Object Center Detection Using Unimodal Confidence Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection and tracking systems for autonomous vehicles face challenges in accurately determining object center points and generating bounding boxes due to noise susceptibility and resource-intensive processing, leading to inefficient navigation and potential safety issues.
Innovation Solution
The use of unimodal confidence values and masking techniques to determine candidate center points and pixels associated with objects, reducing the number of candidate bounding boxes through non-maximum suppression, and training machine-learned models to propagate accurate object data parameters for improved detection and tracking.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional object detection methods are used to generate bounding boxes from sensor data, then comprehensive object coverage is achieved, but noise susceptibility increases and measurement precision deteriorates
Solution Approach 1:
The patent extracts and focuses on the center point of objects rather than processing the entire bounding box. By identifying and prioritizing center point detection, the system removes unnecessary processing of peripheral regions that contribute more noise than signal, thereby improving measurement precision while reducing noise susceptibility.
Solution Approach 2:
The patent applies different processing qualities to different regions of the detection space. The center region receives higher processing priority and more refined detection algorithms, while peripheral regions use standard processing. This local quality differentiation improves overall measurement precision without proportionally increasing computational resources.
2Reliability
If all candidate bounding boxes are processed to ensure complete object detection, then detection completeness is improved, but processing time and computational resources increase
Solution Approach 1:
The patent extracts only the essential information needed for reliable detection - specifically the center point coordinates and confidence scores - rather than processing all candidate bounding boxes in full detail. This extraction approach maintains detection completeness for high-confidence objects while significantly reducing processing time.
Solution Approach 2:
The patent applies partial processing to candidate bounding boxes based on their confidence scores. High-confidence candidates receive full processing to ensure detection completeness, while low-confidence candidates receive minimal processing or are discarded. This partial action approach maintains reliability for critical detections while reducing overall processing time.
3Reliability
If multiple candidate bounding boxes are generated to ensure object detection, then detection coverage is improved, but device complexity and computational resources increase
Solution Approach 1:
The patent extracts only the center point information and confidence scores from multiple candidate bounding boxes, rather than processing all full bounding box parameters. This extraction simplifies the data structure and reduces processing complexity while maintaining detection reliability through confidence-based filtering.
Solution Approach 2:
The patent changes the parameter representation from full bounding box coordinates to center point coordinates with confidence scores. This parameter transformation simplifies the computational requirements and reduces device complexity while maintaining the ability to reliably detect and track objects through the confidence scoring mechanism.
Data Source
AI summary
Techniques for detecting and tracking objects in an environment are discussed herein. For example, techniques can include detecting a center point of a block of pixels associated with an object. Unimodal (e.g., Gaussian) confidence values may be determined for a group of pixels associated with an object. Proposed detection box center points may be determined based on the Gaussian confidence values of the pixels and an output detection box may be determined using filtering and/or suppression techniques. Further, a machine-learned model can be trained by determining parameters of a center pixel of the detection box and a focal loss based on the unimodal confidence value which can then be backpropagated to the other pixels of the detection.