Natural-scene image recognition method based on deep convolutional network of VGG
A natural scene image and recognition method technology, applied in neural learning methods, character and pattern recognition, biological neural network models, etc., can solve the problems of low recognition rate of complex natural scene images, low recognition rate of difficult scenes, etc. The effect of improving the recognition accuracy and improving the recognition ability
Active Publication Date: 2018-08-14
北京理工雷科电子信息技术有限公司
View PDF8 Cites 17 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0004] In view of this, the present invention provides a natural scene image recognition method based on VGG deep convolution network, which can solve the problem of low recognition rate of complex natural scene images, especially the low recognition rate of difficult scenes
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
[0034] The present invention will be described in detail below with reference to the accompanying drawings and examples.
[0035] The present invention provides a kind of natural scene image recognition method based on VGG depth convolutional network, and the present invention method is by adding BatchNorm strategy respectively before the first two maximum pooling layers of VGG19 network, makes training network easier to fit; The network performs detection on the training samples and verification samples, reclassifies the samples with the category probability interval as the node, and conducts augmentation and retraining on the subdivided samples to achieve fine distinction between different natural scene images and improve the difficulty of distinguishing scenes and misclassification. The recognition rate of the scene, and then improve the recognition rate of the entire network for natural scene images. Specifically include the following steps:
[0036] Step 1. For n types o...
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 natural-scene image recognition method based on a deep convolutional network of VGG. According to the method of the invention, network training is enabled to be more liable tofitting through respectively adding a BatchNorm strategy before first two largest pooling layers of the VGG19 network; and detection is carried out on training samples and verification samples through the trained network, class probability intervals are used as nodes to finely reclassify the samples, augmented retraining is carried out on the finely reclassified samples, an effect of finely distinguishing different natural-scene images is achieved, a recognition rate of difficultly classified and falsely classified scenes is increased, and then a recognition rate of the entire network on thenatural-scene images is increased.
Description
technical field [0001] The invention belongs to the field of image target recognition, and in particular relates to a natural scene image recognition method based on a VGG deep convolution network. Background technique [0002] Natural scene target recognition has very important applications in image and video retrieval, tourism navigation, urban monitoring and planning, etc. However, unlike targets with relatively fixed geometric features, natural scene image targets are more complex, often consisting of multiple types of small targets. The complexity and variability of different scenes undoubtedly increase the difficulty of classification and recognition, especially for similar different types of scenarios. [0003] At present, the methods for natural scene image classification are mainly methods based on low-level features combined with traditional machine learning classifiers and machine learning methods based on convolutional neural networks. The former generally need...
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 Applications(China)
IPC IPC(8): G06K9/00G06N3/04G06N3/08
CPCG06N3/08G06V20/00G06N3/047G06N3/045
Inventor 曾大治董安冉赵艳霞刘英杰
Owner 北京理工雷科电子信息技术有限公司