Method and device for target detection

• US 10,474,935 B2
• Filed: 01/17/2017
• Issued: 11/12/2019
• Est. Priority Date: 01/25/2016
• Status: Active Grant

• Current Assignee: Zhejiang Shenghui Lighting Company Limited (Tongxiang Shenghui Lighting Electrical Company Limited)
• Original Assignee: Zhejiang Shenghui Lighting Company Limited (Tongxiang Shenghui Lighting Electrical Company Limited)
• Inventors: Zhong, Jinyan, Shen, Jinxiang
• Primary Examiner(s): Tsai, Tsung Yin
• Application Number:
  US15/749,724
• Publication Number:
  US 20180232607A1
• Time in Patent Office: 1,029 Days
• CPC Class Codes: G06T 2207/20 Special algorithmic details
• CPC Class Codes: G06V 10/255 Detecting or recognising potential candidate objects based on visual cues, e.g. shapes
• CPC Class Codes: 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
• CPC Class Codes: G06V 30/196 using sequential comparisons of the image signals with a plurality of references
• CPC Class Codes: G06V 30/2504 Coarse or fine approaches, e.g. resolution of ambiguities or multiscale approaches
• CPC Class Codes: G06V 40/161 Detection; Localisation; Normalisation

Abstract

The present disclosure provides a method for target detection, including: obtaining an image captured by a camera, and zoning the image based on a mounting angle of the camera to obtain at least one image block; determining target-detection algorithms for the at least one image block based on positions of the at least one image block in the image; and performing target detection on the at least one image block based on the target-detection algorithms.

Claims

1. A method for target detection, comprising:

obtaining an image captured by a camera, and zoning the image based on a mounting angle of the camera to obtain at least one image block;

determining target-detection algorithms for the at least one image block based on positions of the at least one image block in the image, comprising;

determining whether a distance between a geometric center of one of the at least one image block and a geometric center of the image is smaller than a preset distance;

when the distance is smaller than the preset distance, selecting applying a multi-scale and multi-cascade based target-detection algorithm on the one of the at least image block, and

when the distance is greater than the preset distance, selecting applying a single-scale and single-cascade based target-detection algorithm on the one of the at least image block; and

performing target detection on the at least one image block based on the target-detection algorithms.

2. The method according to claim 1, wherein:

determining target-detection algorithms for the at least one image block based on positions of the at least one image block in the image includes determining a target-detection algorithm for each one of the at least one image block based on a position of each one of the at least one image block in the image; and

performing target detection on the at least one image block based on the target-detection algorithms includes performing a target detection on each one of the at least one image block based on the target-detection algorithm applied on each one of the at least one image block.

3. The method according to claim 1, wherein zoning the image based on a mounting angle of the camera to obtain at least one image block includes:

determining the mounting angle of the camera is within a preset range; and

when mounting angle of the camera is within the preset range, zoning the image to three rectangular blocks from top to bottom of the image, the three rectangular blocks having a same length as the image and having a width ratio of about G;

H;

I from top to bottom of the image, G, H, and I being positive integers and I>

H>

G.

4. The method according to claim 3, further including:

determining whether the mounting angle of the camera is greater than a preset angle; and

when the mounting angle of the camera is greater than the preset angle, zoning the image to three rectangular blocks from top to bottom of the image, the three rectangular blocks having a same length as the image and having a width ratio of about A;

B;

C from top to bottom of the image, A, B, and C being positive integers and B>

C>

A, and

when the mounting angle of the camera is not greater than the preset angle, zoning the image to three rectangular blocks from top to bottom of the image, the three rectangular blocks having a same length as the image and having a width ratio of about D;

E;

F from top to bottom of the image, D, E, and F being positive integers and F>

D>

E.

5. The method according to claim 1, wherein the at least one image block is categorized to a plurality of types based on significance of the at least one image block such that a portion of one of the at least one image block is proportionally correlated to a significance of the one of the at least one image block.

6. The method according to claim 5, wherein one of the at least one image blocks having a greatest portion in the width ratio is applied with a multi-scale and multi-cascade based target-detection algorithm having low speed and high precision;

one of the at least one image block having a moderate portion in the width ratio is applied with a multi-scale based target-detection algorithm having large-window; and

one of the at least one image block having a smallest portion in the width ratio is applied with a single-scale and single-cascade based target-detection algorithm having low precision and high speed.

7. The method according to claim 1, wherein performing target detection on the at least one image block based on the target-detection algorithms includes:

performing target detection on one of the at least one image block based on the target algorithm to obtain a detection result of at least one to-be-determined target;

multiplying the detection result by a corresponding weight value to obtain a weighted detection result; and

determining whether the weighted detection result is greater than a preset threshold value,

when the weighted detection result is greater than the preset threshold value, determining the at least one to-be-determined target to be final targets, and

when the weighted detection result is not greater than the preset threshold value, determining the at least one to-be-determined target to be non-final targets.

8. The method according to claim 1, wherein before zoning the image to obtain the at least one image block, further includes:

obtaining a height of the camera; and

determining whether the height of the camera is within a preset height range.

9. A device for target detection, comprising:

a memory and a processor coupled to the memory, the processor being configured to;

obtain an image captured by a camera, and zone the image based on a mounting angle of the camera to obtain at least one image block;

determine target-detection algorithms for the at least one image block based on positions of the at least one image block in the image, comprising;

determining whether a distance between a geometric center of one of the at least one image block and a geometric center of the image is smaller than a preset distance;

when the distance is smaller than the preset distance, selecting applying a multi-scale and multi-cascade based target-detection algorithm on the one of the at least image block, and

when the distance is greater than the preset distance, selecting applying a single-scale and single-cascade based target-detection algorithm on the one of the at least image block; and

perform target detection on the at least one image block based on the target-detection algorithms.

10. The device according to claim 9, wherein the processor is further configured to:

determine the mounting angle of the camera is within a preset range; and

when the mounting angle of the camera is within the preset range, zone the image to three rectangular blocks from top to bottom of the image, the three rectangular blocks having a same length as the image and having a width ratio of about G;

H;

I from top to bottom of the image, G, H, and I being positive integers and I>

H>

G.

11. The device according to claim 10, wherein the processor is further configured to:

determine whether the mounting angle of the camera is greater than a preset angle; and

when the mounting angle of the camera is greater than the preset angle, zone the image to three rectangular blocks from top to bottom of the image, the three rectangular blocks having a same length as the image and having a width ratio of about A;

B;

C from top to bottom of the image, A, B, and C being positive integers and B>

C>

A, and

when the mounting angle of the camera is not greater than the preset angle, zone the image to three rectangular blocks from top to bottom of the image, the three rectangular blocks having a same length as the image and having a width ratio of about D;

E;

F from top to bottom of the image, D, E, and F being positive integers and F>

D>

E.

12. The device according to claim 9, wherein the processor is further configured to:

categorize the at least one image block to a plurality of types based on significance of the at least one image block such that a portion of one of the at least one image block is proportionally correlated to a significance of the one of the at least one image block.

13. The device according to claim 12, wherein the processor is further configured to:

apply one of the at least one image blocks having a greatest portion in the width ratio with a multi-scale and multi-cascade based target-detection algorithm having low speed and high precision;

apply one of the at least one image block having a moderate portion in the width ratio with a multi-scale based target-detection algorithm having large-window; and

apply one of the at least one image block having a smallest portion in the width ratio with a single-scale and single-cascade based target-detection algorithm having low precision and high speed.

14. The device according to claim 9, wherein the processor is further configured to:

perform target detection on one of the at least one image blocks based on the target-detection algorithm applied on the image block, to obtain a detection result corresponding to at least one to-be-determined target.

15. The device according to claim 9, wherein the processor is further configured to obtain a height of the camera and determine whether the height of the camera is within a preset height range.