Pedestrian detection method based on deep learning technology

Abstract

The invention discloses a pedestrian detection method based on a deep learning technology. The method comprises the steps that firstly, a binary classification model is trained through a step-by-step migration strategy on the basis of transfer learning to initialize final model parameters; secondly, pedestrian detection work is completed by adopting and modifying a currently popular and efficient Faster RCNN frame, and on the basis of the CNN characteristics of the frame, not only can images with any scales be processed, but also the detection speed is high. Compared with the prior art, the method has the advantages that the network does not need to be specially designed, existing available data is fully utilized, a good experiment result still can be achieved by adopting a general network structure, the advantages of a deep convolution network are fully achieved, and the advantages of being simple in design, good in robustness, high in detection accuracy and low in omission ratio are achieved.

Description

technical field [0001] The invention relates to a pedestrian detection method based on deep learning technology, which belongs to the technical field of image processing and computer vision. Background technique [0002] With the advancement and development of science and technology, our way of life is also slowly changing. Many tasks that used to require a lot of manpower can now be completed by computers. In recent years, due to the rapid development of Internet technology, communication technology, and Internet of Things technology, a large amount of video image information has been generated, and these massive amounts of information contain huge information volume and commercial value. [0003] An important source of human perception of the world is through visual information. Studies have shown that about 80% to 90% of the information that humans obtain from the outside world comes from the visual information obtained by human eyes. Human beings have a high ability to ...

AI Technical Summary

Problems solved by technology

This method also has obvious deficiencies: (1) It is necessary to pre-normalize the image to a specific size, so that the aspect ratio and scale of the input image are ignored. When the obtained model is used for an image of any size, it is often necessary to crop the original image or do some geometric transformation, which will affect the accuracy and robustness of the model due to the loss of a large amount of useful information or the introduction of geometric distortion. ; (2) The neural network structure adopted by this method is very simple, and the information contained in the data cannot be well mined

However, this method also has some shortcomings: (1) the original image needs to be pre-cropped, which is not only labor-intensive, but also some useful information cannot be used; (2) the original image needs to be specially pre-processed, and the process is relatively Complicated; (3) The data is relatively limited and single, and only one database (such as Caltech, ETH) is used, and the available data is not comprehensively utilized; (4) Although the network has been specially designed (such as components, occlusion design), but Generally speaking, it is not deep enough, it is difficult to mine the deep information of the data, and it does not give full play to the characteristics of CNN features such as strong robustness and wide applicability

Images