Image key point detection method based on feature pyramid network

Abstract

The invention belongs to the technical field of computer image processing, and particularly relates to an image key point detection method based on a feature pyramid network. The method comprises thefollowing steps: extracting high-characterization image features through a feature pyramid network, wherein the high-characterization image features are robust to scale, visual angle geometric transformation, illumination, blurring and the like; generating a training data set suitable for key point detection; in a training phase, taking the grayscale image as the input of the network model, initializing network model parameters by using weight parameters obtained by pre-training on an ImageNet data set, performing fine adjustment on the network parameters by using a training data set, and finally outputting a probability graph with the same size as an input image, wherein each value in the graph is between 0 and 1, and the larger the value is, the more suitable the point is as a key point;in a test stage, using a non-maximum suppression algorithm to prevent points with large response values from accumulating in a small area, and setting threshold values with different sizes to controlthe number of key points. The quality of the key points is ensured.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and in particular relates to an image key point detection method. Background technique [0002] Computer vision technology is based on perceiving images to make useful decisions about objective objects and scenes. Key point detection, also known as feature point or interest point detection technology, is a key technology in the field of computer vision and is applied to many tasks such as image matching, image retrieval, and visual simultaneous positioning and mapping. The key point usually refers to the more prominent and highly distinguishable pixel points or image area blocks in the image. As one of the most important local features in the image, it has rotation invariance, viewing angle invariance, and scale invariance. And many other excellent properties, so it is widely used in various computer vision tasks. [0003] Although the research on the key point detection problem has mad...

AI Technical Summary

Problems solved by technology

These algorithms are manually designed based on a specific environment, so they lack certain generalization and cannot adapt to the transformation of the environment well.

Images