Human Detection Using Depth Window Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing human detection methods in images are limited by variations in human appearance, pose, and viewing angle, leading to high false alarm and missed detection rates, and are computationally expensive, requiring large training databases and fast processors.
Innovation Solution
The use of RGBD images from depth cameras to segment images into varying-sized windows, estimate subject distance, filter windows based on expected human size, and apply a cascade classifier with linear verification using context features, reducing computational overhead and improving detection accuracy across different angles and poses.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing human detection methods are used, then detection can be performed in certain situations, but the methods are computationally expensive and require large training databases
Solution Approach 1:
The image is segmented into multiple windows of varying sizes, and each window is processed independently through filtering and classification stages. This segmentation allows the system to focus computational resources on relevant regions while reducing overall complexity through parallel processing of smaller units.
Solution Approach 2:
A filtering operation is performed before classification to reject windows that fall outside the expected size range for human bodies at given distances. This preliminary action eliminates obviously non-human windows early in the process, reducing the computational burden on subsequent classification stages.
2Reliability
If existing human detection methods are used, then detection can be performed, but false alarm and missed detection rates are unacceptable
Solution Approach 1:
The system incorporates depth information from RGBD images as an additional dimension beyond traditional 2D image data. This extra dimensional information provides geometric constraints that help distinguish human bodies from other objects, improving detection accuracy and reducing false alarms while being particularly effective for detecting humans from various viewing angles.
Solution Approach 2:
The system changes the parameter space by using depth values and calculating expected window size ranges based on distance and focal length. By transforming the detection problem into a parameter-based filtering approach rather than relying solely on appearance features, the system achieves better robustness to variations in human appearance, clothing, pose, and illumination.
3Reliability
If depth information is used to improve detection robustness, then detection accuracy improves, but processing complexity may increase
Solution Approach 1:
The system extracts only the essential depth information needed for distance estimation and window size filtering, rather than processing all depth data. By selectively using depth values to calculate expected size ranges and filter windows, the system leverages depth information for improvement while avoiding the processing complexity of comprehensive depth analysis.
Data Source
AI summary
Techniques are provided for detection of humans in images that include depth information. A methodology embodying the techniques includes segmenting an image into multiple windows and estimating the distance to a subject in each window based on depth pixel values in that window, and filtering to reject windows with sizes that are outside of a desired window size range. The desired window size range is based on the estimated subject distance and the focal length of the depth camera that produced the image. The method further includes generating classifier features for each remaining windows (post-filtering) for use by a cascade classifier. The cascade classifier creates candidate windows for further consideration based on a preliminary detection of a human in any of the remaining windows. The method further includes merging neighboring candidate windows and executing a linear classifier on the merged candidate windows to verify the detection of a human.