Linear object detection method and system based on unmanned plane remote sensing

Abstract

The invention discloses a linear object detection method and system based on unmanned plane remote sensing. The method comprises the following steps: a, obtaining unmanned plane sequence images, and carrying out the extraction of features in a superposed region of the sequence images; b, carrying out the matching of the extracted features, generating a stereopair, and calculating the elevation information according to the stereopair; c, recognizing the edge information of a to-be-recognized target according to the elevation information and the color information of the image, and extracting a to-be-detected target according to the edge information. A data collection scheme is formulated according to the target features, and the sequence images with the high overlapping degree are obtained. A robust SIFT algorithm is employed for feature point matching, and the obtained elevation data is combined with the image features for the recognition of the target edge, thereby achieving the extraction of a conventional target. The method can greatly save the cost, and facilitates the improvement of the detection timeliness and accuracy of a linear target in the images.

Description

technical field [0001] The invention belongs to the technical field of linear target detection, and in particular relates to a linear target detection method and system based on unmanned aerial vehicle remote sensing. Background technique [0002] The extraction of linear objects in remote sensing images is of great significance and extensive research, such as ditches, rivers, roads and power lines, etc. In practical applications, high-resolution satellite images or aerial images are mostly selected for linearization according to the size of linear objects. Target information detection and extraction. There are many linear target detection methods in remote sensing imagery, usually on the image that has been geometrically corrected, the two-dimensional color image appearance features are used to segment, which has certain applicability and limitations; The detection algorithm is based on full search, which requires a lot of calculation and storage space, which is not conduc...

AI Technical Summary

Problems solved by technology

There are many linear target detection methods in remote sensing imagery, usually on the image that has been geometrically corrected, the two-dimensional color image appearance features are used to segment, which has certain applicability and limitations; The detection algorithm is based on full search, which requires a lot of calculation and storage space, which is not conducive to real-time processing; while linear target detection based on mathematical morphology, linear target detection based on region growth and linear target detection based on neural network Although the calculation method is simple and can realize real-time processing, but due to the restriction of the edge detection algorithm to a certain extent, it cannot deal with discontinuous straight lines.

Tracking and detection using high-resolution satellite images and aerial images often requires the purchase of large-scale remote sensing images for identification and extraction, resulting in unnecessary image purchase costs and the burden of a large amount of data processing

Images