GNR633Project_QGISPlugin

Created a first iteration (simplistic and preliminary version) of a QGIS Plugin to implement the fuzzy inference system directly within QGIS instead of the current method of exporting to MATLAB, performing computation, and then re-importing into GIS systems.

Features of the plugin realized -

1. Identify Factors : Identifying the factors that affect the system (inputs as number of factors and corresponding number of variables for each of them)

   • NOTE: Assumption here - Hardcoded 2 input factors and 3 variables for each of these 2 factors as well as the output factor.
   • Reason and feasibility of the assumption to be overcome -
     • Reason - Researched a lot but couldn’t figure out how to make a dynamic GUI such that QLabels and QLineEdits appear on the plugin interface. The nearest I could get to a dynamic GUI was pasting in QLineEdits but that wouldn’t help here.
     • Feasibility to overcome - When we can figure out how to make the GUI dynamic.

2. Membership Functions : Provision to label each of the variables for the inputs as well as output factor

3. Membership Functions (contd.) : Choosing a membership function for each of them from the following options - Piecewise Linear, Sigmoid, Gaussian, Quadratic and Cubic polynomial

   • NOTE - Assumption here - Hardcoded Piecewise linear function=> y=0.1x (Domain: [0,10] and Range: [0,1]
   • Hardcoded parameters of the gaussian=> mean = 0 and standard deviation = 1
   • Reason and feasibility of the assumption to be overcome -
     • Reason - Just the time constraint
     • Feasibility to overcome - Can easily modify a few lines of code to allow users to define all parameters of the functions they want.

4. Fuzzy Rules : Provision to add fuzzy rules governing the system.

   • NOTE - Assumption here - Hardcoded provision for only 1 fuzzy rule for now.
   • Reason and feasibility of the assumption to be overcome -
     • Reason - Just the time constraint
     • Feasibility to overcome - Very simple. All I have to do is add to the GUI and put a condition before calculation of the final fuzzy output to add up all of these answers from each fuzzy rule.

5. Feature Vector : Series of steps - Input data —-> calculate the fuzzy values for each factor —-> apply the operation defined in the fuzzy rule —-> get the fuzzy value of the output factor —-> Defuzzify by the centroid location method.

   • NOTE - Assumption here -
   • Defuzzification by the centroid method is implemented only for the piecewise linear - piecewise linear - piecewise linear combination of the 2 input factors and 1 output factor.
   • Reason and feasibility of the assumption to be overcome -
     • Reason - Couldn’t figure out how to implement the centroid method without extensive geometrical considerations and mathematical calculations.
     • Feasibility to overcome - A little unclear to me at the moment, but I believe I can use integration and make the problem simpler in future iterations.

Images of the working plugin -

1. 
2. 
3. 
4.