Image Processing Techniques for Extraction of Wind Fields from Lidar Aerosol Backscatter

Open Access
Author:
Bickel, Brady Richard
Graduate Program:
Electrical Engineering
Degree:
Master of Science
Document Type:
Master Thesis
Date of Defense:
July 13, 2012
Committee Members:
  • Timothy Joseph Kane, Thesis Advisor
Keywords:
  • aerosol lidar
  • bilateral filter
  • registration
  • backscatter
  • nonlinear filter
  • correlation
Abstract:
The tracking of winds and atmospheric features has many applications, from pre- dicting and analyzing weather patterns in the upper and lower atmosphere to monitoring air movement from pig and chicken farms. Doppler lidar systems exist to quantify the underlying wind speeds, but cost of these systems can sometimes be relatively high, and processing limitations exist in these systems. There is also interest in deployment of these systems where Doppler lidar cannot be easily used. Currently, image correlation techniques are used to calculate wind vectors with lidar aerosol backscatter. By using data from a simple aerosol lidar system, we attempt to extend these processing methods to improve the resulting correlations. By preconditioning the aerosol backscatter im- agery, wind calculation uncertainty can be reduced, leading to potential bene ts such as reduction of noise sensitivity, and shorter lidar scan times. Image thresholding and bilat- eral ltering are applied to the backscatter imagery, and it is shown how these techniques help to improve the correlation quality. Also, use of an image registration method for wind calculation is tested and results are compared to the correlation techniques. Use of these preprocessing and registration methods show promise in the improvement of the wind detection problem using inexpensive aerosol lidar setups with sophisticated processing algorithms.