Sky Detection Method Based on Contextual Reasoning

Abstract

The invention discloses a sky detection method based on context reasoning, including the establishment of sky sample data sets, scene segmentation algorithms, superpixel feature extraction, classifier training and the establishment of context reasoning models, which can be used to train classifiers for preliminary detection of sky regions ; Established the CRF context reasoning model, using context constraints to further improve the detection accuracy, and achieved higher detection accuracy than the existing similar algorithms; a good compromise was achieved in the detection accuracy and detection speed, which can meet the actual needs.

Description

technical field [0001] The invention belongs to the technical field of scene understanding, and relates to a sky detection method based on context reasoning. Background technique [0002] With the development of computer and robot technology, intelligent robots have been widely used not only in manufacturing, but also in military, civilian, scientific research and many other aspects. This paper discusses in detail the important preprocessing step in vision systems such as ground robots—sky detection—and proposes corresponding algorithms. Sky regions are an important part of outdoor images and provide important information about the surrounding environment. In a ground-based autonomous robot developed by Stanford University, a simple sky recognition algorithm improves road detection. The existing sky detection algorithms mainly include methods based on color prior, methods based on gradient prior and methods based on model fusion. Color prior-based sky recognition algorith...

Images