Automated interpretation of 3D and image data
Today, high-performance cameras or laser scanners are used for many surveying tasks and condition monitoring. They deliver high-resolution images and very accurate, geo-referenced measurement data. The data are usually interpreted manually. Advances in artificial intelligence enable the use of innovative techniques and algorithms to automate the evaluation process.
Today, complex learning algorithms based on the concept of »deep learning« with artificial neural networks (ANN) are used for the evaluation of 3D data. These algorithms have been shown to be superior to traditional methods of object recognition. Just a few years ago, the situation was different: The training of bespoke algorithms took weeks or even months. Today, thanks to massive parallelization, this process can be accomplished in just a few hours. The evaluation of new data sets based on a trained ANN is then even conducted in real time. In ANN, the fed-in information passes through a large number of interconnected artificial neurons, where it is processed and passed to other neurons. ANNs learn the output patterns corresponding to specific input patterns with the help of manually annotated training data. Based on this »experience«, new types of input data can then be analyzed in real time. ANNs have proven to be very robust to variations on characteristic colors, edges and shapes.