An Object Detection Method Based on Object Space Knowledge and Two-Stage Predictive Learning

Abstract

The invention discloses a target detection method based on target space knowledge and two-stage predictive learning, using various data transformation methods to increase the number of samples, increase the diversity of samples, and train SSD and the newly designed RefineNet two deep neural networks network; for prediction targets with higher probability in the preliminary prediction results of SSD, the accuracy of discrimination is further improved through RefineNet; by formulating target-specific spatial structure constraint rules to reduce wrong predictions, the final detection results are obtained. Compared with some existing methods, the present invention takes into account the visual characteristics and spatial characteristics of remote sensing targets at the same time, and uses a deep network with excellent feature extraction capabilities to realize end-to-end target candidate, feature extraction and classification positioning, which significantly improves remote sensing. The detection rate of the target is reduced, and the false alarm rate is reduced.

Description

technical field [0001] The invention belongs to the technical field of optical remote sensing image processing, relates to applications in the field of image target detection, and in particular to a target detection method based on target space knowledge and two-stage predictive learning. Background technique [0002] Object detection is a fundamental problem in the field of aerial and satellite imagery analysis and plays a vital role in numerous applications such as environmental monitoring, geological hazard monitoring, land use and cover mapping, geographic information system updates, precision agriculture, and urban planning. [0003] Looking back at the development of object detection in optical remote sensing images, there are four main types of methods: object detection based on template matching, object detection based on knowledge, object detection based on object image analysis, and object detection based on machine learning. At present, with the development of aer...

AI Technical Summary

Problems solved by technology

[0005] The technical problem to be solved by the present invention is to provide a target detection method based on target space knowledge and two-stage predictive learning, which is used for feature extraction and target detection of high-resolution optical remote sensing images. The existing high-resolution optical remote sensing image target detection methods lack effective feature extraction, and the problems of low detection rate and high false alarm rate for large scene targets

Images