Semi-Automatic Target Initialization Using Visual Saliency

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing tracking algorithms require predetermined target location and size for initialization, leading to performance issues due to user errors and mislearning of target appearance, especially in real-time applications where erroneous input is common and can result in premature track losses.

Innovation Solution

A semi-automatic target initialization method using visual saliency, which selects the most salient region bounding box through center-surround difference based initial window selection, geodesic saliency map calculation, and thresholding to compensate for user input errors and enhance discrimination between target and background.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If automatic segmentation and saliency detection algorithms are used to determine target location and size, then target initialization accuracy is improved, but processing time increases and real-time performance deteriorates

Engineering Contradiction:
Improvetarget initialization accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent applies partial action by using a simplified saliency detection approach that focuses only on the most critical visual features rather than comprehensive analysis. The system performs saliency detection on a downsampled version of the image or only on key regions, achieving sufficient target initialization accuracy without the full computational cost of complete automatic segmentation, thus maintaining real-time performance.

Inventive Principle:
Principle #16Partial or excessive action

Solution Approach 2:

The patent segments the image processing task into multiple stages: first performing coarse target location using simplified saliency detection, then refining the target bounding box in a second stage. This multi-stage segmentation allows the system to achieve accurate target initialization while controlling overall processing time by distributing computational load across stages.

Inventive Principle:
Principle #1Segmentation

2Measurement precision

If comprehensive saliency detection with multiple features is used, then target discrimination accuracy is improved, but computational complexity increases

Engineering Contradiction:
Improvetarget discrimination accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies local quality by using different feature extraction strategies for different regions of the image. For regions identified as potentially containing the target, the system applies comprehensive multi-feature saliency detection. For other regions, simplified or no processing is performed. This localized approach maintains high target discrimination accuracy while reducing overall computational complexity.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The system uses partial action by selectively applying complex saliency detection features only where necessary for target discrimination. Rather than computing all possible visual features across the entire image, the system identifies key regions and applies enhanced feature analysis only to those areas, achieving sufficient discrimination accuracy with reduced computational burden.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP3039645B1A semi automatic target initialization method based on visual saliency
Publication Date: 2017.05.17 ASELSAN ELEKTRONIK SANAYI & TICARET ANONIM SIRKETI
  • EP3039645B1 patent drawingFigure 1
  • EP3039645B1 patent drawingFigure 2
  • EP3039645B1 patent drawingFigure 3

AI summary

Target initialization can dramatically change the performance of the tracker, since the initial window determines for the tracker what to track. In order to achieve a better tracking performance; The present invention relates to a method of semi automatic target initialization based on visual saliency for a given point coordinate in the vicinity of target by the user. Performance boost of tracker is mainly based on two key features of target initialization algorithm: It is capable of compensating erroneous user input; also selecting the most distinctive, salient part of object as target, so better discrimination is achieved between the target and background. Experimental results show that tracking performance is boosted in scenarios, in which the tracking is initialized by the proposed algorithm. Very low computational cost and requirement of only a point coordinate as input in the neighborhood of the target make this approach preferable in real time tracking applications.