Comparison of CNN and MLP classifiers for algae detection in underwater pipelines

Abstract

Artificial neural networks, such as the multilayer perceptron (MLP), have been increasingly employed in various applications. Recently, deep neural networks, specially convolutional neural networks (CNN), have received considerable attention due to their ability to extract and represent high-level abstractions in data sets. This article describes a vision inspection system based on deep learning and computer vision algorithms for detection of algae in underwater pipelines. The proposed algorithm comprises a CNN or a MLP network, followed by a post-processing stage operating in spatial and temporal domains, employing clustering of neighboring detection positions and a region interception framebuffer. The performances of MLP, employing different descriptors, and CNN classifiers are compared in real-world scenarios. It is shown that the post-processing stage considerably decreases the number of false positives, resulting in an accuracy rate of 99.39%.