Target detection algorithm based on twin neural network

Abstract

The invention discloses a target detection algorithm based on a twin neural network. The target detection algorithm comprises the following steps: collecting images continuously shot at a fixed visualangle; calculating the similarity between the to-be-detected image and the reference image by using a twin neural network; and quickly finding out changed targets in the scene by utilizing a similarity analysis result between the to-be-detected image and the reference image, and classifying the targets. Aiming at the characteristics of the continuously acquired images, the information relevance among the plurality of images is considered, the information among the plurality of continuous images is fully explored, and the detection speed is improved while the detection precision is improved. The method is suitable for target detection of images which are continuously shot at a fixed angle.

Description

technical field [0001] The invention relates to an image target detection method, in particular to a twin neural network-based target detection algorithm suitable for fixed-angle, continuously captured images. Background technique [0002] Target detection refers to identifying a target object in an image, and then marking the position of the object in the image. Currently, a rectangular frame is used to mark the target object. [0003] The accuracy and speed of traditional target detection algorithms in the current application cannot keep up with the target detection algorithms based on deep learning. Therefore, the current mainstream target detection algorithms are basically based on deep learning. Most of the algorithms are A convolutional neural network is used. These algorithms use convolutional neural networks to extract the features of objects in the image, but in this process, there is a big disadvantage that after performing multiple convolution and pooling and oth...

AI Technical Summary

Problems solved by technology

[0006] In the actual application scene of continuous shooting, the current target detection algorithm is insufficient for the information of the captured image, and the recognition accuracy is not enough. The purpose of the present invention is to provide a target detection algorithm based on the twin neural network. The characteristics of the image, considering the information correlation between multiple images, fully explore the information between multiple consecutive pictures, and improve the detection accuracy while also improving the detection speed

Images