An algorithm for registering a camera image with a database image
Project description
This module is a pure-Python implementation of the Iterative Closest Point algorithm for registering a camera image with a database image.
An application scenario would be the registration of an image recorded by a UAV-mounted camera flying over a terrain with an image extracted from GIS (Geographical Information System) database.
Version 1.1 includes a new option for the constructor that lets the system decide as to which image to use as the model and which image as the data. In general, for color and grayscale images, you get superior registration if the image that results in fewer pixels for ICP processing is used as the data image.
Typical usage syntax:
import ICP
- icp = ICP(
model_image = “modelterrain.jpg”, data_image = “cameraimage.jpg”, binary_or_color = “color”, iterations = 40, connectivity_threshold = 5, output_image_size = 100, display_step = 1, debug = 1 )
icp.icp()
icp.display_results()
The module also includes a static method gendata() to illustrate how one can create simple synthetic images to experiment with the code. A call to this method looks like
import ICP
ICP.ICP.gendata(“triangle”, (80,80), (10,10), 30, “newtriangle2.jpg” )
for creating an 80x80 image of a triangle whose position offset is (10,10) and whose orientation is 30 degrees.
