Man-machine hybrid enhanced small target detection and tracking method and system

Abstract

The invention discloses a man-machine hybrid enhanced small target detection and tracking method and system, and the method comprises the steps: positioning a target region, and obtaining a small target region image; extracting specific image information of the small target area image to obtain a small target detail image; constructing a small target detection and tracking model based on wearableequipment, designing and training a lightweight neural network target detection model, combining a context information technology and a Kalman filtering algorithm, and using a small target detail image to optimize the detection and tracking capability of a small target; circularly obtaining new image information by utilizing the steps to serve as a training set for enriching the small target detection and tracking model, and performing migration training on the small target detection and tracking model by utilizing the updated training set to obtain a small target detection and tracking enhancement model. The system comprises an acquisition and identification module, a sensor module, a detection and tracking module and a communication module. According to the invention, the accuracy of small target detection and tracking in the image is improved.

Description

technical field [0001] The invention relates to small target detection and tracking in real-time images, in particular to a human-machine hybrid enhanced small target detection and tracking method and system. Background technique [0002] At present, real-time and high-precision target detection is an important and difficult technology that must be faced and solved in the field of computer vision. In recent years, the research and technology of neural network and deep learning technology in the field of computer vision have developed rapidly. Especially for the target detection task, the end-to-end real-time target detection neural network model has been realized, which is far superior in real-time and accuracy. traditional object detection methods. However, when applied in the actual military field, when faced with small platforms with limited computing power such as wearable devices, the existing deep network model cannot effectively detect small-scale targets after light...

AI Technical Summary

Problems solved by technology

[0006] (1) In view of the fact that the embedded device on the wearable device has limited computing power for the neural network model, it directly limits the use of the enhanced model backbone (backbone) to enhance the model's ability to detect small targets, which represents the common use in embedded devices. Under the ability, the target detection model can only focus more on data enhancement for the improvement of small targets

At the same time, data enhancement generally considers enhancing the overall data set, which has great limitations in improving the small target detection ability of the model;

[0007] (2) The existing ability to improve small target detection is usually only aimed at the structure of the deep learning network model itself (for example, adding the FPN context feature fusion part, etc.), without considering the capabilities of other auxiliary equipment (multi-focal segment sensors, etc.) , limited to the ability of the model itself, the detection ability of small targets is limited

Images