Abstract The present invention provides an image recognition method whereby the influence of noise due to fluctuations of the ambient environment can be minimized and which has good response and high recognition accuracy, without using binary processing. In the image recognition of the present invention, search models where the shape and luminance value distribution of an object are modeled, are stored as image data in advance, and when multi-grayscale image data including the object is input (Step 1 ), the plurality of search models are distributed in the image sensing area of this input image data (Step 2 ), the goodness-of-fit for indicating correlation with the object is calculated for each one of the search models (Step 3 ), the search models are rearranged based on the goodness-of-fit (Step 5 ), selected (Step 6 ), crossed and regenerated (Step 7 ), and mutated and regenerated (Step 8 ), and through executing this genetic control, the optimum solution of the position and orientation information of the object is searched.
An image recognition method comprising the steps of: storing search models used for searching area, where all or a part of shape and luminance value distribution of an object are modeled in an area state, as image data in advance; obtaining input image data including the object; distributing the search models in the area of said input image data, and assigning the position and orientation information to each one of said search models as individual information; determining a function value to indicate a correlation with said object for each one of said search models; and searching the solution of at least the position information of the object from said input image data by evolving said search models using a genetic algorithm based on said function values.
Claim 2
Dependent Claim
126 words
An image recognition method comprising the steps of: storing search models used for searching area, where all or a part of shape and luminance value distribution of an object are modeled in an area state, as image data in advance; obtaining input image data including the object; distributing the search models in the area of said input image data, and assigning the position and orientation information to each one of said search models as individual information; determining a function value to indicate a correlation with said object for each one of said search models; and searching the solution of at least the position information of the object from said input image data by evolving said search models using a genetic algorithm based on said function values.
Claim 3
Dependent Claim
26 words
The image recognition method according to claim 1 , wherein a plurality of search models are randomly distributed in the area of said input image data.
Claim 4
Dependent Claim
41 words
The image recognition method according to claim 1 or claim 2 , wherein a sum of values determined from the luminance values of said input image data in the internal area of the search model is used for said function value.
Claim 5
Dependent Claim
44 words
The image recognition method according to claims 1 or 2 , wherein constraints are assigned to the evolution of the position and orientation information of said search models so as to limit the search range of the object included in said input image data.
Claim 6
Dependent Claim
43 words
The image recognition method according to claims 1 or 2 , wherein the position and orientation information of a search model having a highest function value at a desired control timing is regarded as an optimum solution without waiting for convergence of evolution.