Wood floor defect detection method based on cascade centroid algorithm

The invention discloses a method for detecting surface defects of a wood floor. The method comprises the following specific processing steps: scaling and standardizing an acquired wood floor image, extracting features by applying depth separation and convolution, inputting the extracted features into a specific calculation module for processing to realize further feature extraction and feature fusion, applying non-maximum shielding to a generated prediction frame, and performing prediction on the obtained prediction frame. And screening the wood floor defect bounding box by a confidence shielding method. The method is based on the structure of a target detection algorithm commonly used for defect detection, improvement suitable for wood floor surface defect detection is carried out, and the method comprises the steps of designing C-ELAN, R-SAM, Mini-Head, RCU and CCC-NORM modules or methods, so that the beneficial effects of improving the accuracy rate of identifying defect categories and the positioning prec