Dynamic Scene Occupancy Analysis With Split Static-Dynamic Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Bayesian Occupancy Filter (BOF) algorithms face challenges in efficiently analyzing dynamic scenes, particularly in moving vehicles, due to spatio-temporal aliasing issues and the need for high-resolution grids to detect slow movements and small objects, which increases computational and memory requirements.
Innovation Solution
The method determines the second probability of occupancy and movement probabilities by solving an equation that splits the calculation into static and dynamic parts, using fictitious particles and updating their positions and speeds based on previous iterations, reducing the computational load and memory resources needed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-resolution occupancy grids and motion histograms are used to detect slow movements and small objects, then measurement precision is improved, but device complexity and computational workload increase significantly
Solution Approach 1:
The patent segments the motion distribution grid into multiple sub-grids, where each sub-grid corresponds to a specific velocity range. This segmentation allows the system to maintain high-resolution detection capabilities for specific speed ranges while reducing the overall computational burden by distributing calculations across multiple smaller sub-grids rather than processing a single high-resolution grid.
Solution Approach 2:
The patent applies partial action by selectively updating motion histograms only for cells that contain dynamic objects, rather than updating all cells in the occupancy grid. The system identifies cells with moving objects and performs detailed motion analysis only on those specific cells, reducing unnecessary computations in static areas while maintaining detection precision for dynamic elements.
2Measurement precision
If high-resolution occupancy grids are used to detect slow movements and small objects, then measurement precision is improved, but memory resources increase
Solution Approach 1:
The motion distribution grid is divided into multiple sub-grids, each storing probability distributions for specific velocity ranges. This segmentation reduces the memory footprint of each individual sub-grid while collectively maintaining the resolution needed for detecting slow movements and small objects, as each sub-grid only needs to store data for a portion of the velocity spectrum.
Solution Approach 2:
The patent applies local quality by maintaining different resolution levels in different parts of the occupancy grid. Cells containing dynamic objects receive higher-resolution motion analysis, while static cells use lower-resolution or cached representations. This allows the system to optimize memory usage by allocating high precision only where dynamically needed rather than uniformly across the entire scene.
3Adaptability or versatility
If the size of the motion distribution grid is increased to represent speeds from -2Vmax to +2Vmax, then adaptability is improved, but computational workload increases significantly
Solution Approach 1:
The motion distribution grid is segmented into multiple sub-grids, each responsible for a specific velocity range (e.g., -Vmax to 0, 0 to +Vmax, or further divided). This segmentation maintains the adaptability to represent the full speed range from -2Vmax to +2Vmax while reducing computational workload by processing smaller sub-grids in parallel or sequentially, rather than processing one large grid at full resolution.
Solution Approach 2:
The patent implements dynamic adaptation by adjusting the number and resolution of sub-grids based on the actual velocity ranges observed in the scene. When objects move at high speeds, the system allocates more computational resources to relevant velocity ranges, while reducing resolution for ranges where no dynamic objects are present, thereby maintaining adaptability while optimizing computational efficiency in real-time.
Data Source
Figure 1~2
Figure 3~5
AI summary
A method for analysing a dynamic scene partitioned into cells which involves determining a probability of occupancy of a cell and a probability or probabilities of movement of the cell by solving the equation P(OV|ZC)= ΣA0-1V-1 P(CA00-1VV-1Z) / ΣA00-1VV-1 P(CA00-1VV-1Z) comprising the determination of the speeds and positions of dummy particles in the grid depending on those determined at the (k-1)th iteration and the probability P(V|V- 1); the determination of the particles located in each cell depending on the determined positions, and the solving of the equation, for a cell, is split into the solving of a static part corresponding to P(0=empty, V=0|ZC) and P(0=occupied, V=0|ZC) and the solving of a dynamic part corresponding to P(0=occ, V= v k i ,|ZC), i = 1 to nk, in which nk is the number of particles determined in cell C for the kth iteration.