Image anomaly detection method based on depth feature histogram

The invention discloses an image anomaly detection method based on a depth feature histogram. The method comprises the steps of collecting normal images to construct a training set and performing preprocessing; performing feature extraction on the preprocessed training set image; performing feature selection on the extracted features; constructing a histogram anomaly detection model for the selected features; obtaining an anomaly score graph of the to-be-detected image through the anomaly detection model; and performing post-processing on the abnormal score graph to obtain an abnormal detection and positioning result of the to-be-detected image. Compared with an existing method, the method has the advantages that a histogram anomaly detection model is constructed on the channel dimension of the image depth features, dependence on pixel positions is avoided, the detection effect of an algorithm in an image non-alignment scene is improved, and occupied storage space is greatly reduced. Experimental results show