Ship target confirmation method based on local shape matching

Abstract

The invention discloses a ship target confirmation method based on local shape matching, and the method comprises the steps: obtaining a region image of a target detection frame in an original image predicted by a neural network detection model, carrying out the preprocessing, and extracting a maximum outer contour diagram of a target binary image in the preprocessed image; obtaining a binarized image of the template image, and extracting a maximum outer contour diagram of the binarized image of the template image; respectively carrying out point sampling on the two maximum outer contour diagrams to obtain a point set of a prow / tail of a template image and a point set of a left / right area of a target binary image; and constructing a feature descriptor of a bow and a stern of the template image and feature descriptors of left and right areas of the target image, performing local shape matching, and finally determining whether the area image of the target detection frame in the original image is a ship target. According to the method, the problem that a large number of false alarms exist in a detection result due to the uncertainty of a deep neural network detection model result can be effectively solved; and the method has the advantages of fast speed, low cost and high precision.

Description

technical field [0001] The invention belongs to the technical field of digital image processing, and relates to a method for eliminating false alarms based on the detection results of a neural network target detection model, in particular to a method for confirming a ship target based on local shape matching. Background technique [0002] Remote sensing technology has now penetrated into all aspects of our lives, providing effective means for ground object observation, area of ​​interest monitoring, remote sensing target detection and recognition, etc., especially in the aspect of target detection and recognition in remote sensing images. Many aspects play an important role. In recent years, thanks to the support of a large amount of data, neural network models have developed rapidly. Various neural network models have surpassed traditional hand-designed algorithms in many aspects, which provides important technical support for remote sensing target detection and recognition...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to solve a large number of virtual scene problems caused by the neural network detection model, and to meet the fast operation of a lower hardware environment, a local based A ship target confirmation method based on shape matching, which has a small amount of calculation and high accuracy, and can be well deployed on various low-performance hardware platforms

Images