Ground unmanned platform autonomous following system based on deep learning detection tracking

Abstract

The invention discloses a ground unmanned platform autonomous following system based on deep learning detection tracking. The system comprises a ground unmanned platform, a three-axis self-stabilization holder, a depth camera, a data processing module and a power supply module. According to the invention, the target to be tracked can be autonomously selected from a plurality of candidate targets, and long-time stable and accurate following is realized.

Description

technical field [0001] The invention belongs to the technical field of autonomous follow-up systems for ground unmanned platforms, and in particular relates to an autonomous follow-up system for ground unmanned platforms based on deep learning detection and tracking. Background technique [0002] The autonomous tracking system of the ground unmanned platform can independently obtain the position information of the target to be tracked and follow the movement of the target to be tracked. The ground unmanned platform independently follows the user's movement, which improves the human-computer interaction capability, can greatly reduce the user's control burden and reduce control errors; free the user's hands, and the user can engage in other work. The present invention can be used in the military field, such as unmanned weapon equipment accompanying mobility, transporting the wounded, transporting materials, etc.; can be used in the commercial field, such as accompanying servi...

AI Technical Summary

Problems solved by technology

Among them, autonomous tracking based on lidar and autonomous tracking based on point cloud maps have problems such as high hardware cost, limited tracking accuracy, and the need for auxiliary personnel to manually specify targets.

Autonomous tracking based on visual images currently uses methods such as target position continuity, target color features, and target detection networks. It has the advantages of high target positioning accuracy, high tracking accuracy, and good system robustness, but there are still some problems that need to be solved urgently: Before starting the tracking, it is necessary for the assistant to manually specify the target; after the target is lost, it is difficult to recapture the designated target, and the assistant needs to re-designate the target; using the continuity of the target position and the autonomous follow-up of the color characteristics of the target, it is easy to be blocked by the target during the tracking process. Interference and other influences make it difficult to follow independently for a long time; using the autonomous follow of the target detection network, its network structure is mostly designed for multi-type target detection, the structure is complex, and the real-time performance is limited

The ground platform is greatly affected by the terrain, and it is often difficult to move to the theoretical pose under complex terrain, which may cause the tracked target to move out of the sensor's field of view and be lost

Images