Introduction
In this lab we where introduced to the 3d mapping and model software knows as Pix4d. Using images that where previously collected we uploaded them to the software in order to create a 3d replication. By uploading photos of the subject which was a truck in this case, the software was able to analyze the data using information from the photos such as that as GPS information as well as by creating different point clouds from the images.
Methods
Before any data processing could be started data needed to be collected in the field. For this lab everyone in the class used the same data set that was provided by our professor. The data set included about 70 photos that where taken in an orbital pattern around the subject which was his truck. Once the file containing the data was accessed using a Windows based computer, I selected the images to upload and began the initial data processing using the program Pix4d. The initial processing did not take too long as the data set was rather small so after about 25 minutes a initial data report was created automatically by the program which I then analyzed for any deiscremancies in the data set. None where found so the data processing contained so that all point clouds where generated and then finally I was abler to enable triangular mesh which finds matching sides to blend together to make the model look more realistic and complete.
Discussion
After the data processing was complete and the final model was analyzed there were several key areas that are worth noting and discussing specifically in regards to the surface construction of the truck in some areas such as the windshield and one of the doors. In the images below you will be able to see some areas that appear to be incomplete or have holes in it. After analyzing these spaces and thinking about what may have caused it I think it can be attributed to the surface of the windshield being glass and how it is at a different angle than the rest of the truck so when the light from the sun hits it, it bounces off in a way not uniform to the rest of the image. Something else worth discussing is how enabling different settings on the left side of the. the screen allows for the 3d model to b3come more defined in nature specifically when enabling the triangular mesh.
Results
Figure 1: This 3d model is very raw just finishing its initial processing and having no setting changed to better establish a more complete and defined truck. You will see another angle with the same basic initial processing setting below in figure 4 . One step that could be done however to make a more defined model is by allowing for point clouds to be enabled as well as the triangular meshes.
Figure 2: Truck coming into focus
In figure 2 the truck is becoming more defined and starting to become more of the focus point of the 3d model which is ultimately the goal. The one setting changed in this image compared to figure # 4 is the fact that point clouds where enabled. By enabling point clouds we are able to achieve more high resolution data without distorting the model/data quality. The model could still be improved if we where to turn on the triangular mesh but this will be explained further in the figure below in figure # 3.
Figure 3: Point cloud and Mesh enabled.
By enabling the mesh I am allowing for the image rays of data points that share common edges and corners to be reflected in a ore solid appearing way which helps eliminate some of the blurriness observed earlier.
Figure 4: No triangular meshes or point clouds are enabled.
With no triangular point clouds or meshes enabled it leaves you only to be able to see a more raw unfinished 3d model of the truck. It is important to note the texture of the image appearing as more of a grainy surface due to no mesh or point clouds being enabled.
Comments