VR video real-time detection method for garden monitoring

Abstract

The invention discloses a VR video real-time detection method for garden monitoring, and the method comprises the steps: firstly carrying out the noise reduction of several non-fused images at different angles photographed by a VR camera by adopting a quick median filtering method, and carrying out the color conversion by adopting a DiagonalOffset color conversion model, so as to remove the impacton a picture from illumination; carrying out color mean value processing based on spatial correlation on the picture, extracting a smoke candidate region by using an RGB color model, extracting RGB static characteristics of smoke, extracting motion characteristics of smoke by using an improved frame difference method, carrying out hole filling on a result, and carrying out shadow elimination on the improved frame difference method by combining the RGB color model; extracting texture characteristics of a smoke motion area by using a direction gradient histogram (HOG) and a local binary pattern(LBP); training an SVM classifier, and then performing classification; and finally, performing post-processing, determining the smoke movement direction, speed and diffusion degree, and obtaining various kinds of smoke diffusion information.

Description

Technical field[0001]The invention relates to a real-time detection method of VR video for garden monitoring, belonging to the field of computer vision and garden-oriented intelligent VR monitoring.Background technique[0002]With the development of information technology, Internet+ thinking runs through all walks of life. Smart gardens are to integrate emerging information technologies such as the Internet of Things, smart terminals, and big data cloud computing with modern ecological gardens to achieve mutual inductance, mutual knowledge, and interaction between man and nature. However, there are still many problems or pain points in the process of merging emerging technologies and ecological gardens. The invention proposes a VR video real-time detection method for garden scenes based on the computer vision field and the smart garden field, and specifically relates to the detection, tracking and evaluation of smoke in the smart garden scene.[0003]Traditional fire detectors usually j...

AI Technical Summary

Problems solved by technology

[0004] Video image fire detection has broad application prospects in practical applications, and some achievements have been made at present. B.U. Toreyin et al. proposed a real-time detection of video smoke based on wavelet transform; ZhouBL et al. proposed a video smoke detection algorithm that combines dynamic and static features , the algorithm can quickly and accurately detect fire smoke, but when there are non-smog objects similar to smoke, false alarms will be generated, and it is difficult to distinguish between smoke and false smoke; L.Wang proposed a fire smoke based on multi-feature fusion of video smoke In the early identification method, the Gaussian mixture model and the background difference method were used to extract the suspicious area of ​​the smoke in the foreground, and then the digital image processing technology was used to extract the color, background blur, contour irregularity and motion of the suspicious area, and finally the feature vector was formed as a support Vector machine (Support Vector Machine SVM) classifier input, while using particle swarm optimization algorithm to optimize support vector machine (SVM) classifier to improve the detection accuracy of smoke; Automatically extract the deep features of the smoke target area and then classify it to complete the identification of fire smoke

Images