Target detection method, model and device, trajectory tracking method and device, and electronic equipment

Abstract

The invention provides a target detection method, model and device, a trajectory tracking method and device, and electronic equipment, and the method enables a target detected by the model to cover more receptive fields through improving the structure of a target detection model and adding a hole coding module, thereby improving the detection efficiency of a small-size target or a target in a low-resolution feature map, reducing the false alarm rate of target detection throughthe head prediction module, improving the precision of target detection in the video, and can be particularly applied to detection of small targets in the video.

Description

technical field [0001] The invention relates to the field of target detection, in particular to a target detection, track tracking method, model and electronic equipment thereof. Background technique [0002] Computer vision has gradually become more and more important in people's lives. It is widely used in robot navigation, intelligent video surveillance, industrial inspection, aerospace and many other fields. Computer vision can reduce the consumption of human capital, which has important practical significance. In computer vision, target detection and target tracking is a popular direction of computer vision and digital image processing, and it is also the core part of the intelligent monitoring system. Through the target detection and target tracking of intelligent monitoring images, the monitoring requirements of various actual scenes can be realized. . [0003] Target detection refers to the use of computer vision technology to accurately detect the required target f...

AI Technical Summary

Problems solved by technology

[0003] Target detection refers to the use of computer vision technology to accurately detect the required target from the image. However, the current method of target detection in the field of computer vision mainly has the following problems: 1. When performing target detection on multiple frames of video frames, often The occurrence of object detection false positives, which means that current convolutional neural networks have difficulty suppressing false positives from poorly recognized locations

Specifically, take the situation of high-altitude parabolic or high-altitude falling objects as an example: take a bag of 250 grams of clothes falling freely from a height of 15 meters to do experiments, the results show that the clothes can make 3-5 mm thick tempered glass fall apart, It can be seen that high-altitude parabolic objects or falling objects are extremely harmful to society; however, the current identification of high-altitude parabolic objects or high-altitude objects often faces problems such as difficult investigation and evidence collection, and the inability to determine specific infringers, and the current monitoring technology cannot be accurate. It cannot accurately identify small moving targets in high altitude, and it cannot be well applied to the monitoring of high-altitude parabolic scenes.

Images