Target re-identification method based on hypersphere embedding in densely connected convolution networks

Abstract

The invention provides a target re-identification method based on hypersphere embedding of densely connected convolution network, at first, DenseNet is used to extract the features of underwater deformation object in the video sequence according to the secret-level connected convolution network, so that the gradient disappear is greatly reduced, feature propagation is enhanced, feature reuse and parameter learning processes are supported , from the viewpoint of fine-grained classification, from local integration to global integration, the characteristics of underwater deformation targets are extracted by means of grouping average pooling, the more accurate feature expression ability of underwater deformation target is obtained, in order to avoid directly measuring the Euclidean distance between the coding features of underwater deformed individual targets, a complete and continuous underwater deformed individual target re-recognition model based on multi-point placement is constructedby using hypersphere loss, i.e. Angular triple loss, to focus on the inter-class difference and intra-class difference of underwater deformed individual targets and avoid directly measuring the Euclidean distance between coding features. The invention finally completes the close supervision and process tracking of the underwater deformation target individual in the close-range multi-field observation.

Description

technical field [0001] The invention relates to an underwater deformation and moving target re-identification method based on densely connected convolutional network hypersphere embedding, and belongs to the technical fields of intelligent information processing and target detection and recognition. Background technique [0002] my country has a vast sea area and rich fishery resources, so marine monitoring is very necessary. Not only can it detect abnormal underwater targets in time and take countermeasures, but it can also improve the marine environment in time to avoid pollution to the marine environment and damage to the ecological environment. Therefore, it is of great significance to carry out long-term and effective multi-camera and multi-angle monitoring of the main resources in the marine environment, and also provides a basis for aquaculture fisheries and marine fishing industries. The behavioral analysis of underwater deformable targets provides data and informati...

AI Technical Summary

Problems solved by technology

That is, the main problems existing in the prior art: (1) Due to the complex optical imaging conditions of the ocean, factors such as post-scattering of underwater imaging, turbid water, sediment, ocean currents, phytoplankton, and swinging aquatic plants will cause optical image quality to decline or The introduction of interference will cause errors in target re-identification; (2) The rapid swimming and large-scale shape changes of underwater deformable targets will also cause difficulties in the monitoring of underwater deformable targets

Images