Wild animal appearing and disappearing detection method based on recursive multi-feature fusion

Abstract

The invention discloses a wild animal appearing and disappearing detection method based on recursive multi-feature fusion, and the method comprises the steps of collecting a video through a camera disposed in a natural reserve, carrying out frame extraction processing on the video, obtaining a large number of pictures, taking the pictures as a sample data set, and marking the class of each animal of the sample pictures in the sample data set; establishing a Cascade R-CNN target detection model fused with a recursive Re-BiFPN structure; training the established target detection model by using the sample data set to obtain a wild animal appearing and disappearing detection model; and inputting the video pictures acquired by the camera in the natural reserve into the wild animal appearing and disappearing detection model for wild animal detection. According to the method, the feature extraction capability of a network can be effectively improved, the accuracy of wild animal target detection is improved, and the wild animals can be effectively identified under the bad video collection conditions, such as animal shielding, poor light conditions, etc.

Description

technical field [0001] The invention belongs to the technical field of wild animal detection, in particular to a wild animal detection method based on recursive multi-feature fusion. Background technique [0002] Wild animals occupy an important position in the species composition of the ecosystem, and the protection of wild animal resources in nature reserves is an important part of protecting the ecological balance. For wild animals, protecting their individuals can effectively control the number of wild animals, demonstrate the effectiveness of wild animal protection, and warn the status quo of wild animal protection. Therefore, the collection and analysis of wildlife information is an important means to grasp the living conditions of wild animals. The current detection and protection of wild animals is through the deployment of a large number of camera equipment in nature reserves to conduct fixed-point video recording and save the video. However, due to the characteri...

AI Technical Summary

Problems solved by technology

[0004] However, in the actual application scenarios of nature reserves, there are situations where wild animals are blocked, and the color of animal hair is similar to the background color of trees and other background colors. In addition, some animals like to come and go at night, which is not conducive to Cascade R-CNN The network extracts the characteristics of animals to affect the accuracy of model detection

Images