3D Point Cloud Tracking with Recurrent Neural Network
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing 3D point cloud tracking systems fail to reconstruct and predict point clouds when objects are shaded or out of LiDAR's field of view, especially in complex environments with multiple moving targets, leading to loss of obstacle avoidance and tracking functions.
Innovation Solution
A three-dimensional point cloud tracking apparatus and method utilizing a recurrent neural network (RNN) that receives and processes LiDAR-scanned point clouds to reconstruct and predict the environment by combining observed and memory point clouds, employing sparse convolution operations and weight updates to handle incomplete data and complex movements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If LiDAR scanning is used to obtain three-dimensional point clouds, then the surface shape of scanned objects can be described, but the system cannot establish point clouds when objects are shaded or out of field of view
Solution Approach 1:
The system performs preliminary actions by storing historical point cloud data in advance and using it to predict and reconstruct current point clouds when LiDAR scanning fails, thereby maintaining continuous environmental representation despite shading or field of view limitations
Solution Approach 2:
The processor acts as an intermediary by combining historical point cloud data with current LiDAR scanning data to reconstruct complete point clouds when direct scanning is blocked, mediating between incomplete current data and the need for complete environmental representation
2Adaptability or versatility
If traditional tracking methods are used, then simple environments can be tracked, but the system fails in complex environments with multiple moving targets
Solution Approach 1:
The system applies dynamics by using a recurrent neural network that can adapt to changing environmental conditions and multiple moving targets, allowing the tracking system to dynamically adjust to complex scenarios while maintaining reliable tracking functionality
Solution Approach 2:
The system changes parameters by transforming point cloud data into a format suitable for neural network processing and using learned parameters from historical data to improve tracking performance in complex environments with multiple moving targets
3Measurement precision
If point cloud reconstruction is performed using recurrent neural network, then accurate prediction in shaded areas can be achieved, but the device complexity increases
Solution Approach 1:
The system replaces traditional mechanical LiDAR scanning with a neural network-based prediction system that uses historical data to reconstruct point clouds, achieving accurate measurement in shaded areas while the increased computational complexity is managed through efficient algorithm design
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables accurate reconstruction and prediction of 3D point clouds even in shaded areas and complex environments, effectively tracking moving objects with non-constant velocities and maintaining tracking functionality over time.
Implementation Method 1
three-dimensional laser scanners are also referred to as 'LiDARs,' which rapidly acquire a large number of points on the surface of a scanned object mainly using a sensed reflected laser beam
Data Source
AI summary
The embodiments of the present invention provide a three-dimensional point cloud tracking apparatus and method using a recurrent neural network. The three-dimensional point cloud tracking apparatus and method can track the three-dimensional point cloud of the entire environment and model the entire environment by using a recurrent neural network model. Therefore, the three-dimensional point cloud tracking apparatus and method can be used to reconstruct the three-dimensional point cloud of the entire environment at the current moment and also can be used to predict the three-dimensional point cloud of the entire environment at a later moment.