Pedestrian re-recognition method based on multi-task learning

Abstract

The invention provides a pedestrian re-recognition method based on multi-task learning. The method comprises the steps: obtaining a cross-camera pedestrian image, and constructing a pedestrian re-recognition training data set; constructing a multi-task learning network, wherein the network can combine an attribute task and an identity recognition task, all parameters and losses can be optimized ina combined mode, and therefore the purpose of improving pedestrian re-recognition accuracy is achieved. The attribute task and the recognition task respectively comprise a verification step, a verification step, a classification step and a verification loss to optimize the distance of the sample, the recognition loss is used for constructing a large class space and verifying the loss at the sametime, and the space is optimized by minimizing similar images and maximizing the distance between different images so as to improve the recognition precision.

Description

technical field [0001] The invention relates to a pedestrian re-identification method, in particular to a pedestrian re-identification method based on multi-task learning, which belongs to the technical field of deep learning. Background technique [0002] Pedestrian re-identification recognizes pedestrian images captured by multiple cameras without overlapping. It takes the target pedestrian captured by one camera as the retrieval object, accurately recognizes pedestrians from images captured by other cameras, and extracts robust representation features and adopts an effective measurement model. Recognition is composed of two processes. Most of the research work focuses on pedestrian appearance image representation features and similarity measurement model. The main task of pedestrian re-identification is to find the effective expression of pedestrian images. [0003] Before the emergence of deep learning technology, early person re-identification research mainly focused on...

AI Technical Summary

Problems solved by technology

[0006] 1) The representation features of different levels and different parts represent different semantics, and images are described through different views, different levels, and different partial visual color features, structural features or depth features, reflecting different semantics. These description data, which generate The methods are different, the information carried is different, the sensitivity to various environmental changes is different, and they are highly complementary. The research results show that using a single type to represent features and a single level to represent features, the recognition effect is not as good as multiple features and multi-level features The fusion effect is good. However, if the fusion of multiple features is used without distinction, its robustness is not good. It is necessary to study the feature fusion and utilization technology with strong environmental adaptability. The role of more robust features in this scene reduces the role of features with poor robustness in this scene. For this reason, it is necessary to study and analyze the representation features applicable to different changes

[0007] 2) The deep convolutional neural network shows excellent performance in image classification. Using the deep convolutional neural network to learn the deep features of pedestrian representation has good discriminative ability and good robustness to noise. However, network training requires A large amount of labeled data is time-consuming and labor-intensive. It is necessary to study how to increase the training data of the deep network and improve the robustness of deep features.

The reason for these sample differences is that cross-camera will bring complexity, and it is difficult for the model to learn and match these features

Images