Detection System: Iteration 2 Research

Problem Statement

Our current system is only capable of tracking and flying towards one target. Further, target detection, location and selection occurs in the same process.

Requirements

• New target detection system with independent target identification and selection.

  • Therefore, targets are stored than selected, rather than taken and selected as they come.

  • Require a diagram explaining how targets are detected then selected.

  • Require an overview of the data structure for storing targets.

• Review of target identification and geolocation subsystems.

  • Require a diagram and explanation of new systems, if current system should be improved.

Scope

Must have:

• High level block diagram of how a new target system would work. Including data into each module, data out from each module, and methods in each module.

• Verification that concepts are possible to implement in code.

• Data structure for storing target information.

Nice to have:

• Implementation details on possible libraries.

Not in scope:

• Any implementation in code.

Milestones and deadlines

August 19th Discussion

• Busy Situations → If we’re flying over targets in an area with a lot of points of interest that can result in false detection, a possible approach is to use image segmentation with filtering to identify points of interests before scanning,

• Target Confidence → Mapping system can also hold target “confidence” which can come from both the model’s confidence in the target and the # of times we see the target.

Reference materials

Research on UAV ground targets:

• https://www.mdpi.com/1424-8220/20/1/272

• https://journals.sagepub.com/doi/pdf/10.5772/58989

• https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7037756/