Image matching apparatus, control method, and non-transitory computer-readable storage medium

A computer based cross-view matching method performed by a computer, comprising:

• acquiring a ground-view image and an aerial-view image;
• extracting features from the ground-view image;
• extracting features from the aerial-view image;
• extracting a plurality of partial aerial regions from the aerial-view image;
• extracting features from each of the plurality of partial aerial regions;
• for each of the plurality of partial aerial regions, computing a combined aerial feature by combining the features of the partial aerial region and the features of the aerial-view image;
• and
• for each of the plurality of partial aerial regions, determining whether the partial aerial region matches the ground-view image by comparing the combined aerial feature of the partial aerial region and the features of the ground-view image,
• wherein the features from the ground-view image, the features from the aerial-view image and the features from each of the plurality of partial aerial regions are extracted based on one or more models, and
• wherein the comparing the combined aerial feature of the partial aerial region and the features of the ground-view image comprises:
• computing a degree of similarity between the combined aerial feature of the partial aerial region and the features of the ground-view image;
• and
• determining that the partial aerial region matches the ground-view image based on a determination that the computed degree of similarity is equal to or larger than a predefined threshold.

A non-transitory computer-readable storage medium storing a program that causes a computer to execute computer based cross-view matching operations comprising:

• acquiring a ground-view image and an aerial-view image;
• extracting features from the ground-view image;
• extracting features from the aerial-view image;
• extracting a plurality of partial aerial regions from the aerial-view image;
• extracting features from each of the plurality of partial aerial regions;
• for each of the plurality of partial aerial regions, computing a combined aerial feature by combining the features of the partial aerial region and the features of the aerial-view image;
• and
• for each of the plurality of partial aerial regions, determining whether the partial aerial region matches the ground-view image by comparing the combined aerial feature of the partial aerial region and the features of the ground-view image,
• wherein the features from the ground-view image, the features from the aerial-view image and the features from each of the plurality of partial aerial regions are extracted based on one or more models, and
• wherein the comparing the combined aerial feature of the partial aerial region and the features of the ground-view image comprises:
• computing a degree of similarity between the combined aerial feature of the partial aerial region and the features of the ground-view image;
• and
• determining that the partial aerial region matches the ground-view image based on a determination that the computed degree of similarity is equal to or larger than a predefined threshold.