Papers
Topics
Authors
Recent
Search
2000 character limit reached

Whole-body Detection, Recognition and Identification at Altitude and Range

Published 9 Nov 2023 in cs.CV | (2311.05725v1)

Abstract: In this paper, we address the challenging task of whole-body biometric detection, recognition, and identification at distances of up to 500m and large pitch angles of up to 50 degree. We propose an end-to-end system evaluated on diverse datasets, including the challenging Biometric Recognition and Identification at Range (BRIAR) dataset. Our approach involves pre-training the detector on common image datasets and fine-tuning it on BRIAR's complex videos and images. After detection, we extract body images and employ a feature extractor for recognition. We conduct thorough evaluations under various conditions, such as different ranges and angles in indoor, outdoor, and aerial scenarios. Our method achieves an average F1 score of 98.29% at IoU = 0.7 and demonstrates strong performance in recognition accuracy and true acceptance rate at low false acceptance rates compared to existing models. On a test set of 100 subjects with 444 distractors, our model achieves a rank-20 recognition accuracy of 75.13% and a TAR@1%FAR of 54.09%.

Citations (4)

Summary

  • The paper introduces an end-to-end system achieving a 98.29% F1 score for whole-body detection and recognition at distances up to 500m and pitch angles up to 50°.
  • It employs a pre-trained detector fine-tuned on the challenging BRIAR dataset to ensure robust performance in indoor, outdoor, and aerial scenarios.
  • The approach reduces false acceptance rates and integrates multimodal techniques to enhance biometric identification from varying altitudes and ranges.

The topic of whole-body detection, recognition, and identification at altitude and range involves the application of advanced computer vision and biometric techniques to identify individuals from a distance and at various angles. This is particularly challenging due to factors such as varying lighting conditions, occlusions, and the complexity of whole-body features. Let's synthesize findings from relevant papers to better understand the current advancements in this area.

Whole-body Detection and Recognition Systems

The paper "Whole-body Detection, Recognition and Identification at Altitude and Range" (2311.05725) provides a comprehensive end-to-end system for detecting and identifying individuals from distances up to 500 meters and at pitch angles up to 50 degrees. The system pre-trains its detector on common image datasets and fine-tunes on the challenging Biometric Recognition and Identification at Range (BRIAR) dataset. This approach ensures robustness across indoor, outdoor, and aerial scenarios, achieving high performance with an average F1 score of 98.29% at IoU = 0.7. The system's recognition accuracy is demonstrated to be effective even with low false acceptance rates.

Whole-body Pose Estimation

Pose estimation plays a critical role in both detection and recognition at a distance. The "Whole-Body Human Pose Estimation in the Wild" paper (Jin et al., 2020) introduces the COCO-WholeBody dataset with annotations covering the entire human body. The ZoomNet model, designed in this paper, significantly reduces model complexity and dataset biases by localizing dense landmarks across different body parts. Additionally, "AlphaPose: Whole-Body Regional Multi-Person Pose Estimation and Tracking in Real-Time" (Fang et al., 2022) offers techniques for accurate whole-body keypoint localization and tracking. Using methods such as Symmetric Integral Keypoint Regression (SIKR) and Parametric Pose Non-Maximum-Suppression (P-NMS), AlphaPose improves both speed and accuracy in real-time applications.

Applications and Methodological Enhancements

Further advancements are found in papers like "Person Re-identification in Aerial Imagery" (Zhang et al., 2019), which focuses on re-identifying individuals captured by UAVs. Utilizing a dataset named PRAI-1581, the paper proposes subspace pooling to create discriminative feature representations, highlighting the challenge and uniqueness of aerial person re-identification.

Similar methodologies are employed in "DOPE: Distillation Of Part Experts for whole-body 3D pose estimation in the wild" (Weinzaepfel et al., 2020), where independent experts for body, hand, and face are combined using a deep network. This enhances whole-body pose detection's accuracy and efficiency, crucial for applications requiring detailed human interaction analysis.

Complementary Technologies

Other relevant technologies include the use of mmWave radars in "mm-Pose: Real-Time Human Skeletal Posture Estimation using mmWave Radars and CNNs" (Sengupta et al., 2019). This method leverages radar to detect skeletal joints robust against adverse weather and lighting conditions, suitable for scenarios where visual data may be compromised.

In summary, whole-body detection, recognition, and identification at a range and altitude are advancing through integrative approaches combining robust datasets, real-time processing techniques, and multimodal sensor data. These advancements promise superior performance in real-world applications such as security, surveillance, and biometric identification from aerial platforms.

Paper to Video (Beta)

No one has generated a video about this paper yet.

Whiteboard

No one has generated a whiteboard explanation for this paper yet.

Open Problems

We haven't generated a list of open problems mentioned in this paper yet.

Continue Learning

We haven't generated follow-up questions for this paper yet.

Collections

Sign up for free to add this paper to one or more collections.