LiDAR Obstacle Detection via Sub-Range Segmentation and Idle Block Allocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current LiDAR systems face challenges in improving real-time performance and accuracy of obstacle detection due to limitations in scanning methods and resource allocation, particularly when processing high-resolution image and video data using deep learning technologies.
Innovation Solution
The proposed solution involves dividing the LiDAR detection range into multiple sub-ranges, detecting them in a preset sequence, and accumulating confidence in real-time to output identification results when the confidence exceeds a threshold, while also optimizing data processing by allocating calculation resources efficiently through idle block utilization and boundary integration of detection results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If LiDAR uses traditional scanning methods to detect obstacles, then the detection coverage is complete, but the real-time performance is poor due to long scanning cycles
Solution Approach 1:
The patent divides the LiDAR detection range into multiple detection sub-ranges (first detection sub-range, second detection sub-range, etc.) and processes them independently. This segmentation allows the system to detect obstacles in different regions simultaneously or with different priorities, reducing the overall detection time and improving real-time performance while maintaining complete coverage through sequential processing of all sub-ranges.
2Measurement precision
If deep learning technology is used to process image data for sensing, then the sensing accuracy is improved, but the calculation resources cannot be properly allocated leading to poor real-time performance
Solution Approach 1:
The patent segments the detection range into multiple sub-ranges and processes them in parallel using multiple calculation resources. This allows deep learning models to process different spatial regions simultaneously, properly allocating calculation resources across multiple tasks and improving both sensing accuracy through deep learning and real-time performance through parallel processing.
Solution Approach 2:
The patent introduces a temporal dimension by processing detection sub-ranges at different times or with different priorities. By organizing the processing sequence and resource allocation in a temporal dimension, the system can maintain high accuracy through deep learning while achieving real-time performance through efficient time-managed resource distribution.
3Productivity
If the LiDAR detection range is divided into multiple sub-ranges, then the real-time performance is improved, but the device complexity increases
Solution Approach 1:
The patent divides the detection range into multiple sub-ranges and assigns them to different processing threads or calculation blocks. This segmentation improves real-time performance by enabling parallel processing, while the complexity is managed through software-based task distribution rather than hardware complexity, making the system scalable and configurable.
Data Source
AI summary
This application provides a data processing method and apparatus and a storage medium. The data processing method includes: obtaining K idle calculation blocks in real time, where K is greater than or equal to 1; invoking first K pieces of detected data from a cache stack in a preset priority sequence of detected data, and inputting the detected data into the K idle calculation blocks; sequentially processing K pieces of detected data on the K idle calculation blocks in the preset priority sequence; and integrating sensing calculation results of the K pieces of detected data in real time based on a boundary relationship between detection ranges of the K pieces of detected data, and outputting a sensing result.