Image Tracking via Dynamic Masking Map for Partial Occlusion

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing object tracking systems face accuracy issues when target objects are partially masked by other objects, particularly in crowded scenes or lane changes, leading to reduced tracking precision.

Innovation Solution

An image processing device that generates a masking map to quantify the degree of masking for each block of the target object, allowing for weighted template matching to improve tracking accuracy by adjusting weights based on masking levels and updating the masking map dynamically.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If template matching is performed using conventional methods, then tracking can be implemented, but tracking accuracy is significantly reduced when the target object is masked by other objects

Engineering Contradiction:
Improvetracking accuracyVSAvoidmasking effect
Core Design Contradiction:
Measurement precisionVSObject-affected harmful factors

Solution Approach 1:

The template is divided into multiple blocks, and masking detection is performed for each block independently. This segmentation allows the system to identify which specific regions are masked while maintaining information from unmasked regions, thereby preserving tracking accuracy even when partial masking occurs.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The invention applies different processing strategies to different regions of the template based on their masking status. Unmasked blocks are used for reliable matching, while masked blocks are handled differently (e.g., excluded or given lower weight). This local differentiation maintains overall tracking accuracy by leveraging reliable regional information.

Inventive Principle:
Principle #3Local quality

2Reliability

If multiple templates are used to prevent accuracy reduction, then tracking robustness improves, but the method fails when the target vehicle is widely masked

Engineering Contradiction:
Improvetracking robustnessVSAvoidtracking accuracy under wide masking
Core Design Contradiction:
ReliabilityVSMeasurement precision

Solution Approach 1:

By segmenting the template into blocks and evaluating masking status for each block, the system can determine whether sufficient unmasked regions remain for accurate tracking. This approach maintains reliability by adapting to different masking scenarios rather than relying on a fixed multiple-template strategy.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The invention dynamically changes the masking detection parameters (such as threshold values or block sizes) based on the current masking situation. When wide masking is detected, the system adjusts its parameters to maximize the use of available unmasked regions, thereby maintaining tracking accuracy in previously failing scenarios.

Inventive Principle:
Principle #35Parameter changes

3Measurement precision

If verification of tracking correctness is performed after individual vehicle tracking, then final result accuracy can be determined, but accuracy cannot be maintained in cases of wide masking or complicated vehicle behavior

Engineering Contradiction:
Improvetracking verification accuracyVSAvoidhandling of wide masking and complicated behavior
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The invention performs preliminary masking detection and evaluation before final template matching. By identifying masked blocks in advance and adjusting the matching strategy accordingly, the system prepares for potential accuracy issues rather than attempting verification after tracking fails. This preliminary action enables the system to handle wide masking and complicated behaviors more effectively.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3483834B1Image processing device and image processing method
Publication Date: 2021.09.22 OMRON CORP
  • EP3483834B1 patent drawingFigure 1
  • EP3483834B1 patent drawingFigure 2
  • EP3483834B1 patent drawingFigure 3

AI summary

An image processing apparatus for tracking a target object to be tracked included in a plurality of consecutively captured images comprises an image acquisition unit configured to acquire an image; a template acquisition unit configured to acquire a template corresponding to a target object to be tracked included in the image; a masking map generation unit configured to generate a masking map serving as a map indicative of a degree of masking the target object to be tracked, which is included in the image, by another object; and a tracking unit configured to detect the target object to be tracked included in the image by using the template, and acquire over a plurality of images an amount of movement of the target object to be tracked and a direction of the movement of the target object to be tracked, based on both of the template and the masking map.