Island coastal wetland waterfowl identification method, distribution query system and medium

Abstract

The invention belongs to the technical field of marine ecological environment monitoring, and discloses an island coastal wetland waterfowl identification method, a distribution query system and a medium. A novel regional convolutional network model and a full convolutional network model are used for automatically detecting and positioning the whole target waterfowl and key parts thereof in the image; and carrying out probability modeling on the high-level semantic features by using a deep convolutional probabilistic neural network model, enhancing the expression of the image content, and identifying the target waterfowl based on the expressed image. The invention provides an island / coastal wetland important waterfowl variety investigation technology based on artificial intelligence, provides efficient video target tracking, machine learning and target image intelligent detection and identification technologies, and provides a waterfowl image online intelligent detection and identification system based on a computer vision technology. The types, the number, the habitat conditions, the living states and the like of important waterfowls in the target area are monitored and recognized.

Description

technical field [0001] The invention belongs to the technical field of marine ecological environment monitoring, and in particular relates to a water bird identification method, a distributed query system and a medium in an island coastal wetland. Background technique [0002] At present, with the rapid development of computer vision and machine learning technology, image-based bird recognition has attracted more and more attention and attention. The current mainstream methods of bird recognition research are mainly based on a single image. First of all, in terms of basic bird database construction, the most influential in the world is the CUB-200-2011 bird database jointly constructed by the Vision Laboratory of Caltech and Cornell University, which includes a total of 200 species of birds And 11,788 images, each image is marked with the bird's overall position, size, 14 main body parts and 28 sets of attributes. In addition, the Cornell University Laboratory of Ornitholo...

AI Technical Summary

Problems solved by technology

It is impossible to monitor and identify important waterfowl types, numbers, habitat conditions and living conditions in the target area

[0009] (2) In the existing technology for the recognition of dynamic micro-targets such as water birds, the accuracy of image recognition is low; the image processing time is long, which cannot meet the actual needs

Images