# US 2014/0152821 A1 · Pedestrian detection systems and methods

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2014/0152821 A1
- Document ID: US20140152821A1
- Status: Published application
- Kind: A1
- Application number: 13/689,770
- Filed: 2012-11-30
- Published: 2014-06-05
- Earliest claimed: 2012-11-30
- Assignees: National Chung Shan Institute of Science and Technology
- Inventors: LI-CHEN FU; Pei-Yung Hsiao; Yi-Ming Chan; Min-Fang Lo
- Web page: https://patentgrep.com/patent/US20140152821A1

## CPC

- G06V 40/103: {Static body considered as a whole, e.g. static pedestrian or occupant recognition}
- G06V 10/50: by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
- G06V 20/58: Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING

## Abstract

A pedestrian detection system and method includes: dividing an image to a plurality of granules, and counting magnitude difference value of each granule in diagonal orientation to obtain features of HOGG. And the HOGG and the HOG captured can work together to improve the detection rate and reduce the false alarm rate, which is the ultimate goal of the vision based pedestrian detection.

## Claims

### Claim 1

A pedestrian detection system, comprising:
a camera module, mounted in a vehicle, and providing a image captured around the vehicle; and
a process module, having two feature capturing modules and a classifier, and receiving and analyzing the image so as to determine whether having a pedestrian in the image, and wherein each feature capturing module adopts histogram of oriented gradients and histogram of gradient of granule feature to capture features of the image.

### Claim 2 (depends on 1)

The pedestrian detection system of claim 1, wherein the classifier performs classifications to determine whether having the pedestrian in the image.

### Claim 3 (depends on 1)

The pedestrian detection system of claim 1, wherein the process module uses a display module to display results of the image analyzed.

### Claim 4

A pedestrian detection method, comprising:
capturing a image from a camera module;
capturing gradient of granule features of the image, and converting the image to a HOGG feature; and
classifying the HOGG feature by a classifier for determining whether having a pedestrian.

### Claim 5 (depends on 1)

The pedestrian detection method of claim 1, wherein the step of “capturing gradient of granule features of the image, and converting the image to a HOGG feature” further comprises:
dividing the image to a plurality of granules;
obtaining a magnitude average value of each granule;
obtaining a magnitude average difference value in diagonal orientation of the each granule in the plurality of cells;
taking a block as unit and counting feature vectors of each block so as to obtain the HOGG feature.

### Claim 6 (depends on 5)

The pedestrian detection method of claim 5, wherein size of the each granule is smaller than the each unit, and size of the unit is smaller than the each block.

### Claim 7 (depends on 4)

The pedestrian detection method of claim 4, wherein the step of “capturing gradient of granule features of the image, and converting the image to a HOGG feature” further comprises:
Simultaneously capturing a HOG feature, and converting the image to a HOG feature.

### Claim 8 (depends on 7)

The pedestrian detection method of claim 7, wherein the HOG feature and the HOGG feature are combined by a combination module.
