CN103577797B - Image recognition system and image recognition method thereof - Google Patents

Abstract

The invention discloses an image identification system and an image identification method thereof. The image recognition system comprises a setting module and a recognition module. The setting module divides an image setting area into a plurality of search areas and selectively sets a search interval for each search area respectively. The identification module is connected with the setting module, and when the identification module identifies the image in a first identification mode, a search window moves in each search area according to the corresponding search interval so as to detect and identify the pixel point corresponding to the position to which the search window moves. The invention can effectively improve the efficiency of the whole image identification and can maintain the quality of the identification and detection.

Description

Image recognition system and image recognition method thereof

technical field

The present invention relates to an image recognition system and its image recognition method, in particular to a fast and accurate image recognition system and its image recognition method which can perform image recognition with a specific moving mode or a mask .

Background technique

With the development of science and technology, image processing technology has become more and more diversified, and its uses have become more and more extensive. The image processing software/system can be installed or embedded in devices such as personal computers, mobile phones, or cameras, or self-assembled into an image processing device. The image processing technology includes an image recognition function, which can be used to detect specific features in an image photo for identification and comparison with other reference images. For example: it can be used to compare the facial features of people, and then perform identity checks, etc.

At present, in terms of existing image recognition technologies, when performing image recognition operations, most of them use a single or fixed method to sequentially move a search window to search for each image unit (including 22*22 pixels) in the image. The identification process, such an inflexible operation method, will increase the amount of data to be processed, and at the same time, the time required for access and execution will be relatively lengthy. As shown in FIG. 1 , the image 100 can be divided into several image units 101 , and when performing image recognition, the search window 200 is moved to perform image recognition on each image unit 101 . Wherein, this figure is only a simplified notation, and actually adjacent image units 101 partially overlap each other. For example, assuming that the pixel position of the upper left corner of the first image unit 101 on the upper left is the coordinate (0,0) in the image 100, the pixel position of the upper left corner of the adjacent right image unit 101 is the coordinate (1,0). , the pixel position of the upper left corner of the image unit 101 below is the coordinate (0,1). The search window 200 uses one pixel as its moving distance. For example, the first point performs identification processing on the image unit 101 at the pixel position (0,0) in the upper left corner. After the execution, it moves one pixel to the position in the upper left corner ( 1, 0) image unit 101 for identification.

In addition, in order to further speed up the image recognition processing, some image recognition methods allow the search window 200 to move at a search interval 300 to identify the moved image unit 101 , as shown in FIG. 2 . However, this method can only set a single search interval 300 for the entire image 100, allowing the system to use this search interval 300 to detect and identify the image 100, but it is not possible to set different search intervals for different regions of the image 100 300, the processing flexibility is relatively insufficient, and it may cause the problem of inaccurate image recognition.

Contents of the invention

In view of the above-mentioned known technical problems, the purpose of the present invention is to provide an image recognition system and its image recognition method to solve the processing efficiency caused by the known image recognition technology that can only move the search window sequentially in a single or fixed mode for recognition. Slow, and only a single search interval can be set for the full image, resulting in insufficient processing flexibility.

To achieve the above object, the present invention adopts the following technical solutions:

An image recognition system includes a setting module and a recognition module. The setting module divides an image setting into several search areas, and selectively sets a search interval for each search area. The identification module is connected to the setting module. When the identification module identifies the image in a first identification mode, a search window is moved in each search area according to the corresponding search interval, so as to determine the location where the search window moves to. Corresponding pixel points are detected and identified. Wherein, the setting module also selectively divides each search area into a plurality of search blocks, and each search block includes a plurality of search units. Setting the module and setting a corresponding search mask for each search box, so that when the identification module recognizes the image in a second identification mode, each search mask is used to respectively search for each corresponding search box Several search units selectively detect and identify.

Preferably, the setting module can provide the user with the flexibility to separately set the range size of each search area and each corresponding interval of the search interval.

Preferably, the setting module also sets each search mask to include at least one search window, so that when the recognition module uses each search mask to perform image detection on each corresponding search box, only the corresponding search window in each search box Perform detection and recognition on the pixels of the search unit in the search window.

Preferably, the setting module can allow the user to flexibly set the range size of each search box respectively, so as to further determine the number of search units corresponding to each search box, and provide the user to flexibly set the range of each search box respectively. The number of search windows included in the mask.

Preferably, the setting module can provide users with options to set the identification module to perform identification in the first identification mode or the second identification mode.

According to the purpose of the present invention, an image recognition method is proposed, which includes the following steps: using a setting module to divide an image setting into several search areas, and setting a search interval for each search area; and when When a recognition module recognizes an image, it moves a search window in each search area according to the corresponding search interval, so as to detect and recognize the pixel corresponding to the position where the search window moves. Wherein, the user is also allowed to respectively set the horizontal search interval and the vertical search interval in each of the search areas through the setting module.

According to the purpose of the present invention, an image recognition method is proposed, which includes the following steps: using a setting module to divide an image setting into several search areas, and setting each search area into several search squares , each search box consists of several search units; use the setting module to set a corresponding search mask for each search box; and when an identification module identifies an image, it uses each search mask to respectively A plurality of search units in each corresponding search block are selectively detected and identified.

Based on the above, according to the image recognition system and its image recognition method of the present invention, it may have one or more of the following advantages:

(1) This image recognition system and its image recognition method can provide users with the ability to divide the image setting into several search areas, and then set each search area to a different search interval, so as to use this search interval for image recognition The processing can avoid wasting time searching for unimportant areas, and mainly identify and detect areas that may have features. In this way, the efficiency of overall image recognition can be effectively improved, and the quality of recognition and detection can be maintained. .

(2) This image recognition system and its image recognition method can provide users with setting different recognition modes, so as to further set the image into different image areas, blocks, units, etc., and can use the recognition and detection processing more effectively resources to effectively improve efficiency.

Description of drawings

FIG. 1 is a first schematic diagram of a known image recognition method of the present invention.

FIG. 2 is a second schematic diagram of a known image recognition method of the present invention.

FIG. 3 is a block diagram of the image recognition system of the present invention.

FIG. 4 is a first flowchart of the image recognition method of the present invention.

FIG. 5 is a second flowchart of the image recognition method of the present invention.

FIG. 6 is a schematic diagram of a first embodiment of the image recognition system and its image recognition method of the present invention.

FIG. 7 is a schematic diagram of a second embodiment of the image recognition system and its image recognition method of the present invention.

FIG. 8 is a schematic diagram of another embodiment of the image recognition system and its image recognition method of the present invention.

FIG. 9 is a schematic diagram of another embodiment of the image recognition system and its image recognition method of the present invention.

detailed description

In order to illustrate the technical features, content and advantages of the present invention and the effects that can be achieved, the present invention will be described in detail below in conjunction with the accompanying drawings, and in the form of embodiments, and the accompanying drawings used herein are only intended to The purpose of illustration and auxiliary description is not necessarily the true proportion and precise configuration of the present invention after implementation, so it should not be interpreted or limited to the scope of rights of the present invention in actual implementation based on the proportion and configuration relationship of the drawings, and it should be stated in advance.

Please refer to FIG. 3 , which is a block diagram of the image recognition system of the present invention. In the figure, the image recognition system 1 includes a setting module 11 and a recognition module 12 , and the setting module 11 is connected to the recognition module 12 . The image recognition system 1 can be software or circuit chips, etc., to be installed or embedded in devices such as personal computers, mobile phones, or cameras, or self-assembled as an image processing device. The setting module 11 may include physical keys or virtual keys and corresponding processing circuits.

In the above, when the user wants to detect and identify an image 2, he can use the setting module 11 to divide the image 2 into several search areas 21, and selectively set a search interval for each search area 21 3. After the user completes the settings in this way, the identification module 12 can identify the image 2 in the first identification mode 121, so that a search window can be searched in each search area 21 according to the search interval 3 set by the user. Move to detect and identify the pixels of image 2 corresponding to the position to which the search window is moved. Among them, the setting module 11 can provide the user to set the range size of each search area 21 and the interval of each corresponding search interval 3 respectively, and the setting module 11 can provide the user to set the range of each search area 21 respectively. Horizontal search interval 3 and vertical search interval 3.

In addition, the image recognition system 1 can also detect and recognize the image 2 in another form. The user can use the setting module 11 to set and divide the image 2 into several search areas 21, and then divide each search area 21 into several search blocks 211, and make each search block 211 contain several search blocks 211. Unit 2111. The user can set a corresponding search mask 4 for each search box 211 through the setting module 11 . Then, the identification module 12 can detect and identify the image 2 in a second identification mode 122, so as to use each search mask 4 to selectively detect and identify several search units 2111 in each corresponding search box 211. . In more detail, the setting module 11 can provide the user to set at least one search window for each search mask 4, so that in the recognition module 12, each search mask 4 can be used for each corresponding search box 211 When performing image detection, the recognition module 12 only detects and recognizes the pixels in each search block 211 corresponding to the search unit 2111 of the search window in the search mask 4 . Wherein, the setting module 11 can provide the user to flexibly set the range size of each search box 211 respectively, so that the number of search units 2111 corresponding to each search box 211 can be further determined. And the setting module 11 can also provide the user with the flexibility to respectively set the number of search windows included in each search mask 4 .

Among the above, the setting module 11 can provide the user to set the area division of the image 2 first and store it. At that time, when the user wants to use the image recognition function, the user only needs to select and set the recognition module through the setting module 11 12 is to perform identification in the first identification mode 121 or the second identification mode 122 , and the identification module 12 will perform image identification according to the image 2 setting mode preset by the user.

In the above, the method that the setting module 11 provides for the user to distinguish and set the image 2 is only an example. The setting module 11 can also provide the user to directly set the image 2 and divide it into several search units 2111, and then set the identification The module 12 performs image detection and recognition on part or all of the search units 2111 in the image 2 . All the above should not be limited by the examples of the present invention.

Please refer to FIG. 4, which is the first flow chart of the image recognition method of the present invention, and its process steps are:

S41: Use a setting module to divide an image setting into several search areas, and set a search interval for each search area.

S42: When a recognition module recognizes the image, move a search window in each search area according to the corresponding search interval, so as to detect and recognize the pixel corresponding to the position where the search window moves.

Please refer to FIG. 5, which is a second flow chart of the image recognition method of the present invention, and its process steps are:

S51: Use a setting module to divide an image setting into several search areas, and set and divide each search area into several search blocks, and each search block respectively includes several search units.

S52: Use the setting module to set a corresponding search mask for each search box.

S53: When a recognition module recognizes the image, use each search mask to selectively detect and recognize several search units in each corresponding search block.

Please refer to FIG. 6 , which is a schematic diagram of a first embodiment of the image recognition system and its image recognition method of the present invention. In this embodiment, there is an image A, and the user can set the image A through the image recognition system of the present invention, and can selectively set the image A as the first search area A1 and the second search area A2, the third search area A3 and the fourth search area A4. Wherein, the above setting of the number of search areas and the size of the range may not be limited, and this figure is only for illustration. The user can further set a corresponding search interval C1 , C2 , C3 and C4 for the first search area A1 , the second search area A2 , the third search area A3 and the fourth search area A4 respectively. Then, the image recognition system will place a search window B in the first search area A1, the second search area A2, the third search area A3, and the fourth search area A4 according to the settings of the user. The search intervals C1, C2, C3 and C4 are moved, and the pixel point to which the search window B is moved is detected and identified. As shown in FIG. 6 , the search interval C1 corresponding to the first search area A1 is 2 pixel positions in the horizontal and vertical directions respectively, and the search interval C2 corresponding to the second search area A2 is 3 pixel positions in the horizontal and vertical directions respectively, The search interval C3 corresponding to the third search area A3 is 1 pixel position apart horizontally and vertically, and the search interval C4 corresponding to the fourth search area A4 is 2 pixel positions apart horizontally and vertically. Wherein, the setting of the search intervals C1 , C2 , C3 and C4 is just an example, and is not limited thereto.

Please refer to FIG. 7 , which is a schematic diagram of a second embodiment of the image recognition system and its image recognition method of the present invention. In this embodiment, there is an image D, and the user can set the image D through the image recognition system of the present invention, and can selectively set the image D as the first search area D1 and the second search area D2, the third search area D3 and the fourth search area D4. The user can also divide each search area into several search blocks. Taking the first search area D1 as an example, the user can set the division as the first search block D11, the second search block D12, and the third search block D13 and the fourth search block D14. Each search block further includes several search units, taking the first search block D11 as an example, it includes nine search units D111. Then, the user can further set a corresponding search mask for each search box, take the first search box D11 as an example, set a search mask E for it, and selectively select At least one search window is set, here, the search windows E1, E2, E3 and E4 are set. Wherein, the setting of the position of each search window may not be limited thereto, and this is only an embodiment. After setting the corresponding search masks for each search box, the image recognition system will use the set search masks to selectively select the search units in each corresponding search box Detect and identify. Taking the first search block D11 as an example, the first search block D11 is divided into nine search units D111, and its corresponding search mask E is also divided into nine blocks, some of which are set with search windows E1, E2, E3 and E4, however, the image recognition system then performs image detection and recognition on the first search block D11 according to the search mask E, so as to identify whether the search unit D111 corresponding to the positions of the search windows E1, E2, E3 and E4 is Has a characteristic presence (characteristics, etc.).

Among the above, the image recognition system and its image recognition method of the present invention, by setting the image, and then detect and recognize the image with different search intervals or search masks, mainly to improve the efficiency of image recognition, and at the same time The accuracy of recognition is not lost. Allows users to set a higher search density for positions where features are more likely to appear in the image, and for positions where features are less likely to appear, the search density can be set to decrease, which can effectively improve the efficiency of image recognition , as shown in Figure 8. In the figure, if the square in the image has a dark background with oblique lines, it is the identification point 500 to be detected (which may be the search window described in the present invention).

In addition, when determining the search mask for each search box, it can be determined according to the following two methods, but this is only an implementation example and is not limited thereto.

method one:

Using software settings, the software can select the applicable search mask according to various needs and different applications. The flexibility of this method is large, what kind of search mask you want to use, you can use the software judgment method to prepare the search mask to be used in advance. When the hardware is executed, image processing will be performed according to the prepared search mask, but this method consumes CPU resources of the central processing unit, and a more suitable search mask can be determined according to different needs.

Method 2:

The hardware determines which pixels in each 3x3 search mask correspond to the search unit (the search unit is, for example, a 22x22 window, but not limited thereto) to be processed according to the weight (priority). For example, if we set the 5th, 1st and 9th pixel points as high weight (that is, the search window set according to the present invention), when the search unit corresponding to these three points contains face information or it is important information If there are existing images, the search units corresponding to the remaining points are then processed, as shown in FIG. 9 . If the search units corresponding to these three points (5, 1, and 9) have no face information or are not important, then other points corresponding to the 3x3 search mask (for example, 4, 6, 2, 3, 8 and 7 pixels). Of course, we can also set only one pixel, the fifth pixel, as a high weight, and the other pixels will be judged according to their respective weights, and they will be processed when it is judged that there is important information. In this way, the hardware can automatically process the image according to a rule, and the CPU does not need to consume resources to make judgments.

To sum up the above, the image recognition system and its image recognition method of the present invention can provide users with settings for images, and further set the required recognition mode, and can detect images at different search intervals or search masks Recognition, in this way, can effectively improve the efficiency of image recognition, and can also take into account the quality of detection and recognition at the same time.

The above descriptions are illustrative only, not restrictive. Any equivalent modifications or changes made without departing from the spirit and scope of the present invention shall be included in the scope of the appended claims.

Claims (9)

1. An image recognition system, characterized in that it comprises: a setting module, which divides the image setting into several search areas, and selectively sets the search interval for each of the search areas; and The identification module is connected to the setting module. When the identification module identifies the image in the first identification mode, the search window is moved in each search area according to the corresponding search interval, so that the search window The pixels corresponding to the moved position are detected and identified; Wherein, the setting module also selectively divides each of the search areas into several search boxes, and each of the search boxes contains a number of search units, and the setting module sets each of the search boxes respectively. Determining a corresponding search mask, so that when the recognition module recognizes the image in the second recognition mode, each search mask is used to selectively select the search units in each corresponding search box Detect and identify. 2 . The image recognition system according to claim 1 , wherein the setting module allows the user to dynamically set the range size of each of the search areas and the corresponding interval of the search interval. 3. The image recognition system according to claim 1, wherein the setting module further sets each of the search masks to include at least one search window, so that the recognition module uses each of the search masks to When each corresponding search block performs image detection, only the pixel points corresponding to the search unit of the search window in each search block are detected and identified. 4. The image recognition system according to claim 3, wherein the setting module allows the user to dynamically set the range size of each of the search boxes, so as to further determine the search corresponding to each of the search boxes. The number of units, and provide the user with the flexibility to separately set the number of the search windows included in each of the search masks. 5 . The image recognition system according to claim 1 , wherein the setting module provides users with options to set the recognition module to perform recognition in the first recognition mode or in the second recognition mode. 6 . 6. An image recognition method, characterized in that it comprises the following steps: Use the setting module to divide the image setting into several search areas, and set the search interval for each search area; and When the identification module identifies the image, the search window is moved in each of the search areas according to the corresponding search interval, so as to detect and identify the pixel corresponding to the position to which the search window moves; Wherein, the user is also allowed to respectively set the horizontal search interval and the vertical search interval in each of the search areas through the setting module. 7. An image recognition method, characterized in that it comprises the following steps: Use the setting module to divide the image setting into several search areas, and divide each of the search areas into several search boxes, and each of the search boxes includes several search units; using the setting module to set a corresponding search mask for each of the search boxes; and When the identification module identifies the image, each of the search masks is used to selectively detect and identify the search units in the corresponding search blocks. 8. The image recognition method according to claim 7, further comprising the following steps: Use the setting module to set at least one search window for each of the search masks, so that when the identification module uses each of the search masks to perform image detection on each of the corresponding search boxes, only for each of the search boxes The pixel points corresponding to the search unit in the search window are detected and identified. 9. The image recognition method according to claim 8, further comprising the following steps: The setting module is provided to allow users to dynamically set the range size of each of the search boxes, so as to further determine the number of the search units corresponding to each of the search boxes, and provide the user to actively set each of the search boxes The number of search windows to include in the search mask.