Depth-Based Reference Shape Selection for Detection Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing audience measurement systems face inefficiencies and inaccuracies due to the need to compare detected object outlines to a large number of reference shapes, leading to high computational costs and a high rate of false positives, which affects the accuracy of media exposure data.
Innovation Solution
The system reduces the number of comparisons by using depth information to select a subset of candidate reference shapes from a library, organizing the reference shapes by distance, and comparing detected object outlines only to those likely to match based on their depth values, thereby improving detection efficiency and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the system compares detected object outlines to a large number of reference shapes, then detection accuracy is improved, but computational cost and processing time increase significantly
Solution Approach 1:
The patent segments the large reference shape library into multiple subsets based on depth information. Instead of comparing detected outlines to all reference shapes, the system first determines the depth of the detected object and then compares it only to reference shapes at similar depths, dramatically reducing computational complexity while maintaining detection accuracy.
Solution Approach 2:
The system performs preliminary organization of reference shapes by depth before the actual detection process. By pre-grouping reference shapes into depth-based subsets, the system prepares the data structure in advance so that during detection, only relevant subsets need to be queried, avoiding unnecessary comparisons and reducing processing time.
2Measurement precision
If the system compares detected object outlines to a large number of reference shapes, then detection accuracy is improved, but computational cost increases significantly
Solution Approach 1:
The patent segments the large reference shape library into multiple subsets based on depth information. Instead of comparing detected outlines to all reference shapes, the system first determines the depth of the detected object and then compares it only to reference shapes at similar depths, dramatically reducing computational complexity while maintaining detection accuracy.
Solution Approach 2:
The system changes the parameter used for reference shape selection from a comprehensive approach to a depth-based filtering approach. By using depth as a filtering parameter, the system reduces the search space from the entire reference shape library to a small subset, significantly lowering computational cost while preserving detection accuracy through depth-matched comparisons.
3Measurement precision
If the system compares detected object outlines to a large number of reference shapes, then comprehensive detection is achieved, but false positive rate increases
Solution Approach 1:
The patent applies local quality by matching reference shapes to detected objects based on their specific depth characteristics. Instead of using a uniform comparison approach for all objects, the system selects reference shapes that are locally appropriate to the detected object's depth, improving detection reliability and reducing false positives while maintaining comprehensive detection coverage.
Solution Approach 2:
The system changes the parameter used for reference shape selection from a comprehensive approach to a depth-based filtering approach. By using depth as a filtering parameter, the system reduces the search space from the entire reference shape library to a small subset, significantly lowering computational cost while preserving detection accuracy through depth-matched comparisons.
Data Source
AI summary
Methods, apparatus, and articles of manufacture to detect shapes are disclosed. Example methods disclosed herein include determining a first likelihood of detection for a reference shape at a first distance from an image capturing device, and storing the reference shape in a first group of a plurality of groups in the database based on a comparison of the first likelihood of detection with a first threshold, the first group associated with the first distance. Disclosed examples also include selecting reference shapes of the first group to compare to an object outline detected in an image in response to a query, the query including a depth value associated with the detected object outline, the selecting of the reference shapes of the first group based on the depth value.