An image processing method for expanding a data set under a small sample

An image processing and sample image technology, applied in the field of image processing, can solve the problems of increasing difficulty in image recognition and inability to recognize, and achieve the effect of increasing a reasonable number of samples and high recognition accuracy

Inactive Publication Date: 2019-02-12
成都网阔信息技术股份有限公司
View PDF5 Cites 26 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] The object of the present invention is to provide an image processing method for expanding data sets under small samples, which solves the problem of insufficient samples, and the image features in the training stage are not enough to effectively represent the changes in image features, thereby increasing the difficulty of image recognition, even Problems with unrecognized phenomena

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
  • An image processing method for expanding a data set under a small sample
  • An image processing method for expanding a data set under a small sample

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0032] An image processing method for expanding data sets under small samples, including small sample sets, further comprising the following steps:

[0033] S1. Classify the samples in the small sample set;

[0034] S2. For the classification in step S1, formulate an image transformation strategy for each category;

[0035] S3. Transform the samples in the small sample set according to the transformation strategy formulated in step S2 to obtain virtual samples, and add both the virtual samples and the original samples to the large sample set to realize the expansion of the sample set.

[0036] In order to solve the problem of poor recognition accuracy of small-sample images, the present invention adopts sample expansion technology to overcome the problems caused by insufficient samples, and provides a more general image processing method. This method translates, rotates, mirrors, zooms, perspectives, transforms contrast and brightness, adds noise, and blurs the original train...

Embodiment 2

[0038] The difference between this embodiment and Embodiment 1 is that further, the classification criteria in step S1 include:

[0039] S101: Determine whether the sample image includes the target image, and classify the sample image into two types: images including the target and images not including the target;

[0040] S102: Carry out feature judgment on the images classified in step S101, determine the proportion of the feature image in the sample image, and classify again the images including the target and the images not including the target according to the proportion.

[0041] Further, the transformation strategy of the image in the step S2 includes:

[0042] S201, setting the number of transformations: controlling the number of extensions by setting the sampling rate for each transformation, and controlling the number of virtual samples generated by the transformation according to the requirements of different training tasks;

[0043] S202. Setting the degree of tra...

Embodiment 3

[0050] Let's take the samples required for training to recognize a driver's phone call as an example. All sample image data are 32*32 in size, and the samples in the training set are divided into four categories according to the specific characteristics of the samples, and customized image transformation processing is performed respectively.

[0051] S103 indicates that among the heads of people making calls, the size of the head is relatively small, that is, there is room for panning around the head, and the ear or mobile phone will not be panned out of the picture.

[0052] S104 indicates that among the heads of the people making the call, the size of the head is relatively large, that is, there is almost no extra space for translation around the head.

[0053] S105 indicates that among the heads of people who did not make a call, the size of the head is relatively small;

[0054] S106 indicates that among the heads of the people who did not make a call, the head size is re...

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

No PUM Login to view more

Abstract

The invention discloses an image processing method for expanding a data set under a small sample. In order to solve the problem of poor identification accuracy of a small sample image, the invention adopts a sample expansion technology to overcome the problem caused by insufficient samples, and provides a more general image processing method. The method comprises the steps that the original training samples are translated, rotated, mirrored, scaled and transmitted, the contrast and the brightness are transformed, the noise is added, the virtual training samples are generated, the number of training samples is increased by generating virtual samples, and then the virtual samples are fused with the original training samples. Through a large number of experiments, the method of the inventionhas excellent recognition effect on a small sample training set, and has better recognition performance. If the sample is insufficient, then the image features in the training phase are not enough toeffectively represent the image features changes, so that the difficulty of image recognition is increased, and even the phenomenon that the image can not be recognized appears.

Description

technical field [0001] The invention relates to the field of image processing, in particular to an image processing method for expanding data sets under small samples. Background technique [0002] In recent years, feature learning methods based on CNN (Convolution Neural Network, Convolutional Neural Network) have achieved great success in image classification and object detection, and have attracted great attention in the field of computer vision. [0003] In practical applications, due to factors such as limited storage space and limited time to obtain samples, the number of training samples is often too small. Each image will have differences in clarity, illumination, contrast, etc. Small samples mean that the number of training samples is too small to cover different scenarios, so the recognition accuracy rate will inevitably decrease or cannot be recognized. [0004] If the training samples are sufficient, the trained image features will contain multiple variations of...

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
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/62
CPCG06F18/241G06F18/214
Inventor 刘宏基
Owner 成都网阔信息技术股份有限公司
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap