US 12,450,907 B2
Smoke detection method based on deep learning, device and storage medium
Yingying Zhu, Guangdong (CN); and Junqi Zhao, Guangdong (CN)
Assigned to SHENZHEN UNIVERSITY, Shenzhen (CN)
Filed by SHENZHEN UNIVERSITY, Guangdong (CN)
Filed on Feb. 17, 2023, as Appl. No. 18/110,903.
Claims priority of application No. 202210901046.5 (CN), filed on Jul. 28, 2022.
Prior Publication US 2024/0037950 A1, Feb. 1, 2024
Int. Cl. G06V 20/52 (2022.01); G06T 5/20 (2006.01); G06T 5/50 (2006.01); G06T 7/11 (2017.01); G06T 7/174 (2017.01); G06T 7/90 (2017.01); G06V 10/26 (2022.01); G06V 10/56 (2022.01); G06V 10/77 (2022.01); G06V 10/774 (2022.01); G06V 10/82 (2022.01); G06V 20/40 (2022.01)
CPC G06V 20/52 (2022.01) [G06T 5/20 (2013.01); G06T 5/50 (2013.01); G06T 7/174 (2017.01); G06V 10/26 (2022.01); G06V 10/7715 (2022.01); G06V 10/774 (2022.01); G06V 20/41 (2022.01); G06T 2207/10016 (2013.01); G06T 2207/10024 (2013.01); G06T 2207/20021 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/20224 (2013.01)] 9 Claims
OG exemplary drawing
 
1. A smoke detection method based on deep learning, comprising:
acquiring a smoke image and an indoor image, wherein the smoke image comprises image transparency;
determining a size and a position relationship of the smoke image in the indoor image, and performing image enhancement processing on the smoke image based on the size and the position relationship of the smoke image in the indoor image to obtain a basic smoke image;
acquiring an illumination image of an indoor scene at a corresponding position of the basic smoke image, and performing color transfer processing on the basic smoke image based on the illumination image to obtain a color transferred smoke image;
superimposing the color transferred smoke image on the indoor image based on the smoke transparency to obtain an initial image, and performing screening processing and detection frame updating processing on the initial image to obtain a target image;
performing feature extraction on the target image through a multilayered network to obtain a smoke feature image, and performing layer calibration processing on the smoke feature image to obtain a smoke calibrated feature image;
performing segmentation processing, detection frame prediction processing, detection frame classification processing and image classification processing on the smoke calibrated feature image in sequence according to a step-by-step feature learning mode to obtain a smoke image set, wherein the smoke image set comprises confidence degrees and union-intersection parameters of smoke features; and
screening the smoke image set based on the confidence degrees and the union-intersection parameters of the smoke features to obtain a target smoke image, and outputting a smoke detection result based on the target smoke image.