Raster Ranger
The Problem
Raster Ranger seeks to better integrate aerial imagery analysis and geospatial machine learning into leading field conservation software platforms (SMART and Earthranger) to protect vulnerable wildlife and contribute to global conservation protection efforts. Conservation protection encompasses several areas of activity, including: counter wildlife trafficking, protected area management, local and global anti-poaching efforts, law enforcement, and intelligence activities.
Our Proposal
Raster Ranger programmatically connects field data (e.g., from a SMART patrol report or Earthranger deployment) to available satellite/aerial imagery through a series of automated queries, then queues those associated images and data for distributed review and labeling, to create high-quality datasets that can be used for downstream geospatial machine learning applications. The Raster Ranger outputs are intended to be used either directly back in the source applications (e.g., SMART/Earthranger dashboards) or as inputs for additional analyses. Raster Ranger is built using best-in-class free and open source software developed by the greater geospatial community. The project is currently in the prototype stage and several conservation partners are interested and ready to collaborate on implementation and evaluation in the field.
We Assume that...
Local conservation protection stakeholders generally have concerns about geospatial data quality, availability, and security within their areas of responsibility.
Local conservation protection stakeholders generally have limited time and attention to develop new technical capabilities, especially capabilities that take them away from fieldwork.
There is significant demand for more, better, high-quality, and highly available labeled training datasets for conservation applications of geospatial machine learning.
It is critical for our solution to prioritize data interoperability and build on pre-existing solutions and standards to encourage adoption and facilitate ease of use.
Constraints to Overcome
Our solution addresses current significant barriers to the widespread adoption and application of powerful new geospatial machine learning techniques for conservation protection by providing and promoting two important capabilities: 1) Automating the process of connecting ground truth observations to available remote sensing imagery and 2) Distributing the development of high-quality labeled training data for pattern detection (without overburdening conservation stakeholders in the field) Removing these barriers will facilitate new opportunities to leverage geospatial machine learning for better, faster, and safer conservation protection.
Current Work
We have developed a model architecture for the Raster Ranger application and should be ready for simulated field testing soon. Several conservation stakeholders are interested to collaborate and implement Raster Ranger in the field for additional evaluation. We're working to develop, refine, and document the Raster Ranger application to facilitate easier integration with the SMART conservation software platform. This will be completed as part of the Con X Tech Prize award. After that development and documentation is complete, we hope to work directly with the SMART development team to design an integration API or UI that lets users connect, request, and receive/display relevant results and info from Raster Ranger within SMART.
Current Needs
Our primary needs at this point are critical feedback from interested field partners and cloud computing resources to continue Raster Ranger development. Communication support would also be great - anyone interested in the project or related topics is welcome to help out!