardized, the characteristics of the object in the images are improved with the use of noise, contrast, and histogram adjustment filters, the detection algorithm is executed and the area occupied by the bottle in pixels is calculated so as to finally present the area of the bottles detected in the analysis zone.