Massive case data efficient retrieval method based on differentiated storage media

Abstract

The invention relates to a massive case data efficient retrieval method based on a differentiated storage medium, and the method comprises the steps: storing pedestrian or cyclist structural semanticinformation in a relational database table, and storing face feature data in a non-relational database table; performing target detection and face detection on the pedestrian or the rider of the to-be-retrieved picture, and extracting face feature data; selecting the pedestrian or rider type and one or more fields of the structured semantic information, querying a corresponding target unique identifier in a relational database table, and efficiently and quickly retrieving face feature data meeting conditions in a face feature table in a non-relational database in combination with the target type of the pedestrian or cyclist. According to the method, massive cases are quickly and accurately retrieved in a retrieval mode of combining the face feature data and the human body structured semantics.

Description

technical field [0001] The invention relates to the storage and retrieval of massive data, in particular to an efficient retrieval method for massive case data based on differentiated storage media. Background technique [0002] Video tracking has become one of the important means to solve cases. In the process of video tracking, it is necessary to retrieve the target object from the video data. [0003] At present, for the storage of video data, different storage media are used according to the type of data structure. At present, the storage of face feature vector data (unstructured data) is still based on files and relational databases as storage media. The space is large, and the performance when reading the face feature vector data is low. [0004] However, in the traditional case-by-case retrieval method, there are too many case results of the same type in the case database retrieved (through the search of structured semantic fields in the case), and the retrieved effe...

AI Technical Summary

Problems solved by technology

[0003] At present, for the storage of video data, different storage media are used according to the type of data structure. At present, the storage of face feature vector data (unstructured data) is still based on files and relational databases as storage media. The space is large, and the performance of reading the face feature vector data is low

[0004] However, in the traditional case-by-case retrieval method, there are too many case results of the same type in the case database retrieved (through the search of structured semantic fields in the case), and the retrieved effective case results are not necessarily accurate, especially when the case database When storing a large number of cases with a long history in the medium, the case handlers often spend a long time and a lot of energy to search. Even if the results are re-screened by the attributes again, it may not be possible to achieve more accurate retrieval. Effective cases, in the case of massive data, will consume a lot of energy, and even delay the time to handle the case, which is very unintelligent

Images