Unlock instant, AI-driven research and patent intelligence for your innovation.

A convolutional neural network target detection method based on an RGB-D camera

A convolutional neural network and target detection technology, which is applied in biological neural network models, neural architectures, image data processing, etc., can solve problems such as limited improvement and achieve the effect of improving detection accuracy

Active Publication Date: 2019-06-18
HANGZHOU DIANZI UNIV
View PDF9 Cites 13 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] However, the convolutional neural network algorithm that only relies on color images has limited improvement in accuracy, and its growth has flattened in recent years.

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
  • A convolutional neural network target detection method based on an RGB-D camera
  • A convolutional neural network target detection method based on an RGB-D camera
  • A convolutional neural network target detection method based on an RGB-D camera

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0036] The following combination figure 1 To further illustrate the present invention, the present invention includes the following steps:

[0037] Step (1): Use RGB-D camera to obtain color image and depth image

[0038] The RGB-D camera is used to shoot the scene containing the target object, and a color image and a depth image corresponding to the color image pixels are obtained.

[0039] Step (2): Use convolutional neural network to predict the position of the target object

[0040] (a) Collect the data set containing the target object first, and manually calibrate the target frame so that the target frame can just contain the target object. Calculate the aspect ratio of the target frame in the data set, and use k-means clustering to generate k aspect ratio values. Then, k anchor frames with an area of ​​1 are generated, and the aspect ratios of the anchor frames correspond to the k values ​​generated by the clustering, and k anchor frames with different shapes are obtained.

[00...

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 relates to a convolutional neural network target detection method based on an RGB-D camera. At present, more and more target detection algorithms utilize a convolutional neural network to position a target object. However, most of the convolutional neural network frameworks only use the color camera to predict the position of the target object. However, using only RGB information, itis very difficult to make the convolutional neural network achieve high detection accuracy. It is difficult to achieve comprehensive consideration of convolutional neural network modeling, training programs and other factors. The invention utilizes a depth map acquired by an RGB-D camera, and the auxiliary convolutional neural network predicts the position of the target object. By using the distance information in the depth image, the size of the target object can be estimated in advance, the modeling difficulty of the convolutional neural network is reduced, and the detection accuracy of thenetwork is improved.

Description

Technical field [0001] The invention belongs to the field of computer vision, and specifically relates to a convolutional neural network target detection method based on an RGB-D camera. Background technique [0002] Traditional target detection algorithms use artificially designed feature extractors to extract image features, and then use machine learning algorithms such as SVM to classify features in specific areas to obtain detection results. However, due to the limitations of the artificially designed feature extractor, it can only extract part of the features of the object, and it is difficult to fully extract the features of an object. It is also difficult for the machine learning algorithm to learn all the features of the object, resulting in the generalization ability of the entire algorithm. Relatively poor, the recognition accuracy is very low, and it is difficult to meet the expected requirements. [0003] Since 2012, deep learning has been widely used in the computer f...

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
IPC IPC(8): G06T7/70G06T7/50G06N3/04
Inventor 杨宇翔杜宇杰高明煜张敬
Owner HANGZHOU DIANZI UNIV