An Image Retrieval Method Based on Deep Learning and Approximate Object Localization
A target positioning and image retrieval technology, applied in the field of deep learning and image retrieval of approximate target positioning, can solve problems such as unsatisfactory results, and achieve the effect of improving description ability, improving accuracy, and improving accuracy.
Active Publication Date: 2021-12-17
XIAN UNIV OF TECH
View PDF4 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Although CroW can solve most image retrieval problems, the effect can only be unsatisfactory when distinguishing complex backgrounds
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0063] The present invention will be further elaborated below in conjunction with the accompanying drawings and specific embodiments.
[0064] Frame diagram of image retrieval algorithm for deep learning and approximate target localization, such as figure 1 As shown, from the input and output of the algorithm, the present invention inputs two image libraries (query image library, image library to be processed), wherein the two images are weighted through the F-CroW process in the feature extraction and weighting stages to obtain N a similar target area.
[0065] From the perspective of algorithm flow, the present invention generally includes three steps: feature extraction and weighting, approximate target location and image reordering. In the feature extraction and weighting stage, the activation map generated by the last layer of the 'pooling' layer in the convolutional neural network convolution layer is extracted, and the extracted features are subjected to spatial weight...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The invention discloses an image retrieval method for deep learning and approximate target positioning. Focus on searching the approximate object area in the image; use the sliding window of multiple scales on the image to extract the corresponding local CNN features; use these small feature blocks to obtain the feature vector of the image, this image feature can already meet the requirements of image retrieval; in After processing the features, an optimized image representation with several weights has been prepared for image retrieval and rearrangement; a new framework called Approximate Object Location is constructed to search for the most similar domain in the queried image that can be retrieved ; For the retrieval of a specific object, a series of small feature blocks are used to perform approximate target positioning on the database image, and the image retrieval process is completed with the localized area. The invention makes full use of the powerful ability of deep learning, optimizes and innovates on the basis of CNN, and reduces redundancy while improving image retrieval accuracy.
Description
technical field [0001] The invention belongs to the technical field of image analysis and retrieval methods, and in particular relates to an image retrieval method for deep learning and approximate target positioning. Background technique [0002] In recent years, with the continuous promotion and development of the machine intelligence industry and the Internet + industry, image retrieval has attracted more and more attention from researchers, and has gradually penetrated into the bottom layer of neural networks, computer networks, machine learning and other fields, and has become an important tool for the society. The backbone of services, industrial upgrading, etc. [0003] At present, image retrieval in most industrial-grade applications still stays in the time-consuming and labor-intensive method based on text or keywords. Of course, content-based image retrieval (Content-based Image Retrieval) has been widely deployed in more advanced search engines, but often a singl...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & Authority Patents(China)
IPC IPC(8): G06F16/583G06N3/04
CPCG06N3/045
Inventor 廖开阳邓轩郑元林汤梓伟袁晖
Owner XIAN UNIV OF TECH