Image processing apparatus and image processing method

JP2024152431A5Pending Publication Date: 2026-04-02CANON KK
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-04-14
Publication Date
2026-04-02

AI Technical Summary

Technical Problem

Existing tracking methods using deep neural networks (DNNs) struggle with maintaining tracking reliability when the target changes posture or moves rapidly, often leading to incorrect determination of a lost state due to fixed criteria that do not account for object similarity or movement patterns, resulting in excessive template updates and tracking failure.

Method used

A method that combines tracker and detector outputs to generate a dynamic lost state criterion by integrating likelihood and movement penalty maps, using a hierarchical neural network to determine the tracking target's state, incorporating objectness detection and motion estimation to adapt the lost determination to each image's unique conditions.

Benefits of technology

This approach allows for robust tracking by accurately determining the lost state, reducing false positives and negatives, and enabling timely template updates, thereby improving tracking stability and accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 00000000_0000_ABST
    Figure 00000000_0000_ABST
Patent Text Reader

Abstract

To provide a technique for generating an appropriate criterion as a determination criterion for determining whether or not a tracking target is in a lost state in which the tracking target cannot be identified.SOLUTION: An image processing apparatus is configured to: acquire first information including a likelihood of a tracking target among a plurality of first candidates of the tracking target in a target image using a tracking technique for tracking an object in the image; acquire second information including the likelihood of the tracking target among a plurality of second candidates of the tracking target in the target image using a detection technique for detecting the object in the image; generate a determination criterion for determining whether or not it is in a lost state based on the first information and the second information; and determine whether or not it is in the lost state for the target image using the determination criterion.SELECTED DRAWING: Figure 2
Need to check novelty before this filing date? Find Prior Art