Hidden article scene detection method for smeared image

Abstract

The invention discloses a hidden article scene detection method for a smeared image. The method comprises the following steps: S1, acquiring an image; s2, segmentation of a smearing region: segmenting the smearing region of the image acquired in the S1 by adopting an improved DDRNet-23 segmentation network so as to determine the feature and shape of the smearing region; s3, smear feature recognition: recognizing the segmentation result of the smear region in the S2, and sending the obtained segmentation result to a classification network for classification after Resize preprocessing; and S4, whole image scene identification: zooming the source image corresponding to the positive class sample index in the smear feature identification result in the S3 to the size, and sending the source image to a classification network for identification so as to judge whether the image belongs to the corresponding class. According to the hidden article scene detection method for the smeared image, an image scene is described from multiple dimensions by using a semantic segmentation method and a classification network model, so that the problem of rough analysis description of a single-dimension scene is solved.

Description

technical field [0001] The invention relates to the technical field of image scene analysis, in particular to a hidden object scene detection method for smeared images. Background technique [0002] With the rapid development of the Internet and the comprehensive popularization of smart phones, people's lives have been greatly facilitated. Some of them trade some supplies through contactless methods. In real life, the prohibited items are first placed in a pre-designated place, and then the other party is informed of the actual hiding place by taking pictures with the mobile phone and painting them later, and the other party finds the prohibited items according to the pictures, and finally completes the whole process through online transfer. Trading of Prohibited Items. [0003] Currently, there is no hidden object detection method for smeared images. Most of the existing methods are part of the detection process, including: semantic segmentation of smeared images, and cl...

AI Technical Summary

Problems solved by technology

[0004] 1. A single model can only describe one dimension in the image, and it is difficult to describe the multi-dimensional image scene analysis;

[0005] 2. There is no overall process for hidden object detection, and there is a lack of an overall solution

Images