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

Robust and efficient unmanned pedestrian detection method

A technology for unmanned driving and pedestrian detection, applied in the field of unmanned driving environment perception, which can solve the problems of slow speed and inability to achieve real-time performance.

Pending Publication Date: 2020-01-10
TIANJIN UNIV
View PDF0 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Researchers have also begun to apply convolutional neural networks to pedestrian detection, but these methods are very slow and cannot meet the real-time requirements

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
  • Robust and efficient unmanned pedestrian detection method
  • Robust and efficient unmanned pedestrian detection method
  • Robust and efficient unmanned pedestrian detection method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0032] Referring to the accompanying drawings, the present invention presents a robust and efficient unmanned driving pedestrian detection method, which introduces the idea of ​​regression into pedestrian detection based on deep learning and proposes a new one-stage pedestrian detection method based on deep learning, specifically The implementation steps are as follows:

[0033] Step 1: Design feature extraction network;

[0034] Since in unmanned driving scenarios, due to the difference in the size of pedestrians caused by distance and the influence of light, choosing a high-quality basic network is crucial for feature extraction in pedestrian detection. Therefore, we choose easier optimization, performance Better deep residual network as base network. At the same time, the pedestrian detector in the unmanned driving scene should meet the real-time requirements, so the 34-layer deep residual network is selected as the feature extraction network.

[0035] Step 2: Specify a p...

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 robust and efficient unmanned pedestrian detection method. The method comprises the following steps: (1) selecting a deep residual network as a feature extraction network; (2) assigning a positive and negative class label to each candidate box; (3), defining a target loss function; (4) setting the size of a candidate box; (5) adopting a difficult-to-mine sample mining method to enable the ratio of the number of the final positive samples to the number of the final negative samples to be 1: 3; (6), adopting data augmentation; and (7), training a network. According to the method, the single convolutional neural network is used, the coordinates of the bounding box and the probability of pedestrians are directly obtained from the original image, end-to-end training can be achieved, the residual unit used by the deep residual network can improve the optimization process of the deep network model, and the time overhead of model convergence is reduced.

Description

technical field [0001] The invention belongs to the field of unmanned driving environment perception, and in particular relates to a robust and efficient unmanned driving pedestrian detection method. Background technique [0002] Pedestrian detection unit is an irreplaceable part in autonomous driving environment perception. In unmanned driving scenarios, pedestrian detectors not only have to deal with appearance differences caused by posture, clothing, occlusion, and scale in general pedestrian detection, but also must consider issues unique to special application scenarios, such as light changes and The impact of speed. Therefore, the pedestrian detector should be robust and real-time, so that the self-driving car can avoid pedestrians in time. Pedestrian detection is an important and challenging research direction in computer vision, and its wide application has become a research hotspot for companies and related researchers [0003] Traditional pedestrian detection alg...

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): G06K9/00G06K9/62
CPCG06V40/103G06F18/214G06F18/24
Inventor 王一晶郑开辅左志强
Owner TIANJIN UNIV