Shadow Detection Method Based on Orthogonal Decomposition and em Algorithm

Abstract

The invention relates to a shadow detection method based on orthogonal decomposition and an EM algorithm. The shadow detection method includes the following steps of: establishing a system of linear equations through linear models inside and outside shadow areas in an original image; performing orthogonal decomposition on the system of linear equations to obtain a colored illumination-unchangeable image and an illumination-changeable image; classifying the colored illumination-unchangeable image through a K-means algorithm; performing Gaussian mixture model on the illumination-changeable image through the EM algorithm according to a classification result, and extracting the shadow areas; and finally performing optimization on the extracted shadow areas through a morphology operator. According to the method, the shadow areas can be extracted through the simple orthogonal decomposition and the simple EM iterative algorithm, and a complicated characteristic operator learning process can be avoided, so that the time complexity of the algorithm can be greatly reduced, and the method can be directly applied to real-time occasions; and prior knowledge such as scenes and targets is not needed, so that the method has the good universality.

Description

technical field [0001] The invention relates to the fields of computer vision and image processing, in particular to a shadow detection method for a single outdoor image based on orthogonal decomposition and EM algorithm. Background technique [0002] As a ubiquitous physical phenomenon in nature, shadows bring many adverse effects to computer vision tasks. The coverage of shadows will cause image blur and destroy the continuity of gray values, which will affect the robustness of edge detection, object recognition, and image matching algorithms, and will greatly interfere with subsequent image analysis and understanding. According to the number of images used, shadow detection can be divided into multi-image based methods and single image based methods. At present, many scholars at home and abroad have conducted in-depth research on shadows in image sequences, and proposed many effective algorithms. These methods have been widely used in the fields of video surveillance and...

AI Technical Summary

Problems solved by technology

This type of algorithm often only considers the difference of a single feature of the target shadow, and the detection efficiency is not high

For example, many algorithms use the feature that the brightness of the shadow area is lower than that of the surrounding area for shadow detection, but the darker area is not necessarily a shadow; there are also algorithms that simply rely on chromaticity invariance for detection, but when the shadow color is darker, use Chroma invariance makes it impossible to detect shadows

In recent years, the multi-feature shadow detection algorithm based on statistical learning has received more and more attention. The algorithm has better versatility and robustness, but they often require a complex learning process, and the calculation of the algorithm data is large. Time-consuming and difficult to apply to real-time occasions

Images