Track prediction device and method based on DS evidence theory

Abstract

The invention provides a target trajectory prediction device and a method based on a DS evidence theory. The method comprises the steps of performing space-time registration according to a target detected by a sensor, extracting and classifying target features by using an improved convolutional neural network, performing feature registration and fusion on image target features and three-dimensional point cloud features based on the same frame, and judging whether non-common features need to be reserved, wherein the features include common features and non-common features; for the same target,associating the plurality of targets based on continuous front and back frames to obtain the same target of the continuous frames, and establishing a sample matrix for representing the predicted trajectory, including target trajectory prediction of a single sensor, trajectory prediction of target matching among the plurality of sensors, and trajectory prediction after trajectory fusion of the respective targets obtained by the single sensor; and performing correctness judgment on the predicted trajectory based on the spatial features, the image features and the target trajectory information byusing a DS evidence theory, and outputting an optimal trajectory. The accuracy of trajectory prediction can be improved.

Description

technical field [0001] The invention relates to the field of target detection and recognition, in particular to a trajectory prediction device and method based on DS evidence theory. Background technique [0002] The target prediction trajectory is to grasp the dynamics of the target movement in advance, and the accuracy of the planning and prediction trajectory will greatly affect the feasibility and effectiveness of decision-making. In the current field, the production of the planned forecast trajectory is calculated by constructing a target performance model, but there are still some differences between the predicted results and the actual operating results. [0003] The existing trajectory prediction method is to establish a trajectory differential equation solution model based on trajectory planning, and use numerical solution methods such as the Euler method, Runge-Kutta method, etc., to use the given initial value and solution step size. The trajectory differential e...

AI Technical Summary

Problems solved by technology

The advantage of offline forecasting is that it can realize forecasting in real time, and the forecasting speed is fast, but the disadvantage is that the accuracy of the forecasting results is poor, and if the system is slightly disturbed, there will be errors in the forecasting results

Images