ArcGIS Drone2Map leverages multithreaded CPUs and NVIDIA GPUs with CUDA compute capabilities to quickly turn drone imagery into usable data. However, not all steps use systems resources the same way. In addition to your computer’s hardware, the characteristics of your drone flight, such as number of images, overlap between images, and ground sampling distance (GSD), can affect performance. The sections below outline the hardware components that are used during processing and provide guidance for optimizing processing speeds.
Key factors that influence processing speed
The following are the most common factors that influence the speed of processing in Drone2Map projects:
- Hardware components
- CPU
- GPU
- RAM
- Storage
- Flight characteristics
- Number of images
- Image resolution
- Project settings
- Ground sampling distance
- Image scale
- Matching strategy
- Point cloud density
- Output products
Hardware components
Your computer's hardware components are used in varying capacities when processing projects. The following is a summary of the behavior of the main components when processing:
CPU
- Drone2Map uses parallel processing to leverage systems with multicore CPUs.
- The CPU is the most important hardware component to influence processing speeds. All processing steps benefit from upgrading the CPU, so upgrading the CPU should be the highest priority to achieve faster processing times.
- A CPU with higher clock speed will typically perform better than a comparable CPU with more cores.
GPU
- Drone2Map leverages NVIDIA GPUs with CUDA compute capabilities to further increase processing speeds.
- The GPU significantly increases the processing speeds of the Dense Matching and 2D Products steps.
- The GPU provides moderate processing performance for the Adjust Images step.
RAM
- Drone2Map uses the available RAM dynamically to optimize performance.
- Additional RAM increases processing speeds by allowing Drone2Map to do more work in parallel.
- Systems with high CPU core counts and not enough RAM may limit the processing potential of the CPU. It is recommend that you use at least 32 GB of RAM when using CPUs with high core counts.
Storage
- It is recommended that you store projects and images on SSDs or SSD RAIDs for optimal performance.
- Avoid processing projects from USB, network drives, and cloud storage drives, as the increased latency and risk of network outages may degrade performance or cause processing to fail.
- Once processing is complete, you can copy projects to slower HDDs for long-term storage.
Flight characteristics
Drone2Map provides multiple options for managing flights and increasing processing speed. The following tips can be used to ensure fast, high-quality products:
Number of images
- Increasing the number of images in a project increases the time it takes to process the project.
- Remove any images that do not contribute to the project, such as images during take-off or landing and images outside the project area.
- Use the Project Area tool to delineate the region of interest.
Resolution of images—Higher-resolution images generally require more processing power and storage space. Consider the resolution of your drone images when budgeting for hardware components to ensure optimal performance.
Project settings and generated products
Selecting additional output products increases processing time. You may want to select all products when processing a project, but selecting only the products that are within the scope of the project will maximize efficiency.
Drone2Map preserves key processing milestones, allowing for greater flexibility when deciding which products to process. If you decide to create additional products, you’ll only have to process the steps that are required for that product.
As with adding additional products, increasing the quality of processing options generally results in longer processing times. It is recommended that you use the processing template that matches the project requirements, as these templates have been optimized for quality and performance.
The following processing steps have the greatest impact on processing speed:
- General options
- Ground sampling distance
- Point cloud density
- Adjust images options
- Image scale
- Matching neighborhood
Consider the following scenarios to determine the optimal processing options:
Scenario 1
You are out in the field running a drone flight on a laptop or computer with lower-end hardware resources. The flight has completed, but you are unsure of the quality. You want to see how well the flight reconstructs in Drone2Map, but you have limited time to finalize the mission and return to the office.
To obtain the fastest reconstruction, the best option is to start with the Rapid template. In the processing options for the project, you can lower the project resolution on the general tab to 8x GSD. This automatically lowers the point cloud density to low. Additionally, on the Adjust Images tab, you can increase the initial image scale to 1/8 (eighth image scale). Using these settings provides you with the fastest possible orthomosaic.
Scenario 2
Multiple flights were recently completed, and you are now tasked with processing them in Drone2Map. You have a limited window of time for results but want high-quality products. The computer being used is a desktop machine with midrange hardware resources. Each flight needs to have an orthomosaic and both elevation surfaces processed.
The best option to process projects quickly, but at a high quality, is to use the 2D products template. The default settings for this template configure the project resolution to be 1x GSD, process the dense point cloud at high density, and run a second pass at 1x resolution when generating tie points to obtain more matches. You can increase the matching neighborhood under tie point options to potentially gain more matches between neighboring images, but this is only recommended if you are not as confident in the levels of overlap between images.
Scenario 3
You have an Advanced license and a project that you want to process to output high-quality 2D and 3D products. Time is not a concern, and you want the highest quality that Drone2Map can provide.
Considering how 3D products will be processed, it is best to start with the 3D products template. This template is automatically set up to process the project resolution at 1x GSD. You can increase the point cloud density ultra instead of high. The default template tie points settings are configured to run a second pass at full resolution to gather more matches. You can increase the matching neighborhood to Large or X-Large to obtain the maximum amount of tie points. This provides the highest level of processing available in Drone2Map. Keep in mind that depending on your hardware resources, processing takes more time to complete with these settings.