Unlock instant, AI-driven research and patent intelligence for your innovation.
Refined target identification method and system
What is Al technical title?
Al technical title is built by PatSnap Al team. It summarizes the technical point description of the patent document.
A target recognition and fine technology, applied in the field of fine target recognition methods and systems, can solve the problems of a large amount of time and manpower, poor portability, and difficulty in providing manual labeling information for large data sets, so as to improve recognition efficiency, save time and human effect
Active Publication Date: 2017-06-13
CAPITAL NORMAL UNIVERSITY
View PDF1 Cites 11 Cited by
Summary
Abstract
Description
Claims
Application Information
AI Technical Summary
This helps you quickly interpret patents by identifying the three key elements:
Problems solved by technology
Method used
Benefits of technology
Problems solved by technology
[0004] However, in the existing technology, it is necessary to strongly label the attributes and local feature points of the local area, which requires a lot of time and manpower when performing strong labeling, and only a few small data sets provide manual strong labeling information, and most large data sets It is difficult to provide manual annotation information, so the existing methods are poor in portability
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 more
Image
Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
Click on the blue label to locate the original text in one second.
Reading with bidirectional positioning of images and text.
Smart Image
Examples
Experimental program
Comparison scheme
Effect test
Embodiment 1
[0074] The fine target recognition method provided in this embodiment does not need to artificially annotate images strongly, and can detect fine targets and extract distinguishable regions in an unsupervised situation, saving a lot of time and manpower.
[0075] see figure 1 , the present embodiment provides a fine target recognition method, the method comprising:
[0077] Step S101 further includes step S201 to step S203:
[0078] Step S201, changing the size of the image to be recognized to a preset size;
[0079] Step S202, training the DomainNet convolutional neural network model to extract the pool5 layer features of the image to be recognized under the preset size, and obtaining the feature description of the image to be recognized;
[0080] In step S203, the image to be recognized is processed through the feature description of the image to b...
Embodiment 2
[0126] see image 3 , a fine target recognition system provided in this embodiment, the system includes:
[0127] Target saliency map generation module 30, used to extract the feature description of the image to be recognized, and generate the target saliency map;
[0128] The target candidate region determination module 31 is used to process the image to be recognized through the target saliency map to obtain the target candidate region of the image to be recognized;
[0129] The K nearest neighbor image retrieval module 32 is used to retrieve the K nearest neighbor images of the image to be identified, and processes the K nearest neighbor images to obtain the target candidate area of the K nearest neighbor images;
[0130] Similarity calculation module 33, for calculating the similarity of the target candidate area of the image to be recognized and the target candidate area of the K nearest neighbor image;
[0131] The fine target area determination module 34 is conf...
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 provides a refined target identification method and system, and relates to the technical field of image processing. The method comprises the following steps that: extracting the feature description of an image to be identified, and generating a target saliency map; through the target saliency map, processing the image to be identified to obtain the target candidate area of the image to be identified; retrieving the k-nearest neighbor image of the image to be identified, and processing the k-nearest neighbor image to obtain the target candidate area of the k-nearest neighbor image; calculating a similarity between the target candidate area of the image to be identified and the target candidate area of the k-nearest neighbor image; and according to the sum of the similarity and the similarity between the k-nearest neighbor image and the target candidate area, determining the refined target area of the image to be identified. Therefore, the refined target can be described without a situation that the image to be identified needs to be subjected to strong marking, a great quantity of time and manpower can be saved, and refined target identification efficiency is improved.
Description
technical field [0001] The present invention relates to the field of image processing, in particular to a fine target recognition method and system. Background technique [0002] Since the fine target has a small gap between classes, accurate local description is very important for fine target recognition, while the traditional algorithm requires complete manual labeling information in the image to be recognized, but the acquisition of manual labeling information is time-consuming and laborious. It is applied to large-scale image recognition in real life, so how to more easily recognize fine objects has become an urgent problem to be solved. [0003] Related technologies mainly use strong annotations to identify fine objects. First, use the strong annotation information in the training image to predict the local area that may contain objects in the test image; then use the convolutional neural network model to extract the depth features of the local area. ; Finally, feature...
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
Application Date:The date an application was filed.
Publication Date:The date a patent or application was officially published.
First Publication Date:The earliest publication date of a patent with the same application number.
Issue Date:Publication date of the patent grant document.
PCT Entry Date:The Entry date of PCT National Phase.
Estimated Expiry Date:The statutory expiry date of a patent right according to the Patent Law, and it is the longest term of protection that the patent right can achieve without the termination of the patent right due to other reasons(Term extension factor has been taken into account ).
Invalid Date:Actual expiry date is based on effective date or publication date of legal transaction data of invalid patent.
Login to View More 
Login to View More