Tree species image recognition method based on natural background

An image recognition and background technology, applied in character and pattern recognition, instruments, computer parts, etc., can solve the problems of different sizes, inability to ensure the accuracy of image segmentation, and single posture

Inactive Publication Date: 2019-10-18
ZHEJIANG FORESTRY UNIVERSITY
View PDF15 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

It solves the problems of difficult sampling, blurred marks, and different sizes when establishing a crop pest database in the prior art, and also solves the over-fitting problem caused by the small number of samples and single posture during the training process of the deep convolutional neural network model , but this method cannot ensure the accuracy of image segmentation, and this method only segments the image and then builds an image database, and does not disclose how to use these images to train a deep convolutional neural network

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
  • Tree species image recognition method based on natural background
  • Tree species image recognition method based on natural background
  • Tree species image recognition method based on natural background

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0069] A method for tree species image recognition based on a natural background, the flow of the method is shown in the figure, and specifically includes the following steps:

[0070] S1 uses web crawler technology to capture image data of trees, stores them according to different types, and forms a set of data sets;

[0071] S2 uses the GrabCut method to segment images with complex natural backgrounds, and obtains tree images without natural backgrounds;

[0072] S3 uses the convolutional neural network to learn the segmented tree image dataset;

[0073] S4 Use the K-fold cross-validation method to select the model with the smallest average test error in K evaluations, and use this model to identify tree species.

Embodiment 2

[0075] A kind of tree species image recognition method based on the natural background of the present embodiment, based on the first embodiment, the GrabCut method in the step S2 is specifically: assuming that Z represents the collection of pixel points, Z=(z 1 ,z 2 ,…,z n ,…z N ,),z n Represents each pixel in the image, α=(α 1 ,…,α n ,…,α N ,), α n ∈{0,1}, corresponding to each pixel of the image Z=(z 1 ,z 2 ,…,z n ,…z N ,), where a n Represents the mark value of each pixel, the mark value is 0 or 1, 0 represents the background, 1 represents the foreground, the GrabCut method can be described as the value of α when minimizing the E(α, θ) energy function, expressed as: α=

[0076] Using a Gaussian mixture model containing K Gaussian components to describe the image, the energy function E can be expressed as:

[0077] E(α,k,θ,z)=U(α,k,θ,z)+V(α,z)

[0078] In the formula, α represents the set of label values ​​corresponding to each pixel in the image, z represent...

Embodiment 3

[0087] A kind of tree species image recognition method based on the natural background of the present embodiment, based on the second embodiment, in order to minimize the energy function E, the specific iterative process includes the following steps:

[0088] S21 The user obtains an initialized tripartite map by directly selecting the position of the tree in the target image of interest, and the outside of the rectangular frame is set as the background area T B , the area T inside the rectangular box is set as "may be the target" U , foreground T F set to empty, i.e.

[0089] S22 sets the background area T B The mark of the pixel in is initialized to 0, that is, another T B α=0 in the region, the region T U The mark of the pixel in is initialized to 1, that is, T U α = 1 in the region;

[0090] S23 initializes the Gaussian mixture model of the target area and the background area by using the marker α;

[0091] S24 to T U Each pixel in Z n Assign Gaussian components ...

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 a tree species image recognition method based on a natural background. The method comprises the steps of firstly capturing tree image data through employing a web crawler technology, storing the tree image data according to different types, and forming a data set; preprocessing the image, and segmenting a complex background in the original image from a target tree entity byadopting a Grab-Cut image segmentation method; learning the segmented image by using a convolutional neural network; adopting k-fold cross validation for model selection, wherein firstly, known dataare randomly segmented into K subsets which are not intersected with one another and are the same in size; training a model by using the data of K-1 subsets, and testing the model by using the remaining subsets; repeating the process for K times, finally selecting the model with the minimum average test error in K times of evaluation, and identifying the tree species image. According to the method, an image segmentation method is adopted, complex backgrounds which are not used for tree species identification are segmented, and the tree species identification rate is greatly improved.

Description

technical field [0001] The invention relates to a tree species image recognition method, in particular to a tree species image recognition method based on natural background. Background technique [0002] There are many types of trees in nature, and they are still growing. Facing the increasing number of tree species, it is difficult for ordinary people to know each tree. Therefore, researchers have been devoting themselves to proposing an effective and automatic method for identifying tree species. In recent years, with the development of deep learning, image recognition algorithms based on convolutional neural networks have been widely used in tree species recognition due to their good performance. [0003] Tree species recognition algorithms based on convolutional neural networks usually learn features autonomously according to the unique convolutional layers and pooling layers in their networks. In the existing technology, Dou Gang et al. provided a method for tree sp...

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): G06K9/00G06K9/34G06K9/62
CPCG06V20/10G06V10/267G06F18/241
Inventor 冯海林胡明越方益明周国模
Owner ZHEJIANG FORESTRY UNIVERSITY
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