Object Detection System with Adaptive Image Segmentation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing systems fail to accurately separate background and objects in images captured by in-vehicle devices due to changing light conditions and complex textures, leading to inefficient object detection and excessive data communication when transmitting videos over mobile networks.

Innovation Solution

A terminal device separates candidate images containing target objects based on a first evaluation criterion and transmits only these images to a server, where a second evaluation criterion is used for accurate recognition, reducing data communication and enhancing detection efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If all captured images are transmitted to the server for object detection, then the object detection accuracy is improved, but the data communication amount increases excessively

Engineering Contradiction:

Improveobject detection accuracyVSAvoiddata communication amount

Core Design Contradiction:

Measurement precisionVSQuantity of substance

Solution Approach 1:

The system segments the image processing task into two parts: preliminary filtering is performed on the terminal device to identify candidate images containing target objects, and only these candidate images are transmitted to the server for comprehensive analysis. This segmentation reduces the data communication amount while maintaining detection accuracy.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The terminal device performs preliminary image processing and candidate selection before transmission to the server. By pre-processing images locally and identifying potential target objects using initial evaluation criteria, the system reduces the volume of data that needs to be communicated to the server.

Inventive Principle:

Principle #10Preliminary action

2Difficulty of detecting and measuring

If edge detection and color conversion are used to detect objects, then the object detection capability is improved, but the detection accuracy deteriorates under changing light conditions

Engineering Contradiction:

Improveobject detection capabilityVSAvoiddetection accuracy

Core Design Contradiction:

Difficulty of detecting and measuringVSMeasurement precision

Solution Approach 1:

The system dynamically adjusts the evaluation criteria for object detection based on the specific characteristics of each image and the changing light conditions. Rather than using fixed thresholds, the system adapts its detection parameters to accommodate variations in illumination, allowing accurate detection across different environmental conditions.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system changes the parameters used for image evaluation based on the detected conditions. When light conditions change, the system adjusts the evaluation criteria (such as color thresholds, contrast levels, or feature weights) to maintain accurate object detection despite the varying illumination.

Inventive Principle:

Principle #35Parameter changes

3Measurement precision

If comprehensive image processing is performed on the terminal device, then the object detection accuracy is improved, but the device complexity increases

Engineering Contradiction:

Improveobject detection accuracyVSAvoidprocessing complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The system segments the processing complexity between the terminal device and the server. The terminal device performs only the necessary preliminary evaluation and candidate selection, while the computationally intensive comprehensive analysis is performed on the server. This segmentation reduces the processing complexity burden on the terminal device.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The terminal device performs partial processing - enough to identify candidate images with potential target objects, but not complete comprehensive analysis. This partial action is sufficient to filter out non-relevant images, reducing the overall system complexity while maintaining detection accuracy.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS9984303B2Object detection system, object detection method, POI information creation system, warning system, and guidance system

Publication Date: 2018.05.29 HITACHI LTD

AI summary

Provided is a technique for, even when the lightness of an image contained in a video changes, accurately separating the background of the image and an object in the background and thus efficiently collecting images that contain a target object to be detected, and also suppressing the amount of data communicated between a terminal device and a server. In the present invention, a terminal device accurately separates the background of an image and an object (i.e., a target object to be detected) in the background so as to simply detect the object in the background, and transfers to a server only candidate images that contain the detected object. Meanwhile, as such simple target object detection may partially involve erroneous detection, the server closely examines the candidate images to identify the target object to be detected, and thus recognizes the object.