A system and method for road intersection recognition based on deep learning
A deep learning and recognition method technology, applied in the field of image recognition, can solve the problem of unsatisfactory small target detection effect, and achieve the effect of solving inaccurate recognition
Active Publication Date: 2021-03-19
PLA STRATEGIC SUPPORT FORCE INFORMATION ENG UNIV PLA SSF IEU +1
View PDF10 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0013] However, since road intersections are generally small planar targets in remote sensing images, the candidate areas in the image in the existing methods are generated by the candidate area generation network, and the features of the candidate areas are only obtained by pooling the target area in the last convolutional layer. , the detection effect on small targets is not ideal
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
[0041] Since road intersections are generally small planar targets in remote sensing images, in the original FasterR-CNN method, the candidate area is generated by the candidate area generation network RPN, and the features of the candidate area are only passed through the target area by the last convolutional layer. Pooling is obtained, and the detection effect on small targets is not ideal. This embodiment proposes multi-scale detection for the detection of road intersections. The overall process of road intersections is as follows image 3 shown, including:
[0042] 1) Multi-scale feature map fusion
[0043]In a multi-layer convolutional neural network, the low-level features can well represent the details of image textures, edges, etc., and as the network layer deepens, the neuron receptive field becomes larger, and the high-level features can often be very good. To represent the semantic information of the image, the feature maps of each convolutional layer obtain diffe...
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 present invention provides a road intersection recognition system and method based on deep learning. The present invention improves the deep learning network and fuses the feature maps of the low-level and high-level convolutional layers from low to high, so that the low-level convolution The accumulation layer effectively fuses the detailed information such as texture and edge of the image with the semantic information of the image by the high-level convolution layer, and combines the processing ability of the RPN layer corresponding to each convolution layer, which is beneficial to the detection of small objects and solves the problem of The problem of inaccurate recognition of road intersections has been solved. The present invention obtains 16 different area suggestion frames by setting 4 sizes and 4 ratios, which can better cover all types of road intersections and enhance the recognition effect of X-shaped, Y-shaped, and composite intersections .
Description
technical field [0001] The invention relates to the field of image recognition, in particular to a road intersection recognition system and method based on deep learning. Background technique [0002] As the hub of road connections, road intersections provide important information such as accurate location, direction, and topological relationship for the rapid construction of road networks. In remote sensing images, road intersections are generally small planar objects with inconspicuous contour features and are easily disturbed by surrounding objects. The current road intersection detection algorithm needs to introduce more manual intervention, the degree of automation is low, and the detection effect of road intersections under complex backgrounds such as occlusion and adjacent objects with similar colors is not ideal. [0003] In recent years, the deep learning revolution has made remarkable achievements in the fields of computer vision and artificial intelligence, and h...
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 Patents(China)
IPC IPC(8): G06K9/00G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06V20/52G06N3/045G06F18/241
Inventor 金飞王龙飞芮杰刘智徐聪慧官恺王淑香孙启松吕虎
Owner PLA STRATEGIC SUPPORT FORCE INFORMATION ENG UNIV PLA SSF IEU