Object Detection Histogram Analysis Speed

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing object detection systems face challenges with prolonged calculation times and false positives due to inefficient indexing and uncertainty in object position, leading to time-consuming detection and potential false identification of objects.

Innovation Solution

An object detection apparatus that reduces calculation load by determining average dictionary complexities, maximum search pixel skipping amounts, and search window size changes, allowing for faster and more accurate detection by analyzing histograms and adjusting dictionary complexities and search window sizes based on normal distribution patterns.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If histogram analysis is performed on edge extraction frequencies in both horizontal and vertical directions, then object detection accuracy is improved, but calculation time is prolonged

Engineering Contradiction:
Improveobject detection accuracyVSAvoidcalculation time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent segments the histogram analysis by performing it only in the horizontal direction rather than both horizontal and vertical directions. This segmentation reduces the calculation scope while maintaining detection accuracy by focusing on the most critical directional information for object detection.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent extracts and utilizes only the essential histogram information from edge extraction frequencies in the horizontal direction, discarding the redundant vertical direction analysis. This extraction of key features maintains detection effectiveness while reducing computational burden.

Inventive Principle:
Principle #2Taking out (Extraction)

2Reliability

If a complex classifier with multiple stages is used, then detection rate and false positive reduction are improved, but detection speed decreases

Engineering Contradiction:
Improvedetection accuracy and false positive rateVSAvoiddetection speed
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent applies preliminary action by using a pixel skipping mechanism that pre-filters potential object regions before applying the full classifier. This preliminary filtering reduces the number of regions that require complex classification, thereby maintaining accuracy while improving detection speed.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent implements partial action by applying the complex classifier only to selected regions identified by the histogram analysis and pixel skipping mechanism, rather than to the entire image. This partial application maintains detection reliability while significantly reducing processing time.

Inventive Principle:
Principle #16Partial or excessive action

3Adaptability or versatility

If the search window size is increased to cover uncertain object positions, then detection coverage is improved, but calculation load increases

Engineering Contradiction:
Improvedetection coverageVSAvoidcalculation load
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent applies dynamics by making the search window size adaptive rather than fixed. The search window size is dynamically adjusted based on the histogram analysis results and detected object characteristics, allowing the system to cover uncertain positions when needed while reducing calculation load when object positions are well-defined.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP2038843B1Object detection apparatus, method and program
Publication Date: 2017.03.29 TOYOTA JIDOSHA KK
  • EP2038843B1 patent drawingFigure 1
  • EP2038843B1 patent drawingFigure 2
  • EP2038843B1 patent drawingFigure 3

AI summary

An object detection apparatus (1) for detecting an object from an image obtained by taking a front view picture of a road in a traveling direction of a vehicle includes a camera unit (10) for taking the front view picture of the road and inputting the image; a dictionary modeling the object; a search unit for searching the image with a search window; a histogram production unit for producing a histogram by comparing the image in the search window with the dictionary and counting a detection frequency in a direction parallel to a road plane; and a detection unit (22) for detecting the detection object by detecting a unimodal distribution from the histogram.