Wireless DNA hackathon

data at https://filesender.funet.fi/?s=download&token=d05d4253-ef08-439e-968f-a10ef84877a4

Result: Won through using method from Vaisala

Problem Statement

Delete, delete and reconstruct radar interferences.

Objective

Digitalize radar values and locate the img pixels on real locations (lat/lng of pixels) --> because of radar errors Detect interferences Delete interferences Correct interferences Save data into a file in netCDF format.

Data and Assumptions

Real radar data of 5 months saptial resolution of 1km per pixel sampling rate of 10 minutes --> maybe interferences doesnt change between the images

ba, pm, va --> locations then inside is each day with 10 min intervals

each image from each area kinda overlap with each other --> u can correlate and check for interferences. dry: jan, feb wet: apri, may

last half of may for testing is the best

missing images is possible

Digitalization and Visualization Step

Simple reading with pillow and cropping with pillow (including the radar circle).

Interference Detection Step

Methods Tried

• Traditional CV (Canny Edge Detection, Hough Line Transformation and Custom Rule based Approach)
• Deep Learning
• Azimuth Radial Differential detection + Signal Professing (This worked very well )
  • https://ams.confex.com/ams/101ANNUAL/meetingapp.cgi/Paper/383907