Distance Image Access Control for Automotive Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing systems for automotive applications face a trade-off between processing speed and detection accuracy when using methods like thinning out disparity images for object detection, leading to reduced detection accuracy of road surfaces and objects.
Innovation Solution
An information processing apparatus that dynamically adjusts the number of accesses to a distance image based on the distance information, increasing access as distance increases and decreasing access as distance decreases, using a virtual voting method to generate disparity maps while maintaining detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the disparity image is thinned out before access to reduce the number of pixels, then processing speed is improved, but detection accuracy of road surface and objects is reduced
Solution Approach 1:
The patent applies local quality by differentiating processing intensity based on spatial location and object type. Different regions of the disparity image (e.g., road surface areas vs. object areas) are processed with different access densities. Similarly, different object types (pedestrians, vehicles, road equipment) have different detection requirements, allowing the system to allocate processing resources locally rather than uniformly across the entire image.
Solution Approach 2:
The patent implements dynamics by making the pixel access pattern adaptive rather than static. The system dynamically adjusts which pixels to access based on real-time detection needs, processing history, and local image characteristics. This dynamic approach allows the system to optimize between processing speed and detection accuracy by accessing only the necessary pixels for current detection tasks while maintaining the ability to access additional pixels when needed.
2Measurement precision
If all pixels of the disparity image are accessed to ensure detection accuracy, then detection accuracy is improved, but processing time increases
Solution Approach 1:
The patent extracts and processes only the essential pixels needed for detection rather than accessing all pixels. By identifying and extracting key pixels that contain critical detection information (such as pixels likely to contain object boundaries or significant features), the system achieves accurate detection while avoiding the time cost of processing redundant pixels.
Solution Approach 2:
The patent applies partial action by accessing a subset of pixels that is sufficient for accurate detection without being excessive. Rather than accessing all pixels (excessive), the system identifies and accesses only the necessary portion (partial) that provides adequate detection accuracy for the given scenario, optimizing the balance between processing time and detection quality.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
An information processing apparatus (13) includes an acquisition unit (1102) to acquire a distance image; and a data generation unit (1103) to access the distance image to acquire distance information, and generate data based on the acquired distance information. The data generation unit (1103) is configured to access at least one pixel of the distance image to acquire distance information of the accessed at least one pixel, and set a parameter for determining a number of access to the distance image based on a distance indicated by the acquired distance information of the accessed at least one pixel, such that the number of access to the distance image increases as the distance increases, and the number of access to the distance image decreases as the distance decreases.