2020 Geospatial Thesis Topics
We’re looking for graduate students at our Linköping office.
Topic: Utilizing extended multispectral images for ML/AI classification.
Students will investigate the pros and cons of using a range of multispectral bands in the shortwave infrared bands for classification of buildings, materials, and more.
Topic: Towards Global AI/ML coverage.
Students will investigate methods for getting global coverage—e.g. using one global network or several depending on geographic position—to best annotate data from different parts of the globe and measure classification results.
Topic: AI/ML based Digital Terrain Model generator.
Students will investigate if a DTM generator could be created using an ML process that would employ filtering techniques and discard non-ground objects from DSM data; the goal is an AI-based algorithm that would create a cleaner, more easily maintained, and better DTM.
Topic: Map open source vector data to high resolution 3D map.
Students will investigate techniques to align third-party vector data to Vricon’s 3D model to ensure objects and artifacts—e.g., light poles, benches, etc.—are in their proper position.
Topic: AI/ML based Bridge Extraction from 3D-Models.
Students will investigate methods for multi-view classification to get a more precise classification of bridges and extract accurate data related to bridge length, width, height, and thickness.
Topic: AI/ML based area classification.
Vricon seeks to divide its 3D model into higher order classes, such as dessert, dense-urban, suburb, airport, golf course, etc.; students will investigate if a highly detailed classification could be used to train a low-level area classifier, either using detailed classification as input or to generate training data.
Topic: Transitioning Computer Vision Algorithms from CPU to CPU/GPU.
Students will investigate if advanced 3D reconstruction algorithms could be transitioned to GPUs; this would include measuring performance and comparing code complexity among hand-optimized CPU code, CPU/GPU-agnostic code running on CPUs, and CPU/GPU-agnostic code running on GPUs.