Camouflage object detection model based on edge collaborative supervision and multi-level constraint

Abstract

The invention discloses a camouflage object detection model based on edge collaborative supervision and multi-level constraint, provides a mature and complete camouflage object detection method based on the model, preliminarily searches a target potential area through graph-level and pixel-level search, and strengthens spatial response through cascade constraint and an attention mechanism. Meanwhile, an edge segmentation branch is established to guide the model to predict a more accurate contour, a Frelu activation function is utilized to extract the spatial activity of an image at a basic convolution part, camouflage object detection can be effectively and accurately carried out through the model, and a gap is filled in the field of camouflage object detection in the prior art.

Description

technical field [0001] The invention relates to the technical field of computer vision, and more specifically relates to a camouflaged object detection model and method based on edge cooperative supervision and multi-level constraints. Background technique [0002] Camouflaged object detection is a task that has gradually started in the field of computer vision in recent years. It is intended to separate the target object disguised in the scene from the background. This binary classification semantic segmentation technology is similar to salient object detection, but because the Due to the high similarity in color and texture, and the high complexity of the scene, the detection of camouflaged objects is much more difficult than the currently developed salient object detection, and the relevant data and algorithms are extremely scarce. [0003] Therefore, how to propose a camouflaged object detection model and method based on edge cooperative supervision and multi-level const...

AI Technical Summary

Problems solved by technology

[0002] Camouflaged object detection is a task that has gradually started in the field of computer vision in recent years. It is intended to separate the target object disguised in the scene from the background. This binary classification semantic segmentation technology is similar to salient object detection, but because the Due to the high similarity in color and texture, and the high complexity of the scene, the detection of camouflaged objects is much more difficult than the currently developed salient object detection, and the relevant data and algorithms are extremely scarce.

Images