Image Analysis Task Segmentation for Crash-Induced Frame Loss

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image analysis systems suffer from long-time frame loss during video playback due to program crashes on computing nodes, leading to unreliability as entire image analysis tasks are discarded, resulting in significant data loss.

Innovation Solution

The central node splits image analysis tasks into sub-tasks upon a program crash, sequentially transmits these sub-tasks to other computing nodes, and discards only the error sub-tasks, retaining other images, thus minimizing data loss and improving reliability.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If the entire image analysis task is discarded upon program crash, then the computing node can be restored to proceed to the next task, but the entire task's images are lost resulting in long-time frame loss during video playback

Engineering Contradiction:
Improvereliability of image analysisVSAvoidloss of images
Core Design Contradiction:
ReliabilityVSLoss of information

Solution Approach 1:

The patent divides an image analysis task into multiple independent image analysis sub-tasks, each processing a single image. When a program crash occurs, only the failed sub-task is discarded rather than the entire task. This segmentation allows the system to retain and reprocess successful sub-tasks, significantly reducing image loss and frame loss duration during video playback.

Inventive Principle:
Principle #1Segmentation

2Ease of repair

If the entire image analysis task is discarded upon program crash, then the computing node can be restored quickly, but the duration of frame loss during video playback increases

Engineering Contradiction:
Improverestoration of computing nodeVSAvoidduration of frame loss
Core Design Contradiction:
Ease of repairVSLoss of time

Solution Approach 1:

By segmenting the task into sub-tasks, the system can quickly restore the computing node after a crash while minimizing frame loss. Only the failed sub-task needs to be reprocessed, allowing the majority of the video sequence to be recovered without prolonged frame loss.

Inventive Principle:
Principle #1Segmentation

3Productivity

If image analysis tasks are processed in parallel on multiple computing nodes, then processing efficiency is improved, but program crashes can cause entire task loss reducing reliability

Engineering Contradiction:
Improveprocessing efficiencyVSAvoidreliability of image analysis
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent maintains parallel processing efficiency by dividing tasks into sub-tasks that can be distributed across multiple computing nodes. Simultaneously, it improves reliability by ensuring that a crash on one node only affects individual sub-tasks rather than entire tasks, allowing other parallel operations to continue successfully.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent implements a selective discard and recover mechanism where only failed sub-tasks are discarded upon crash, while successful sub-tasks are retained and reprocessed. This minimizes data loss and maintains high reliability in parallel processing environments.

Inventive Principle:
Principle #34Discarding and recovering

Data Source

PatentEP3561761B1Image analysis method and device
Publication Date: 2026.02.11 HANGZHOU HIKVISION DIGITAL TECHNOLOGY CO LTD
  • EP3561761B1 patent drawingFigure 1~2
  • EP3561761B1 patent drawingFigure 3
  • EP3561761B1 patent drawingFigure 4-1~4-2

AI summary

The present application discloses an image analysis method and apparatus, upon a program crash on a computing node, the central node may acquire the image analysis task that results in the crash, split the image analysis task into multiple image analysis sub-tasks, transmit sequentially the image analysis sub-tasks to the computing node until a program crash occurs thereon. The central node then acquires an error image analysis sub-task that results in the crash on the computing node, and discard the error image analysis sub-task. Compared with relevant art, the image analysis method according to the embodiment of the present application discards only the crash resulting image from an image analysis task, while other images in the image analysis task are retained. As such, fewer images are lost, so that the duration of a frame loss during video playback can be reduced and the reliability of image analysis can be improved. The approach is applicable for image analysis.