Real-time scene layout identification and reconstruction method based on an artificial neural network
An artificial neural network and scene technology, applied in the field of real-time scene layout recognition and reconstruction based on artificial neural network, can solve problems such as complex real-time layout recognition and plane 3D reconstruction, and achieve the effect of improving processing capacity and reducing hardware costs
Active Publication Date: 2018-10-12
HANGZHOU DIANZI UNIV
View PDF12 Cites 11 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0005] The purpose of the present invention is to solve the problems of real-time layout recognition and plane three-dimensional reconstruction in complex indoor scenes. Correlation, improve the accuracy of network judgment, after obtaining the layout recognition of indoor scenes (ground and wall, ceiling and wall, wall and wall intersection line), under the assumption of the Manhattan world, use the space One point has the principle of unique projection in the pixel coordinate system, iterative optimization, get the unit normal vector of each plane in the scene and the distance from the plane to the camera center, and render the plane under OpenGL
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
[0032] In order to make the above objects, features and advantages of the present invention more comprehensible, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.
[0033] The real-time scene layout recognition and reconstruction method based on the artificial neural network proposed by the present invention, such as figure 1 shown, including the following three steps:
[0034] 1) To build an artificial neural network model, the convolutional residual neural network is selected as the basic structure of the network in the method. In order to enable the network to output a matrix, several convolutional layers are used at the top of the network to replace the general fully connected layer. In order to overcome the imbalance of training data and better deal with severe occlusions in the scene, network training is divided into three stages:
[0035] I. Phase 1: Training the network to achieve se...
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 discloses a real-time scene layout identification and reconstruction method based on an artificial neural network. The invention uses an artificial neural network model to process the input sequence, and by looking for correlations between individual frames, based on the Manhattan world hypothesis, the unit normal vector of each plane in the scene and the distance from the plane to the center of the camera are obtained by iterative optimization using the principle that there is a unique projection of a point in the space in the pixel coordinate system, and the plane is rendered in OpenGL, after the layout identification of the indoor scene. The invention utilizes the artificial neural network and searches the connection between frames, so that the network output result and the three-dimensional reconstruction result are more accurate. Single-purpose RGB information is used to restore three-dimensional information to reduce the hardware cost. At the same time, the normal vector and distance to the camera center of each plane in the scene are obtained by using the layout information of each frame and optimization algorithm.
Description
technical field [0001] The invention belongs to the technical fields of computer vision and computer graphics, in particular, the invention relates to a real-time scene layout recognition and reconstruction method based on an artificial neural network. Background technique [0002] The ultimate goal of computer vision is to achieve the ability of the human eye and the human brain to understand images, share the processing and analysis of image information for humans in the era of information explosion, and enable more intelligent devices to better complete more tasks. Task. [0003] Indoor scene layout recognition and 3D reconstruction is an important and basic problem in the field of computer vision and computer graphics, which can provide strong prior conditions for other indoor scene tasks, mainly including pedestrian detection, object tracking, human Face recognition, etc., image content retrieval in the field of Internet information, target finding, scene understanding...
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): G06T7/55G06T15/00G06K9/00
CPCG06T7/55G06T15/005G06T2207/10016G06T2207/10024G06T2210/04G06V20/36
Inventor 颜成钢邵碧尧徐枫丁贵广张勇东
Owner HANGZHOU DIANZI UNIV