Deep learning pedestrian detection method based on embedded terminal

Abstract

The invention discloses a deep learning pedestrian detection method based on an embedded terminal. The method comprises the steps that first, at a sample preparation stage, an existing automatic driving dataset is acquired, or manual annotations obtained after a fixed camera and a mobile camera shoot videos are collected; second, at a training stage, a large quantity of training images are used totrain parameters of a constructed convolutional neural network so as to complete detection feature learning; third, at a test stage, a large quantity of test images are input into the trained convolutional neural network, and a detection result is obtained; and fourth, at a porting stage, code level optimization and porting into the embedded terminal are performed. According to the method, the 18-layer convolutional neural network is adopted to perform pedestrian feature learning, and the method has an innovative advantage compared with a traditional machine learning method; and an optimization strategy for the embedded terminal is also proposed, the network scale and algorithm complexity are further reduced, and the method is suitable for ADAS function application.

Description

technical field [0001] The invention relates to a pedestrian detection method, in particular to a deep learning pedestrian detection method based on an embedded terminal. Background technique [0002] In recent years, with the rapid development of deep learning, opportunities have emerged for the development of advanced driver assistance. For advanced assisted driving, the most important thing is object detection, including pedestrian detection. At the same time, pedestrian detection is also of great significance to the fields of intelligent robots and video surveillance. In order to achieve staged progress in advanced assisted driving, pedestrian detection is an unavoidable problem to be overcome. At present, the traditional pedestrian detection method is mainly to manually design features. By extracting features to train classifiers, the effect is impressive. However, it is difficult for artificially designed feature classifiers to adapt to large changes in scenes and can...

AI Technical Summary

Problems solved by technology

Although this method can detect the information of pedestrians, it has a large error in the location information of pedestrians and cannot be applied to the field of advanced driver assistance systems. Advanced driver assistance systems have very strict requirements on location information.

Moreover, this method does not perform well in detecting objects that are very close to each other or in a small group.

[0005] At the same time, the currently disclosed technical solutions have a huge amount of calculation, and the convolutional neural network algorithm is huge and only applicable to servers or computer clusters, and cannot be used for end-to-end real-time applications of embedded devices.

Images