Vehicle re-identification method based on perception cascade context

Abstract

The invention discloses a vehicle re-identification method based on a perception cascade context, and the method comprises the following steps: constructing a feature extraction model of a vehicle image, wherein a basic network uses a pre-trained ResNet-50 classification network; embedding an attention enhancement module based on the perception cascade context into the feature extraction model; inputting a vehicle image data set, and training the constructed feature extraction model; and carrying out a vehicle re-identification task by adopting the trained feature extraction model. According to the invention, the attention enhancement module based on the perception cascade context is embedded in the feature extraction model of the vehicle image, so that richer and more distinctive featureinformation in the vehicle image can be extracted, and the accuracy of a re-identification task is improved; and only one piece of label information of the vehicle ID is used as a supervision signal to carry out model training without depending on any fine-grained label information (such as a license plate, a vehicle model, a color and the like).

Description

technical field [0001] The invention relates to the technical field of intelligent traffic monitoring, in particular to a vehicle re-identification method based on perception cascading context. Background technique [0002] Vehicle re-identification currently belongs to the category of target re-identification in the field of scientific research, and is a subfield task under computer vision. Before the rise of deep learning technology, traditional vehicle re-identification can only be identified by continuously capturing vehicle images and using the license plate as the unique ID of the vehicle. Unfortunately, this technique relies heavily on the recognition accuracy of license plates. In many illegal and criminal incidents, license plates are often blocked, removed or even forged. In this case, the public security personnel can only use human eyes to check the time, place and vehicle characteristics of the vehicle in the video for investigation. [0003] The current main...

AI Technical Summary

Problems solved by technology

However, this method has inherent deficiencies: First, to obtain finer-grained label information may require a higher-definition shooting device, and a larger image size will lead to a decrease in the efficiency of the entire model reasoning; second, it needs to spend more A lot of manpower and material resources to label the fine-grained areas

Images