Method and device for detecting defects of footwear products based on gaussian splash and wavelet transform

Abstract

The application discloses a kind of based on Gaussian splashing and wavelet transform footwear product flaw detection method and device, it is related to image processing field, including: using frequency domain feature extractor to 3D Gaussian model is trained, obtains the 3D Gaussian model of training, in training process, using mixed optimization target loss function and executing adaptive strategy, adaptive strategy includes the adaptive densification strategy based on frequency domain gradient and the spherical harmonic function order dynamic adjustment strategy of frequency energy drive;Optimize the camera pose of the image to be detected, using the camera pose after optimization to the 3D Gaussian model of training is rasterization rendering, obtains standard reference image, based on standard reference image and the image to be detected wavelet residual map and adaptive threshold are calculated, using adaptive threshold and wavelet residual map generates defect mask, based on defect mask detection obtains flaw detection result.The application effectively solves the problem that small defect on the surface of footwear product is difficult to detect.

Images