Motion blur target detection method based on vehicle motion and data augmentation

• Main Authors: ZHANG LIJUN, MENG DEJIAN, HAN YE, HUANG LUYING
• Format: Patent
• Language: Chinese; English
• Published: 02.03.2021
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; HANDLING RECORD CARRIERS; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention relates to a motion blur target detection method based on vehicle motion and data augmentation, and the method comprises the following steps: S1, building a joint model of a vehicle kinematics model and a camera linear model, and obtaining the distribution of blur parameters through simulation; S2, acquiring an image data set, and performing polar coordinate transformation to obtainan image under a polar coordinate system; S3, according to the fuzzy parameter distribution, degenerating the image under the polar coordinate system to obtain a degenerated image matrix; S4, converting the degraded image matrix into a Cartesian coordinate system to obtain augmented image data; S5, performing neural network training by using the original image data set and the augmented image datato obtain a trained motion blur target detection network; S6, inputting the image shot by the camera into the trained motion blur target detection network, and carrying out motion blur target detection. Compared with the prio