Salient target detection methods for RGB-D images

By using an attention mechanism and dynamic weight allocation in the feature extraction stage of RGB-D images, the performance of RGB-D salient object detection under low-quality depth maps is addressed, and more accurate salientity map prediction is achieved.

CN116597169BActive Publication Date: 2026-01-06GUANGDONG UNIV OF TECH
2 Cites 0 Cited by

Patent Information

Application Number
CN202310434477.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-04-21
Publication Date
2026-01-06
Estimated Expiration
2043-04-21

AI Technical Summary

Technical Problem

Existing RGB-D salient object detection methods are inadequate in low-quality depth maps, ignoring the similarity between foreground and background and the quality issues of depth maps in low-contrast environments. This results in unnecessary additional information being treated as noise, making it difficult to generate accurate salient masks.

Method used

In the cross-modal fusion of the feature extraction stage, an attention mechanism is used to determine salient regions, and the weights of RGB and depth map features are dynamically assigned through feature concatenation and convolution operations to form a salient target detection model.

Benefits of technology

It improves the performance of salient target detection in low-quality depth map conditions, and can detect salient targets better and cope with scenes with low depth map contrast.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The present invention aims to provide a salient target detection method for RGB-D images, comprising: sampling RGB image and depth map samples at a specified number to form sampling data, and preprocessing the sampling data using data augmentation techniques to obtain data to be processed; in the cross-modal fusion of each layer in the feature extraction stage, using an attention mechanism to infer salient regions to determine the saliency degree of different regions, and fusing at each layer to obtain low-level features and high-level features; performing concatenation and convolution operations on the high-level features and the low-level features to obtain joint features of RGB features and depth features; dynamically allocating the weights of RGB image and depth map features according to the joint features of RGB features and depth features to obtain a salient target detection model; and realizing the salient target detection result according to the salient target detection model. The method described in this invention can detect salient targets well and handle scenes with low depth map contrast.
Need to check novelty before this filing date? Find Prior Art