TWI743875B - Infrared recognition method for human body - Google Patents

Abstract

The present invention provides an infrared recognition method for human body. The infrared recognition method includes an irradiation step, an acquisition step, an image simulation step, an image analysis step, and a recognition step. The irradiation step is implemented by irradiating to a target to generate a plurality of laser tiles and a plurality of ambient light areas spaced from each other. The acquisition step is implemented by capturing to a basic image for the target. The image simulation step is implemented by obtaining to multiple image information corresponding to the ambient light areas by an interpolation method to obtain multiple simulation information. The multiple simulation information replaces the multiple image information of the corresponding laser areas to generate a simulation image. The image analysis step is implemented by comparing and analyzing to the basic image and the simulated image to generate an identification image. The recognition step is to pass the identification image through a classifier to recognize whether the target is a human body.

Description

Infrared body recognition method

The invention relates to a human body identification method, in particular to an infrared human body identification method that uses infrared laser light to identify whether a target is a human body.

The existing human body recognition method uses an infrared camera (also called an IR camera) to take two shots of a human body to obtain an ambient light image (NIR) without near-infrared light and an infrared light image with near-infrared light ( NIRL), and then identify the position of the human face in the ambient light image and the infrared light image through a face detection (face detection), and then compare the infrared light image corresponding to the human face The image information of the part position replaces the image information corresponding to the position of the human face in the ambient light image to generate a near-infrared differential image (NIRD) used as a recognition image. The classifier analyzes and recognizes the features (reflectivity) in the near-infrared light differential image to confirm whether the human body is a real human.

However, this method needs to take two shots to obtain the ambient light image and the infrared light image to obtain the near-infrared light differential image for identification. However, the human body cannot be Obviously shifted to avoid too large difference in the position of the human body in the ambient light image and the infrared light image, causing the face detector to be unable to distinguish the position of the human body, or caused by the shaking of the human body The near-infrared light differential image has afterimages, so that the classifier cannot confirm the features in the infrared light differential image, and cannot analyze and identify.

Therefore, the inventor believes that the above-mentioned shortcomings can be improved, and with great concentration of research and the application of scientific principles, we finally propose an invention with reasonable design and effective improvement of the above-mentioned shortcomings.

The technical problem to be solved by the present invention is to provide an infrared human body identification method for the shortcomings of the prior art, which can effectively improve the defects that may occur in the existing human body identification method.

The embodiment of the present invention discloses an infrared human body recognition method, which includes: implementing an irradiating step: irradiating a target with an infrared laser light, so that the outer surface of the target forms a plurality of laser blocks arranged at intervals, and The outer surface of the target is formed with an ambient light area irradiated by an ambient light between any two of the laser blocks; implementing a capturing step: using an image information capturing device on the target A basic image is captured on the outer surface, and the basic image has a plurality of first image information corresponding to a plurality of the laser tiles and a plurality of second image information respectively corresponding to a plurality of the ambient light regions ; Implement an image simulation step: in the basic image, the second image information corresponding to any two adjacent ambient light regions is obtained by an interpolation method to obtain a simulation information, and the simulation information replaces the corresponding two The first image information of the laser tiles between the two ambient light regions is used to generate a simulated image that simulates that the target is only illuminated by the ambient light; and an image analysis step is performed: The basic image and the simulated image are compared and analyzed to remove the image information of the ambient light to generate an identification image, the identification image having a plurality of reflections on the target corresponding to a plurality of the laser blocks Implementation of an identification step: the identification image is identified by a classifier; wherein the classifier is identified by a plurality of the reflectance information in the identification image to confirm whether the target is A human body.

The embodiment of the present invention further discloses an infrared human body recognition method, which includes: implementing an irradiation step: irradiating a target with an infrared laser light, so that the outer surface of the target forms a plurality of laser blocks arranged at intervals, and An ambient light area irradiated by an ambient light is formed between any two of the laser blocks on the outer surface of the target; and a capturing step is performed: an image information capturing device is used on the target A basic image is captured on the outer surface of, the basic image has a plurality of first image information corresponding to a plurality of the laser blocks and a plurality of second images respectively corresponding to a plurality of the ambient light regions Information; implementing an image analysis step: analyzing a plurality of the first image information and a plurality of the second image information in the basic image to generate an identification image, the identification image having a plurality of the lasers Block a plurality of reflectance information on the target; implement a recognition step: recognize the recognized image through a classifier; wherein the classifier passes a plurality of the reflectance in the recognized image The information is identified to confirm whether the target is a human body.

In summary, the infrared human body recognition method disclosed in the embodiment of the present invention can generate a plurality of the laser tiles and a plurality of the ambient light regions spaced apart on the target by the infrared laser. Method, the basic image including the infrared laser light and the ambient light image information can be acquired in the capturing step; accordingly, by directly analyzing the first image information in the basic image (That is, image information corresponding to a plurality of the laser tiles) and the second image information (that is, image information corresponding to a plurality of the ambient light regions), to obtain the high-precision recognition image, The classifier can directly recognize the features in the recognized image to maintain an accurate recognition rate.

In order to further understand the features and technical content of the present invention, please refer to the following detailed description and drawings about the present invention. However, the provided drawings are only for reference and description, and are not used to limit the present invention.

The following are specific examples to illustrate the implementation of the "infrared human body recognition method" disclosed in the present invention. Those skilled in the art can understand the advantages and effects of the present invention from the content disclosed in this specification. The present invention can be implemented or applied through other different specific embodiments, and various details in this specification can also be based on different viewpoints and applications, and various modifications and changes can be made without departing from the concept of the present invention. In addition, the drawings of the present invention are merely schematic illustrations, and are not drawn according to actual dimensions, and are stated in advance. The following embodiments will further describe the related technical content of the present invention in detail, but the disclosed content is not intended to limit the protection scope of the present invention.

It should be understood that although terms such as "first", "second", and "third" may be used herein to describe various elements or signals, these elements or signals should not be limited by these terms. These terms are mainly used to distinguish one element from another, or one signal from another signal. In addition, the term "or" used in this document may include any one or a combination of more of the associated listed items depending on the actual situation.

[First Embodiment]

As shown in Figures 1 to 8, which are the first embodiment of the present invention, this embodiment discloses an infrared human body recognition method, and the above infrared human body recognition method is to obtain the ambient light and Image information of infrared light for human body identification. To put it another way, the method of acquiring image information with ambient light and infrared light through more than one capture, or performing human body recognition without using infrared light and ambient light, is not the human body recognition method referred to in this embodiment.

The infrared human body recognition method includes: an irradiation step S110, a capturing step S120, an image simulation step S140, an image analysis step S150, a selection step S160, and an identification step S170. It should be noted that any one of the above multiple steps can be omitted or replaced by a reasonable change according to the needs of the designer.

Implement the irradiation step S110: irradiate a target G with an infrared laser light L, so that the outer surface of the target G forms a plurality of laser blocks LG arranged at intervals, and the outer surface of the target G An ambient light area EA illuminated by an ambient light E is formed between any two of the laser blocks LG. It should be noted that the outer surface of the target G is exposed to the ambient light E, that is, when the infrared laser light L irradiates the outer surface of the target G, each The location of the laser block LG also has the ambient light E instead of the infrared laser light L alone.

Specifically, a plurality of the laser patterns LG generated by the infrared laser light L are presented on the outer surface of the target G in a longitudinal stripe manner, and a plurality of the laser patterns The total number of blocks LG is not less than 6, and the ambient light area EA is formed between any two of the laser blocks LG, and the ambient light area EA is not affected by the infrared laser light L. The irradiated area is not limited to what is contained in this embodiment. For example, in other embodiments not shown in the present invention, a plurality of the laser blocks LG can be adjusted according to the needs of the designer. For example, the plurality of the laser blocks LG can also be horizontal The method is located on the outer surface of the target G, and the number is 20.

Furthermore, the infrared laser light L needs to select a wavelength of not less than 800 nanometers (nm), and the reflectance of the infrared laser light L on the skin surface (outer surface) that irradiates the human body is greater than that of the infrared laser light L. More than 5% of non-skin surface (such as silicone); and in this embodiment, the wavelength of the infrared laser light L is 840 to 860 nanometers, and the infrared laser light L irradiates the skin surface of the human body The reflectivity of is greater than 15% of the reflectivity of the infrared laser light L irradiating the non-skin surface, but it is not limited to that described in this embodiment.

To put it another way, as shown in the graph in Figure 2, this graph is used to show that the infrared laser light L of different wavelengths irradiates light-colored silicone surface, dark-colored silicone surface, light-colored skin surface, and dark-colored skin surface. Reflectance. The horizontal axis of this chart is wavelength (nm), and the vertical axis is reflectance (%); it should be noted that the surface of dark and light skin is used to represent the skin of different skin tones of the human body, while dark and light colored silicone The surface is made of silicone or other similar materials to form fake faces with different skin tones. From the chart, it can be clearly found that when the wavelength of the infrared laser light L is 840 to 860 nanometers, the reflectance of the surface of the dark and light-colored skin is significantly different from the reflectance of the surface of the dark and light-colored silicone.

Implement the capturing step S120: As shown in FIG. 3, a basic image P1 is captured on the outer surface of the target G through an image information capturing device C, and the basic image P1 has respective corresponding multiples. A plurality of first image information P11 of the laser block LG and a plurality of second image information P12 respectively corresponding to a plurality of the ambient light areas EA. In detail, the outer surface of the target G is exposed to the ambient light E, that is, the first image information P11 corresponding to each of the laser tiles LG includes the The image information of the infrared laser light L and the ambient light E, and each of the first image information P11 has only the image information of the ambient light E. It should be noted that the image information capturing device C can emit the infrared laser light L in this embodiment (as shown in FIGS. 1 and 9).

Performing the image simulation step S140: As shown in FIG. 4, in the basic image P1, the second image information P12 corresponding to any two adjacent ambient light areas EA is obtained by an interpolation method. Simulation information A. The simulation information A replaces the first image information P11 corresponding to the laser block LG between the two ambient light areas EA to generate a simulation that the target G is only affected by the environment A simulated image P2 illuminated by light E; to put it another way, the simulated image P2 removes each of the first image information P11 in the basic image P1, and then removes any two adjacent first image information P11 The second image information P12 calculates the analog information A using an interpolation method (the interpolation method), and the analog information A replaces the first image information P11 that has been removed between the two second image information P12.

Perform the image analysis step S150: referring to FIG. 5, the basic image P1 and the simulated image P2 are compared and analyzed to remove the image information of the ambient light E (that is, the plurality of second images Information P12 and a plurality of the simulation information A) are generated to generate an identification image P3, the identification image P3 has a plurality of reflectance information corresponding to the plurality of the laser blocks LG on the target; in detail The identification image P3 is the image information of the ambient light E in the basic image P1 minus the image information of the ambient light E in the simulated image P2, so that the identification image P3 does not have any The image information of the ambient light E only has the reflectance of the plurality of the laser blocks LG (the plurality of the first image information P11) on the outer surface of the target G.

Perform the selection step S160: as shown in FIG. 6, the identification image P3 is divided into a plurality of identification areas P31, each of the identification areas P31 has a plurality of the reflectance information; In the area P31, a part of the identification area P31 corresponding to the target G is selected, and each is defined as a detection area Y, and a plurality of the reflectance information in each detection area Y is calculated to obtain a Average reflectivity.

Specifically, the recognition image P3 defines a first direction D1 and a second direction D2 perpendicular to each other, and a plurality of the recognition regions P31 are arranged in M rows parallel to the first direction D1 and parallel to the first direction D1. In the N columns of the two directions D2, M and N are each a positive integer greater than 2; that is, the plurality of identification regions P31 are arranged in a matrix of M by N (checkerboard shape). As shown in FIG. 6, in this embodiment, a plurality of the identification regions P31 are arranged in 5 rows parallel to the first direction D1 and 4 columns parallel to the second direction D2, that is, a plurality of The number of identification areas P31 is 20, and the plurality of detection areas Y are the plurality of identification areas P31 except for the eyes corresponding to the target G; in detail, as shown in FIG. 6, The plurality of identification areas P31 are defined from top to bottom, from left to right, as a first identification area, a second identification area, a third identification area, ..., and a twentieth identification area, and more In this embodiment, the Y domains of the detection areas are the second recognition area, the third recognition area, the fourth recognition area, the twelfth recognition area, the thirteenth recognition area, the fourteenth recognition area, and the tenth recognition area. Seventh recognition area, eighteenth recognition area, and nineteenth recognition area, that is, a plurality of the detection areas Y are those that correspond to the target G except for the second column from top to bottom in FIG. 6 The identification area P31 is, but not limited to that described in this embodiment.

Furthermore, the identification image P3 is defined as a different identification area P31 at the middle 1/3 position corresponding to the target G from top to bottom (that is, the first direction D1). A detection area N, a plurality of the detection areas Y are not located in a plurality of the non-detection areas N. In detail, referring to FIG. 7, a plurality of the non-detection areas N are located in the second row from the top to the bottom of the identification image P3, but they are not limited to those in this embodiment. For example, in other embodiments of the present invention that are not shown, the identification images P3 are divided into 48 segments, and are arranged in 6 rows parallel to the first direction D1 and 8 rows parallel to the second direction D2. Columns (that is, arranged in a 6 by 8 matrix), the plurality of identification areas P31 located in the third and fourth columns from top to bottom are defined as the plurality of non-detection areas N. To put it another way, the plurality of non-detection areas N correspond to the eye positions of the target G (human body); accordingly, it is possible to avoid capturing the reflectance information of the target G (human body) when wearing glasses.

Perform the recognition step S170: recognize the recognized image P3 through a classifier (not shown in the figure); wherein, the classifier recognizes by the mean value of the reflectance in each detection area Y , To determine whether the target G is a human body. Specifically, the classifier in this embodiment is a support vector machine (Support Vector Machine; SVM), and the classifier is identified by the mean value of the reflectance in each detection area Y to Confirm whether some of the average reflectance values among the plurality of average reflectance values are greater than other average reflectance values for judgment.

In order to enable those skilled in the art to better understand the process of the present invention in the identification step S170, an example will be given below for description, and the target of this example is a fake face made of silicone (not shown in the figure). Show), the fake face simulates the real facial features, and the illuminating step S110, the capturing step S120, the position confirmation step S130, the image simulation step S140, and the image analysis step have been sequentially implemented S150, and the selection step S160; that is, the fake face will perform the recognition step S170.

When the skin surface is frontally illuminated by the image information capturing device C, except for the part corresponding to the forehead of the human body which is a smooth surface, other parts of the skin surface cannot directly face the image information capturing The device C has an angle so that when the infrared laser light L is irradiated on the other parts of the skin surface, it is not easy to be received, resulting in the low reflectivity of the other parts of the skin surface and close to The reflectivity of the fake face, therefore, when the infrared laser light L irradiates the skin surface, the infrared laser light L will have a strong reflectivity at the forehead position corresponding to the skin surface, and The light reflectance of other positions on the skin surface is lower, that is, the plurality of detection areas Y of the forehead corresponding to the skin surface have a higher mean light reflectance, which corresponds to the The plurality of detection regions Y on other positions of the skin surface have a lower mean value of the reflectance; in contrast, the reflectance of the fake face is very similar whether it is the forehead or other parts.

To put it another way, as shown in the graph in FIG. 8, this graph is the reflectivity when the infrared laser light L is irradiated on the fake face and the skin surface at different angles; wherein, the horizontal axis is the angle of incidence, and the vertical axis is the angle of incidence. The axis is reflectivity. From the chart, it can be clearly seen that when the infrared laser light L irradiates the skin surface at different incident angles, the corresponding reflectivity is significantly different, and the infrared laser light L irradiates the fake face at different incident angles. At this time, the corresponding light reflectance is less different than the former (that is, the light reflectance irradiated on the skin surface).

That is to say, when the infrared laser light L irradiates the fake face and the human body at an incident angle close to 0 degrees, the reflectivity of the skin surface will be significantly greater than that of the fake face, especially The forehead position is the most obvious; on the contrary, when the infrared laser light L irradiates the fake face and the human body at an incident angle close to 90 degrees, the reflectivity of the skin surface will obviously approach that of the fake face Reflectivity. Accordingly, through the difference in reflectance of the infrared laser light L on the surface of the silicone and the surface of the skin, the classifier can distinguish that the fake face is a non-real human body through the difference of a plurality of mean values of the reflectance.

In addition, it should be noted that the multiple detection areas Y are not located in the non-detection area N to avoid capturing the average value of the high reflectance of the spectacle mirror surface when the human body wears the spectacles, which may lead to misjudgment by the classifier. Case.

From the above description, it is clear that the infrared human body recognition method only needs to go through the irradiation step S110, the capturing step S120, the image simulation step S140, the image analysis step S150, and the identification step S170. The precise human body recognition effect can be achieved, and if the infrared human body recognition method implements the selection step S160 more, the human body recognition accuracy can be further greatly improved.

[Second Embodiment]

This embodiment is similar to the above-mentioned first embodiment, and the similarities between the two embodiments will not be described again. The main difference between this embodiment and the above-mentioned first embodiment lies in the following: Before performing the image simulation step S140, the infrared human body recognition method further includes performing a position confirmation step S130.

Perform the position confirmation step S130: analyze the basic image P1 to confirm the positions of a plurality of the first image information P11 in the basic image P1. Specifically, the basic image P1 includes red, blue, and green light information (RGB), and has a red channel (Channel R), a blue channel (Channel B), and a green channel (Channel G), The position of a plurality of the first image information P11 in the basic image P1 is confirmed through the difference in the reflection of the infrared laser light L by the red channel and the green channel, but it is not limited to this embodiment Contained. For example, the basic image P1 may also be the difference in reflection of the infrared laser light L by the red channel and the blue channel to confirm that a plurality of the first image information P11 are in the basic image Position in P1. Of course, the method of confirming the positions of the plurality of first image information in the basic image P1 can also be to obtain the plurality of first image information P11 through the characteristic shape or high-frequency position of the basic image P1. The position in the basic image P1.

[Third Embodiment]

As shown in FIG. 9, it is the third embodiment of the present invention. This embodiment is similar to the above-mentioned second embodiment, and the similarities between the two embodiments will not be repeated. The main difference of the embodiment is that in the implementation of the irradiation step S110, a plurality of the laser blocks LG are in a square shape, and the number of the plurality of laser blocks LG is greater than 50, so that they are located at any two The ambient light area EA is formed between the laser tiles LG. However, it should be noted that the number and shape of the multiple laser blocks LG can be adjusted according to the designer's needs. For example, the multiple laser blocks LG can also be triangular, diamond-shaped, or Irregular shapes, etc.

[Technical Effects of Embodiments of the Invention]

In summary, the infrared human body recognition method disclosed in the embodiment of the present invention can generate a plurality of the laser blocks LG and a plurality of the environment spaced apart on the target G by the infrared laser light L With the light area EA, the basic image P1 including the infrared laser light L and the ambient light E image information can be obtained in one shot; accordingly, by directly analyzing all the basic images P1 The first image information P11 (that is, the image information corresponding to a plurality of the laser blocks LG) and the second image information P12 (that is, the image information corresponding to the plurality of ambient light areas EA), to obtain The high-quality recognition image P3 allows the classifier to directly recognize the recognition image P3 to maintain an accurate recognition rate.

The content disclosed above is only the preferred and feasible embodiments of the present invention, and does not limit the scope of the patent application of the present invention. Therefore, all equivalent technical changes made using the description and schematic content of the present invention are included in the application of the present invention. Within the scope of the patent.

G: Goal L: Infrared laser light LG: Laser tiles E: Ambient light EA: Ambient light area P1: Basic image P11: First image information P12: Second image information P2: Analog image A: Simulation information P3: Recognize the image P31: Recognition area D1: First direction D2: second direction Y: detection area N: Do not detect area C: Image information capture device

FIG. 1 is a schematic diagram of a state in which the infrared human body recognition method according to the first embodiment of the present invention is performing the irradiation step.

FIG. 2 is a reference chart of the reflectance of the infrared laser light used in the infrared human body recognition method according to the first embodiment of the present invention when irradiating the silicone surface and the skin surface with different wavelength bands.

3 is a schematic plan view of a basic image of the infrared human body recognition method according to the first embodiment of the present invention.

4 is a schematic plan view of a simulated image of the infrared human body recognition method according to the first embodiment of the present invention.

FIG. 5 is a schematic plan view of a recognition image of the infrared human body recognition method according to the first embodiment of the present invention.

FIG. 6 is a schematic plan view of the first embodiment of the present invention when multiple recognition images of the infrared human body recognition method are divided into multiple recognition areas.

7 is a schematic plan view of undetected areas in multiple recognition images of the infrared human body recognition method according to the first embodiment of the present invention.

FIG. 8 is a reference chart of the reflectance of the infrared laser light in each waveband irradiated on the silicone surface and the skin surface in the infrared human body recognition method according to the first embodiment of the present invention.

FIG. 9 is a schematic diagram of a state in which the infrared human body recognition method according to the third embodiment of the present invention is performing the irradiation step.

G: Goal

L: Infrared laser light

LG: Laser tiles

E: Ambient light

EA: Ambient light area

C: Image information capture device

Claims (10)

An infrared human body recognition method, comprising: implementing an irradiation step: irradiating a target with an infrared laser light so that the outer surface of the target forms a plurality of laser blocks arranged at intervals, and the target An ambient light area irradiated by an ambient light is formed on the outer surface between any two of the laser blocks; implementing a capturing step: capturing on the outer surface of the target through an image information capturing device A basic image, the basic image having a plurality of first image information respectively corresponding to a plurality of the laser blocks and a plurality of second image information respectively corresponding to a plurality of the ambient light regions; implementing an image simulation step: the ambient light in the base region of the image corresponding to any two adjacent second image information obtained information to a an analog interpolation, the two analog information corresponding to the ambient light substituted The first image information of the laser tiles between regions to generate a simulated image that simulates that the target is only illuminated by the ambient light; and an image analysis step is performed: the basic image and the The simulated image is compared and analyzed to remove the image information of the ambient light to generate an identification image, the identification image having a plurality of reflectance information corresponding to a plurality of the laser blocks on the target; implement one The identification step: the identification image is identified by a classifier; wherein the classifier is identified by a plurality of the reflectance information in the identification image to confirm whether the target is a human body. The infrared human body identification method according to claim 1, wherein the wavelength of the infrared laser light is not less than 800 nanometers (nm), and the reflectance of the infrared laser light on the skin surface irradiating the human body is greater than that of the The reflectivity of infrared laser light irradiating non-skin surface is greater than 5%. According to the infrared human body identification method according to claim 1, the wavelength of the infrared laser light is 840 to 860 nanometers (nm), and the reflectivity of the infrared laser light on the skin surface irradiating the human body is greater than that of the infrared ray The laser light irradiates 15% of the reflectivity of the non-skin surface. The infrared human body identification method according to claim 1, wherein a plurality of the laser tiles are striped, so that the ambient light area between any two laser tiles is striped. The infrared human body identification method according to claim 1, wherein the classifier is a support vector machine (SVM). The infrared human body identification method according to claim 1, wherein the number of the plurality of laser blocks is not less than six. The infrared human body recognition method according to claim 1, wherein, between performing the capturing step and performing the image simulation step, the infrared human body recognition method further includes performing a position confirmation step: analyzing the basis Image to confirm the position of a plurality of the first image information in the basic image. The infrared human body recognition method according to claim 7, wherein in the step of performing the position confirmation, the basic image includes red, blue, and green light information (RGB) and has a red channel (Channel R) and a The green channel (Channel G) is used to confirm the position of the plurality of first image information in the basic image through the difference in the reflection of the infrared laser light between the red channel and the green channel. An infrared human body identification method, which includes: Implement an irradiation step: irradiate a target with an infrared laser light, so that the outer surface of the target forms a plurality of laser blocks arranged at intervals, and the outer surface of the target is on any two lasers. An ambient light area illuminated by an ambient light is formed between the blocks; Implementing a capturing step: capturing a basic image on the outer surface of the target through an image information capturing device, the basic image having a plurality of first image information corresponding to a plurality of the laser blocks, respectively And a plurality of second image information respectively corresponding to a plurality of said ambient light regions; Performing an image analysis step: analyzing a plurality of the first image information and a plurality of the second image information in the basic image to generate an identification image, the identification image having a plurality of the laser blocks Multiple reflectance information on the target; A recognition step is performed: the recognition image is recognized by a classifier; wherein the classifier recognizes through a plurality of the reflectance information in the recognition image to confirm whether the target is a human body. The infrared human body identification method according to claim 9, wherein the wavelength of the infrared laser light is 840 to 860 nanometers (nm), and the reflectivity of the infrared laser light on the skin surface irradiating the human body is greater than that of the infrared laser light. The reflectivity of the infrared laser light when irradiating the non-skin surface is 15%.

Images