Multi-modal saliency object detection method based on coding and decoding structure

Abstract

The invention provides a multi-modal saliency object detection method based on a coding and decoding structure. On the basis of an existing color image algorithm model, a depth image is used as priorinformation to be supplemented into an algorithm. The method provided by the invention can be divided into two parts: one part takes a color image as input and realizes saliency detection based on a deep learning technology of a coding and decoding structure; and the other part is used for processing the depth image, learning the depth features by using a lightweight neural network, and supplementing the feature information to the first part, thereby improving the recognition precision of the overall model. According to the method, the depth image can be simply and efficiently utilized, features are learned from the depth image, the recognition precision and stability of the algorithm are enhanced, and a quick and low-cost updating means can be provided for an existing deployed saliency detection algorithm.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and is especially aimed at the task of image salient object detection, and in particular relates to a multimodal salient object detection algorithm based on a codec structure and combining color images and depth images. Background technique [0002] Computer vision technology is an important branch of the field of artificial intelligence. With the rapid development of neural networks and deep learning, many problems in the field of computer vision, such as image classification, target detection, and image segmentation tasks, have made great progress. The accuracy of the algorithm , The real-time performance has been greatly enhanced. At present, computer vision technology has been widely used in actual production and life, especially in the fields of security monitoring, payment, and intelligent manufacturing, to promote social production cost reduction and efficiency improvement, and imp...

AI Technical Summary

Problems solved by technology

At present, the main challenges are divided into two points. On the one hand, new algorithms need to be re-developed to jointly process color images and depth images; on the other hand, simple and effective strategies are needed to integrate depth information into existing algorithms.

Images