Multi-scale target identification method

Abstract

The invention relates to a multi-scale target identification method. The method comprises the following steps: S1, constructing a neural network model; S2, training a neural network model based on the training set to obtain a trained neural network model; S3, performing calculation based on the test set and the loss function to obtain a loss value of the trained neural network model; S4, judging a loss value, if the loss value is smaller than a loss set value, executing S5, and if the loss value is larger than or equal to the loss set value, adjusting hyper-parameters of the neural network model based on a gradient descent algorithm and a verification set, and retraining and repeatedly executing S2-S4 until the loss value of the neural network model is smaller than the loss set value; and S5, realizing target detection and positioning based on the trained neural network model. The target can be accurately identified, the real-time performance is high, and the category and the position of the target can be rapidly returned when the water emergency situation is processed.

Description

technical field [0001] The invention relates to the technical field of target detection, in particular to a multi-scale target recognition method. Background technique [0002] Surface target recognition technology is the key technology for unmanned boats to perceive the environment. Whether it is maritime environmental reconnaissance, coastline patrols, early warning of emergencies, rescue of surface personnel, and obstacle detection, accurate and real-time surface target recognition plays an important role. play an irreplaceable role. In recent years, deep convolutional neural networks have made great progress in the fields of target detection and recognition, and have been successfully applied to medical diagnosis, face recognition and other fields, and have achieved higher and higher accuracy rates in general feature sets. However, due to the complex and changeable water surface scene, the large scale span of water surface targets, and the low image resolution, water su...

AI Technical Summary

Problems solved by technology

Unmanned boats are usually equipped with processors that can calculate data, but due to energy consumption and space constraints, using a large-scale network model will cause delays and freezes, which also poses challenges to the portability and real-time performance of the algorithm. higher requirement

Images