Multi-scale sensing pedestrian detection method based on improved full convolutional network
A fully convolutional network, pedestrian detection technology, applied in the field of multi-scale perception pedestrian detection based on improved fully convolutional network, can solve the problem of pedestrian accuracy not meeting the requirements of the industry.
Active Publication Date: 2018-11-16
JIANGNAN UNIV
View PDF2 Cites 33 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0004] Due to the complexity of the environment, the accuracy of the current algorithm for detecting pedestrians in rainy or dark conditions cannot meet the requirements of the industry
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
[0046] In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention, not to limit the present invention.
[0047] refer to Figure 6 , a multi-scale perceptual pedestrian detection method based on an improved fully convolutional network, including:
[0048] Normalize the size of the input image to a predetermined pixel, input it to the RoIDataLayer of the ResNet-50 network, and learn the characteristics of pedestrians;
[0049] The first four layers of the ResNet-50 network are used to extract the pedestrian area in the image and generate feature maps of different scales;
[0050] In the res5a_branch2b layer, res5b_branch2b layer and res5c_branch2b layer of ResNet-50, the deformable convolution layer ...
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 relates to a multi-scale sensing pedestrian detection method based on an improved full convolutional network, and belongs to the field of pedestrian detection. The method comprises the following steps: firstly, importing a deformable convolutional layer in the full convolutional network structure to expand the sensing the field of a feature map; secondly, extracting a multi-scale pedestrian suggestion region through a cascaded RPN, importing a multi-scale discriminant strategy, defining a scale discriminant layer, and judging a scale category of the pedestrian suggestion region;and finally constructing a multi-scale sensing network, importing a Soft-NMS detection algorithm, and combining a classification value and a regression value of each network output to obtain a final pedestrian detection results. Experiments show that the detection algorithm of the invention produces lower detection errors on benchmark pedestrian detection data sets Caltech and ETH, which is superior to the accuracy of all detection algorithms in the current data set, and is suitable for detecting pedestrians on a far scale.
Description
technical field [0001] The invention relates to the technical field of pedestrian detection, in particular to a multi-scale perceptual pedestrian detection method based on an improved full convolutional network. Background technique [0002] In recent years, with the wide application of intelligent video surveillance, vehicle assisted driving (ADAS), content-based image or video exploration and human behavior analysis, and the emergence of some new application fields, such as home service robots, aerial photography-based The research on pedestrian detection technology has become an important research topic in the field of machine vision. Pedestrian detection technology has great challenges and room for improvement, mainly because pedestrian targets have greater pose changes than face targets, and the size span of different pedestrians in videos or pictures is large. Due to the different backgrounds of pedestrians, different light intensities, and the diversity of clothing, ...
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): G06K9/00G06N3/04G06N3/08G06V10/764
CPCG06N3/08G06V40/20G06V40/10G06N3/045G06V10/454G06V10/82G06V10/764G06N3/048G06N3/04G06F9/545G06F18/2148G06F18/213
Inventor 彭力刘辉闻继伟
Owner JIANGNAN UNIV