Tracking in 2D: OISHI features

To locate the fish in a 2D image, we are looking at several points that “looks like the centre of a fish”.

I call these centres oishi features, because they should be Orientation Invariant and SHape Invariant.

How to find these features? Shortly speaking, type these

import fish_track as ft

kernels = ft.kernel.get_kernels(shapes, axis_indices, cluster_number)
oishi_kernels = ft.oishi.get_oishi_kernels(kernels, 35)
features = ft.oishi.get_oishi_features(image, kernels)

But what is happening? What are the meanings of those variables? What is the business happening inside the code?

Here is an detailed explaination.

Find a Shape in a Image

under construction …

Finding reprsentative Shapes

under construction …

Rotate the Kernels

under construction …

Getting the OISHI features

under construction …

Refining the OISHI features

under construction …