LiDAR Object Recognition with Noise Count Memory
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object recognition technologies using LiDAR add noise to distance measurement points, causing stochastic resonance, but this can result in the overwriting of previously added noise data, leading to loss of feature values.
Innovation Solution
An object recognition device and method that adds noise around point clouds based on distance, using a noise count memory unit to store the number of added noises for each position, ensuring feature values are not lost.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiple noises are added to point clouds to improve recognition rate through stochastic resonance, then the recognition rate is improved, but the feature value added to the object may be lost due to overwriting of noise data
Solution Approach 1:
The patent introduces a new dimension by storing noise data in a three-dimensional structure (spatial coordinates + noise count) rather than overwriting in a two-dimensional plane. The noise count memory unit creates an additional layer of information storage that preserves historical noise data while adding new noise, effectively transforming the data structure from simple overwriting to cumulative accumulation across multiple dimensions.
Solution Approach 2:
The patent changes the parameter of noise data storage from binary (present/absent) to quantitative (noise count). By introducing a noise count memory unit that stores the number of times noise has been added at each position, the system transforms the data from simple presence indicators to cumulative counts, preventing information loss while maintaining the stochastic resonance effect.
2Measurement precision
If noise is added to distance measurement points to enhance feature values, then object recognition accuracy is improved, but the complexity of the data processing system increases
Solution Approach 1:
The patent segments the noise addition process into distinct functional modules: a noise generation unit that creates random noise values, a noise addition unit that applies noise to point clouds based on distance thresholds, and a noise count memory unit that stores cumulative noise information. This segmentation allows each module to perform its specific function independently, reducing overall system complexity while maintaining processing accuracy.
Solution Approach 2:
The patent implements preliminary action by pre-establishing distance thresholds and noise addition rules before processing begins. The system pre-determines which point clouds should receive noise based on their distance from the object, and pre-configures the noise generation parameters. This preliminary preparation simplifies the main processing loop and reduces real-time computational complexity.
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 object recognition by adding noise according to distance without losing feature values, enhancing recognition accuracy.
Implementation Method 1
By adding a noise, a feature value is added to an object, which causes a stochastic resonance phenomenon and thus improves the recognition rate of the object
Data Source
Figure 1
Figure 2A~2B
Figure 3
AI summary
An object recognition device 100 includes: a point cloud measurement unit 10 that measures the positions of point clouds included in an object and the distance to each of the point clouds; a noise adding unit 20 that adds a noise around the measured point clouds according to the distance to each of the point clouds; a noise count memory unit 30 that stores the total number and position of noises added by the noise adding unit; and a recognition unit 40 that recognizes the object based on the point clouds and the noises added to the point clouds.