Small flying target detection method and device

Abstract

The invention relates to the field of target detection and positioning, and discloses a small flying target detection method and device, and the method comprises the steps: obtaining a first image and a second image containing a small flying target through employing a binocular camera; a target detection module of the general processor is utilized to input the first image into the pruned yolov5s network for target detection, and a first image plane coordinate of the target is obtained; based on an epipolar constraint principle of the binocular camera, obtaining a cutting range of the second image by using a preprocessing module of the general processor, cutting the second image according to the cutting range, and then obtaining a second image plane coordinate of the target by using a target detection module; and a target positioning module of the general processor is used to calculate a target space position according to the two image plane coordinates and the internal and external parameters of the binocular camera. Therefore, processing data of the second image can be reduced, memory occupation is reduced, detection efficiency and detection precision are improved, and the method is suitable for a general processor with limited computing resources.

Description

technical field [0001] The invention relates to the field of target detection and positioning, in particular to a small flying target detection method and device. Background technique [0002] Small flying targets usually refer to "low, small and slow" flying targets, which refer to flying targets with slow flight speed, small size, low flight altitude, and are not easily detected by military and civilian radars. In view of safety considerations in low-altitude airspace, it is necessary to detect and locate small flying targets with high efficiency and accuracy. [0003] Due to the high efficiency of self-learning features of convolutional neural networks, it has become the mainstream research direction in the field of small target detection. The convolutional neural network optimizes the network model mainly by deepening the network layer to improve the detection accuracy. However, with the deepening of the network level, the hardware requirements for training the model i...

AI Technical Summary

Problems solved by technology

When the detection target is a small flying target, the scene is often a large-resolution image to detect a small target, which is directly input to the detection network for detection. If simple downsampling is performed, data information is easily lost when the downsampling multiple is too large, and it is difficult for the network to learn the characteristics of the target. Information; the current sampling multiple is too small, and the computing resources required for the large number of feature maps that need to be stored in the memory for the forward propagation of the network cannot be guaranteed, and normal network training cannot continue

Images