Image recognition method and device based on multi-channel fusion and storage medium

Abstract

The invention provides an image recognition method and device based on multi-channel fusion and a storage medium. The method comprises the following steps: an acquisition step: acquiring a plurality of images containing a human oral cavity area of a historical sample; a screening step: screening a specific number of historical images of each historical sample from the plurality of acquired images; a fusion step of performing multi-channel fusion processing on a specific number of historical images screened out from each historical sample to obtain a fused image as an input image of each historical sample so as to establish a training data set; a construction step: constructing an image recognition model based on a deep network algorithm, and training the image recognition model by using the training data set; and an identification step: inputting a to-be-identified image into the trained image identification model to carry out image identification so as to obtain an image identification result. According to the image identification method, multi-channel fusion is carried out on the image, and multi-dimensional feature information can be fully utilized, so that the accuracy of identifying the airway type is improved.

Description

technical field [0001] The present invention relates to the technical field related to computer vision, and more specifically, to an image recognition method, device and storage medium based on multi-channel fusion. Background technique [0002] With the development of machine vision and artificial intelligence technology, artificial intelligence applications based on machine vision have been widely used in various industries, such as face detection, pedestrian tracking, object detection, etc. However, there are still some difficulties in the application of customized artificial intelligence in smart construction sites: on the one hand, artificial intelligence scenarios related to visual applications often require GPU server computing resources to provide computing power support, and the implementation at the deployment end requires the deployment conditions provided by the on-site computer room; on the other hand On the one hand, customized requirements for different applic...

AI Technical Summary

Problems solved by technology

The disadvantages are: 1. The cost is high, and the price of a server plus a graphics card often reaches tens of thousands of yuan; 2. The deployment requirements are high, and the server needs to be deployed in the cabinet of the computer room and the heat dissipation conditions must be guaranteed.

However, the patent only deals with methods for edge computing and does not involve the use of machine models to identify algorithmic aspects

[0004] However, traditional machine learning methods and deep learning methods are usually used to predict airway types, but there are problems such as low recognition accuracy and low reliability, and the value for scientific research and teaching is very limited

The existing literature on airway prediction still has problems such as too small data set size, poor model generalization ability, and low recognition accuracy, and the main reason is that the images used are too single, and only a single image is used to judge whether it belongs to Difficult airway, low reliability

In addition, there is still a lot of room for improvement in image feature extraction and image recognition accuracy in image classification application scenarios.

Images