Target detection method based on micro-aircraft

Abstract

The invention provides a target detection method based on a micro-aircraft. According to the method, the size of the model is reduced by using depth separable convolution; the generalization of the model is improved by using a multi-branch structure, an extra convolution layer is further added to a front-end network MBDSCNN structure of the target detection model, and the size of the extra convolution layer is set to be decreased layer by layer, so that the model can be predicted under a multi-scale condition. The established target detection model can combine the output of each extra convolution layer to replace the full connection layer of the MBDSCNN, and a non-maximum suppression layer (NMS) is added behind the full connection layer to output a detection result. Therefore, compared with the prior art, the method has the advantages that targets under various scales can be accurately recognized, and the detection result can be accurately output.

Description

technical field [0001] The invention relates to the technical field of unmanned aerial vehicles, in particular to a target detection method based on a micro air vehicle. Background technique [0002] UAV system is a technology-intensive industry in the aerospace manufacturing industry, and its design and production process integrates a variety of high-tech systems. [0003] In the military, unmanned aerial vehicles have become the biggest bright spot in the weapons and equipment of various countries. Due to the characteristics of low operating cost, no risk of casualties, and easy use, UAVs can be widely used in anti-terrorism and riot control, emergency rescue and disaster relief, aerial survey and aerial photography, land and resource management, environmental protection, urban planning and management, film and television, and power departments. Line inspection, maritime patrol, highway inspection, fire prevention in forestry department, pest monitoring and control in agr...

AI Technical Summary

Problems solved by technology

The error of UAV attitude recognition limits the control of UAV to UAV to a certain extent, thus affecting the large-scale application of UAV

[0008] Moreover, the target detection algorithm of the existing visual recognition system on the UAV is complex and requires a large amount of computing resources as support. Therefore, the real-time and accuracy of target recognition need to be improved

Images