Pedestrian re-identification method, device, electronic device and computer-readable storage medium

An apparatus for pedestrian re-identification, comprising:

• an acquisition unit, configured to acquire a pedestrian image to be detected;
• a first extraction unit, configured to extract global feature information of the pedestrian image to be detected through multiple convolutional layers of a convolutional neural network;
• a second extraction unit, comprising an intermediate feature extraction unit and a local alignment unit, wherein the intermediate feature extraction unit is configured to extract multiple pieces of intermediate feature information of the pedestrian image to be detected through the multiple convolutional layers of the convolutional neural network respectively, and the local alignment unit is configured to perform local alignment on the multiple pieces of intermediate feature information and merge the multiple pieces of locally aligned intermediate feature information as the local feature information, wherein the multiple pieces of intermediate feature information each corresponds to one of the multiple convolutional layers;
• and
• a determination unit, configured to assign the global feature information and the local feature information as a classification feature of the pedestrian image to be detected and determine a classification result of the pedestrian image to be detected according to the classification feature.

An electronic device, comprising a processor and a memory, wherein the memory is configured to store an application program code, and the processor is configured to perform the application program code to implement a method for pedestrian re-identification, the method for pedestrian re-identification, comprising:

• acquiring a pedestrian image to be detected;
• extracting global feature information of the pedestrian image to be detected through multiple convolutional layers of a convolutional neural network;
• extracting multiple pieces of intermediate feature information of the pedestrian image to be detected through the multiple convolutional layers of the convolutional neural network respectively, and performing local alignment on the multiple pieces of intermediate feature information, and merging multiple pieces of locally aligned intermediate feature information as the local feature information, wherein the multiple pieces of intermediate feature information each corresponds to one of the multiple convolutional layers;
• and
• assigning the global feature information and the local feature information as a classification feature of the pedestrian image to be detected, and determining a classification result of the pedestrian image to be detected according to the classification feature.