KR20170094837A - Apparatus and Method for detecting human using image processing - Google Patents

Abstract

The present invention relates to a human detecting method using image processing which reduces a calculation amount. According to an embodiment of the present invention, the human detecting method using image processing comprises: a first step of acquiring an input image; a second step of detecting an edge in the input image to generate an edge image; a third step of expanding an edge thickness of the edge image to a preset thickness to generate an expansion edge image; a fourth step of copying an image of an area corresponding to a first area of the input image in the first area where an edge of the expansion edge image exists to acquire a detection target image; a fifth step of performing a modified census transform (MCT) on the first area of the detection target image; and a sixth step of detecting as a person a second area corresponding to a preset MCT value range for the first area of the detection target image on which the MCT is performed.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention [0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention [0002] The present invention relates to an apparatus and method for detecting a human using image processing, and more particularly, to an apparatus and method for detecting a pedestrian from an image obtained through a camera mounted on a vehicle.

Although methods for detecting a person using an appearance-based detection algorithm in a photographed image are known, such an algorithm scans a human-shaped object in the entire image area, and thus the amount of computation is large. In particular, it is very difficult to process high-quality camera images in real time.

In general, human detection algorithms operate on PCs and are directed toward optimizing hardware algorithms (especially using Multi Core CPUs and GPUs) to increase speed.

However, black boxes mounted on vehicles or vehicles use embedded processors such as ARM in terms of economic aspect and weight minimization, which have a relatively low processing speed compared to PC. Therefore, it is difficult to embed human detection algorithms which have a large amount of computation.

Therefore, it is required to develop a method that can easily detect a human from a low - end processor.

SUMMARY OF THE INVENTION The present invention has been made to overcome the above-mentioned problems, and it is an object of the present invention to provide a method for detecting a person in an image having a small amount of calculation.

It is to be understood that both the foregoing general description and the following detailed description of the present invention are exemplary and explanatory and are not intended to limit the invention to the precise form disclosed. It can be understood.

According to an aspect of the present invention, there is provided a method of detecting a human using image processing, the method including generating an edge image by detecting an edge of the input image, expanding the edge thickness of the edge image to a predetermined thickness, Generating an extended edge image by copying an image of a region corresponding to the first region of the input image to a first region where an extended edge of the extended edge image exists, Performing MCT (Modified Census Transform) on the first region of the target image and detecting a second region corresponding to a predetermined MCT value range for the first region of the detection subject image on which the MCT has been performed as a person . ≪ / RTI >

The method may further include removing an edge region of the input image before generating the edge image.

Also, in the step of generating an edge image, the edge can be detected by a Canny operation.

In addition, after the step of detecting the second region as a person, the step of performing the MCT after reducing the size of the detection subject image to a predetermined ratio and the step of detecting the second region as a person can be performed again have.

Also, the predetermined MCT value may be determined in an Adaboost learning process.

In addition, the step of detecting the second region as a person may detect the second region by sliding the detection subject image on which the MCT has been performed to a window of a predetermined size.

If the difference between the detected positions of the second regions in the detection object images of different sizes is within a preset distance, The second region may be regarded as a duplicate.

Embodiments of the present invention can provide a user with an in-video person detection method with a small amount of computation.

Concretely, not only a low-specification process can be performed, but also a person can be detected without reducing a high-resolution image to a low resolution.

In addition, the candidate region of the object detection can be remarkably reduced through edge detection which is relatively simple in the image processing algorithm and easy to calculate.

In addition, the speed of the algorithm can be shortened by reducing the area to be discriminated, and unnecessary erroneous detection (for example, detection of a person in the sky) can be reduced.

It should be understood, however, that the effects obtained by the present invention are not limited to the above-mentioned effects, and other effects not mentioned may be clearly understood by those skilled in the art to which the present invention belongs It will be possible.

BRIEF DESCRIPTION OF THE DRAWINGS The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate preferred embodiments of the invention and, together with the description, serve to further the understanding of the technical idea of the invention, It should not be construed as limited.
1 is a flowchart illustrating a method of detecting a person using image processing according to an embodiment of the present invention.
2 is a flowchart illustrating a method of generating a detection subject image from an input image according to an exemplary embodiment of the present invention.
3 shows an input image according to an embodiment of the present invention.
4 shows an edge image according to an embodiment of the present invention.
5 illustrates an extended edge image according to an embodiment of the present invention.
6 shows a detection subject image according to an embodiment of the present invention.
7 is a diagram illustrating a second area of a detection subject image according to an embodiment of the present invention.
8 is a block diagram illustrating a computing system that implements a method of detecting a person using image processing in accordance with an embodiment of the present invention.

Hereinafter, a preferred embodiment of the present invention capable of solving the conventional problems will be described with reference to the drawings. In addition, the embodiment described below does not unduly limit the content of the present invention described in the claims, and the entire structure described in this embodiment is not necessarily essential as the solution means of the present invention.

1 is a flowchart illustrating a method of detecting a person using image processing according to an embodiment of the present invention.

2 is a flowchart illustrating a method of generating a detection subject image from an input image according to an embodiment of the present invention.

FIG. 3 shows an input image according to an embodiment of the present invention, FIG. 4 shows an edge image according to an embodiment of the present invention, FIG. 5 shows an extended edge image according to an embodiment of the present invention, 6 shows a detection subject image according to an embodiment of the present invention.

7 is a diagram illustrating a second area of a detection subject image according to an embodiment of the present invention.

Hereinafter, a method of detecting a person using image processing will be described in detail with reference to Figs. 1 to 7. Fig.

A detection object image is generated from the input image (S100).

In step S100, a step of removing an area in which no human is detected is removed from the input image to reduce the amount of computation. A detailed method for generating the detection object image will be described below with reference to FIG. 2 to FIG.

First, an edge is detected in an input image to generate an edge image (S110).

The edge refers to an outline having a brightness difference between adjacent portions. A canny operation can be performed to generate an edge image as shown in FIG. 4 from the input image shown in FIG.

Next, an edge of the generated edge image is expanded to a preset thickness to generate an extended edge image (S120).

In step S120, an edge of the edge image is enlarged as shown in FIG. 5 to generate an extended edge image.

In FIG. 5, the portion indicated by white corresponds to the extended edge, and the portion indicated by the extended edge corresponds to the first region.

In general, an object with the same size as a human is included in the extended region of the edge portion, i.e., the extended edge. That is, an object having a large size such as a building does not include the whole building on the extended edge even if the edge portion is extended, but a person is included in the extended edge when the edge is small.

Next, the image of the region corresponding to the first region of the input image is copied to the first region where the edge of the extended edge image exists to obtain the detection target image (S130).

6 shows an example of a detection object image in which an original image is copied only in a portion where an extended edge of the extended edge image shown in FIG. 5 exists.

Since a person is included in the expanded edge area by expanding the thickness of the edge in step S120 and the region such as the sky or the road bottom is not included and processed to be excluded from the detection object, a detection object image in which the detection object area is significantly reduced is acquired .

Alternatively, the step of obtaining the input image and the step of setting the region of interest may be performed according to another embodiment of the present invention before performing step S100.

The input image may be obtained by a camera installed in the vehicle such as a black box of the vehicle, but the present invention is not limited thereto, and the input image may be a still image or a moving image.

In addition, the ROI setting step is a step for reducing a calculation amount by removing a portion where no human is detected.

The region of interest can be set to the region of interest excluding the top, bottom, left, and right edges of the image. In the case of a black box image, a sky is often displayed on the top due to a fixed aspect ratio, and a hood of a vehicle or the like appears below the image, so that a human image is often not input.

The detection target image can be generated from the region of interest set as described above.

Next, a process for detecting a person in a detection subject image will be described with reference to FIG.

MCT (Modified Census Transform) is performed on the pixel existence region of the detection subject image (S200).

Here, the pixel presence area corresponds to the first area in which the image copied in the extended area of the edge exists.

In step S200, an MCT is performed on the first area to extract image features.

Next, a second area corresponding to a predetermined MCT value (reference value) range is detected for the first area of the detection subject image in which the MCT is performed, and the second area is determined as a person (S300).

Here, the predetermined MCT value may be extracted through the Adaboost learning process. In step S300, a window of a specific size is moved and it is determined whether the image of the region where the corresponding window is located is human or not.

In this case, the size of the window is preferably the size of the window used to obtain the reference MCT value (for example, 50X50).

When it is determined, the MCT feature image (the detection target image in which the MCT is performed in step S200) is slid with a fixed size window to discriminate. At this time, no pixel-free portion (an area other than the first area in the detection object image) is not discriminated.

If the MCT value of the portion where the corresponding window is located corresponds to the predetermined MCT value range during the determination, the corresponding region (second region) is determined as a human. For example, the second area of FIG. 7 is an area in which the MCT value falls within a predetermined MCT value range and can be detected as a human.

On the other hand, since the size of a human being in the image is variable, the following procedure can be further performed according to another embodiment of the present invention.

The size of the detection object image generated in step S100 is reduced to a predetermined ratio.

In step S300, the size of the window for determining whether a person is a person is constant. If the size of the person is larger than the size of the window in the image, the person may not be judged as a person even though it is a person.

The MCT of step S200 is performed on the detection target image whose size is reduced, and the process of detecting the person of step S300 is performed.

As described above, the number of times of performing S200 and S300 again by reducing the size of the image is not limited to one, and a plurality of results can be obtained.

When a plurality of results are derived, the same person can be detected as a duplicate in detection target images of different sizes. Therefore, according to an embodiment of the present invention, duplicate results can be removed from a plurality of detection results.

That is, when the difference between the detected positions of the detection target images of different sizes is within a preset distance, it is considered that only the detection result is regarded as a valid result by considering the detected persons in the detection target images of different sizes as duplicates can do.

8 is a block diagram illustrating a computing system that implements a method of detecting a person using image processing in accordance with an embodiment of the present invention.

 8, a computing system 1000 includes at least one processor 1100, a memory 1300, a user interface input device 1400, a user interface output device 1500, (1600), and a network interface (1700).

The processor 1100 may be a central processing unit (CPU) or a memory device 1300 and / or a semiconductor device that performs processing for instructions stored in the storage 1600. Memory 1300 and storage 1600 may include various types of volatile or non-volatile storage media. For example, the memory 1300 may include a ROM (Read Only Memory) and a RAM (Random Access Memory).

Thus, the steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by processor 1100, or in a combination of the two. The software module may reside in a storage medium (i.e., memory 1300 and / or storage 1600) such as a RAM memory, a flash memory, a ROM memory, an EPROM memory, an EEPROM memory, a register, a hard disk, a removable disk, You may. An exemplary storage medium is coupled to the processor 1100, which can read information from, and write information to, the storage medium. Alternatively, the storage medium may be integral to the processor 1100. [ The processor and the storage medium may reside within an application specific integrated circuit (ASIC). The ASIC may reside within the user terminal. Alternatively, the processor and the storage medium may reside as discrete components in a user terminal.

The method of detecting a human using image processing as described above is not limited to the configuration and method of the embodiments described above, but the embodiments may be modified so that all or some of the embodiments are selectively As shown in FIG.

1000: Computing System
1100: Processor
1200: System bus
1300: Memory
1310: ROM
1320: RAM
1400: User interface

Claims (1)

Generating an edge image by detecting an edge in an input image;
Generating an extended edge image by expanding the edge thickness of the edge image to a predetermined thickness;
Obtaining an image of a detection object by copying an image of a region corresponding to the first region of the input image to a first region where an extended edge of the extended edge image exists;
Performing MCT (Modified Census Transform) on the first region of the detection subject image; And
And detecting a second region corresponding to a predetermined MCT value range for a first region of the detection subject image on which the MCT is performed as a person.

Images