50mm.at – an objective view to the world

integrated project – near field photogrammetry

Post author By Peter Berger
Post date 20. May 2018

Two well known commercial Photogrammetry Software Suites, Agisoft Photoscan and Pix4D Pix4Dmapper have been compared to a Opensource product called Visual SFM with PMVS plugin. The Results showed about the same amount of point density for the two commercial products, with VSFM behind. The detail of rendering from Pix4D is inferior to Agisoft, espacially in the term of roughness of the produced point cloud. A ground sampling distance of 2,5 cm has been estimated with Pix4D which is competitive to Lidar workflows.

Objective

The main objective of the project was to compare the output quality of different Software products based on the same input dataset. Quality was rated in terms of the local density of the point cloud, point distances compared to the other products and the roughness of the cloud, representing the modeling detail. Ease of use and the ability in Georeferencing was rated as well.

Method

The input dataset consisted of 34 aerial images of an alpine riverbed in Obermieming / Tirol made from an UAV [flight height ~ 95m above ground] and 6 corresponding ground control points measured with differential GPS. A dense pointcloud has been produced in each case, using the integrated georeferncing tools as well. The datasets have been imported to the Software CloudCompare to evaluate the quality of the output by comparing each result with the other ones.

Results

Agisoft and Pix4D modeled about the same extent, with better results in vegetated areas from Pix4D. VisualSFM did only model the unvegetated areas with unsatisfying results in the vegetated areas.

For further reading see attached Poster.

Tags near field photogrammetry, photogrammetry, project, UAV

Courses

spatial simulation – Netlogo bicycle model

Post author By Peter Berger
Post date 20. May 2018

The purpose of this model is to simulate the bicycle traffic in a hypothetic approach, given the fact that pedelec-bikes with limited range have to use the provided charging stations. The routing has to be modified for all cyclists which have not sufficient range to fulfil their route to the target location.

As it can be assumed, in a practical situation it would not make sense to provide charging stations for cyclists in a city, because the range of current pedelec bikes is high enough to do mostly all ways on a whole day in a city. Pedelec bikes are normally charged at home, as their batteries normally can be unmounted from the bike and charged in the living rooms of the user.

This model should be the basis for further investigations to fast-charge electric vehicles, where the street-network of the whole province of Salzburg will be imported in the model.

In a further step it could be implemented to let bicyclists wait at the charging stations for the needed period of time for charging. While the charge-point is probably already busy, the 2^nd or further incoming cyclists have to wait for their charge.

In this context the “stress” while waiting for the charge could be measured in the model to further improve the charging infrastructure to the given demand.

Conceptual model of routing algorithm

The procedure for creating the route from the location to the target of the cyclist is more complicated when the use of a charger is needed. First the shortest route and distance is computed. When this route is shorter than the range everything is fine and the routing process finishes.

If the route is longer than the range first the distance to all chargers need to be calculated, and second all distances from all chargers to the target has to be computed. These distances are stored in a list.

The next step is to sort out the charging-station, which provides the shortest distance (to charger + to target), which also fulfills the condition that the distance from the location to the charger and the distance from the charger to the target are within the range of the agent. Otherwise it gets sort out.

The model can now setup agents with the routing engine using stopover points, which are in this case charging stations. While using the standard routing without the use of chargers and the range parameter, the setup of 1000 cyclists needs about 1:30 minutes, while the setup of 1000 cyclists with charging enabled and 500 [patches] of range the process consumes 2:30 minutes of time. Using 300 [patches] as range further increases the number of agents which use a routing with charging and also the amount of dropped agent rises, the setup of 1000 cyclists takes 4:20 minutes.

The charging stations have been positioned manually by selecting vertices where a high traffic occurs. The simulation was running without charging with 1000 agents. Based on the resulting heatmap the charging-stations (orange) have been setup, this is shown in the following figure.

Tags bicycle, netlogo, spatial simulation, traffic

Courses

spatial data aquisition – orthophoto generation

Post author By Peter Berger
Post date 20. May 2018

The Goal of the project work was to get experience on using Handheld GPS devices, as well as working with the Software Trimble Pathfinder Office and Erdas Imagine LPS to generate an Orthophoto and a digital elevation model out of 6 aerial pictures and some additional data sources.

The aerial pictures are originally film-type, but have been provided already scanned to a digital format. As the pictures are already historic, there was also the challenge to find the pictured scenery in real life. Therefore it was essential to go well prepared into the field without the risk of doing false measurements as this could lead to a second try for the field work.

Data Processing:

The data was processed in the ERDAS Imagine 2013 toolbox “Leica Photogrammetry Suite” (LPS). For projection variables the horizontal projection was set to BMN M31 (EPSG:31258), the vertical projection was set to the Bessel spheroid with the MGI (Hermannskogel) date. As for the camera calibration information an average flying height of 3800m was chosen. This reflects the average flying height of the plane of 4700m above sea level in combination with the average elevation of the study area. Additional camera calibration information was retrieved from an calibration file in .cam- format.

After entering this first set of information about the coordinate reference system and the recording-parameters, the workflow as shown was used.

Result

The final result of the project is the mosaicked orthophoto, overlaid onto the generated DTM. The final output covers the area around the Gaisberg. In the northeast, it is just missing out the main parts of the villages of Schnurnn and Koppl. As for the southwest, it shows the buildings close to “Schloss Aigen”.

Tags course, exercise, GPS, orthophoto