Image recognition system and method for self-dividing regions

Abstract

The present invention discloses an image recognition system and method for self-dividing regions, relates to the technical field of image recognition, and solves the technical problems of region division in the prior art, which is time-consuming and inaccurate. The present invention obtains an original image of a target region through a camera device, and sends the original image to an image processing module, which divides the original image into regions and obtains S sub-region images. The original image is divided into regions, and the accuracy of the region division is improved. The sub-region image label is obtained according to the sub-region image and an image recognition model. The sub-region image and the sub-region image label are sent to the region division module. The region division module divides the original image into regions according to the sub-region image and the sub-region image label. A label can be obtained for each sub-region, so that users can understand the sub-region features.

Description

Technical Field

[0001] The present invention belongs to the field of region division, relates to image recognition technology, and specifically is an image recognition system and method for region self-division. Background Art

[0002] In the study and application of images, people are often only interested in certain parts of an image. These parts of interest generally correspond to specific areas with special properties in the image (can correspond to a single area or multiple areas), which are called targets or foregrounds; while other parts are called the background of the image. In order to identify and analyze the target, it is necessary to isolate the target from an image, which is the problem to be studied in image segmentation. The so-called image segmentation, in a broad sense, is to group and cluster image pixels according to the similarity criteria of certain features or feature sets of the image (including grayscale, color, texture, etc.), and divide the image plane into several non-overlapping areas with certain consistency. This makes the pixel features in the same area similar, that is, consistent; while the features of pixels in different areas have mutations, that is, inconsistency.

[0003] Existing technical solutions generally use manual calibration to distinguish different areas in an image. Automatic methods are generally classic algorithms such as GraphCut, which interact with users to analyze the statistical characteristics of the image in the user-selected area, and then automatically distinguish the foreground and background of the image.

[0004] Existing technologies, such as the method of manually calibrating areas, are obviously time-consuming and inaccurate. Classic algorithms such as Graphcut require users to select objects and backgrounds when processing each image, which is time-consuming and computationally complex. It takes even more time on platforms with poor computing performance, such as mobile phones.

[0005] Therefore, an image recognition system and method with self-region division are proposed. Summary of the invention

[0006] The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, the present invention proposes an image recognition system and method for self-regional division, which solves the problem of region division in the prior art, which is time-consuming and inaccurate.

[0007] To achieve the above-mentioned object, an image recognition system with self-region division is provided according to an embodiment of the first aspect of the present invention, comprising an image acquisition module, an image processing module and a region division module; each module performs information exchange based on a data signal;

[0008] The image acquisition module is used to acquire the original image of the target area and send the original image to the image processing module and the area division module;

[0009] The image processing module is used to receive the original image, and divide the original image into regions to obtain S sub-region images;

[0010] Acquire a sub-region image label according to the sub-region image and an image recognition model; wherein the image recognition model is established based on an artificial intelligence model;

[0011] and sending the sub-region image and the sub-region image label to the region division module;

[0012] The region division module is used to receive the original image, the sub-region image and the sub-region image label, and divide the original image into regions according to the sub-region image and the sub-region image label.

[0013] Preferably, the image acquisition module includes a camera device.

[0014] Preferably, the camera device acquires an original image of the target area and sends the original image to the image processing module.

[0015] Preferably, the image processing module receives the original image and divides the original image into regions to obtain S sub-region images. The specific process includes:

[0016] The image processing module receives the original image;

[0017] Divide the original image into M rows×N columns to obtain M×N standard images;

[0018] Select any standard image and mark it as a first seed standard image, where the first seed standard image is marked as 1;

[0019] Compare the first seed standard image with the adjacent standard images by category,

[0020] The adjacent standard image and the first seed standard image belong to the same category, and the adjacent standard image is marked as 1;

[0021] The adjacent standard image and the first seed standard image do not belong to the same category, and the adjacent standard image is marked as 0;

[0022] After the adjacent standard images of the first seed standard image are marked,

[0023] If all adjacent standard images are marked as 1, the range of adjacent standard images is expanded until all standard images are marked;

[0024] If the labels of adjacent standard images are 0 and 1,

[0025] Obtain an adjacent standard image of a standard image marked as 1, and the adjacent standard image is not marked,

[0026] Comparing the standard image with the first seed standard image by category, and repeating the above process until all standard images are marked;

[0027] Acquire a standard image marked as 1, and integrate the standard image into a first sub-region image;

[0028] Select any standard image marked as 0 and set it as the second seed standard image. The second seed standard image is marked as 2.

[0029] Compare the second seed standard image with the adjacent standard image by category,

[0030] The adjacent standard image and the second seed standard image belong to the same category, and the adjacent standard image is marked as 2;

[0031] The adjacent standard image and the second seed standard image do not belong to the same category, and the adjacent standard image is marked as 0;

[0032] After the adjacent standard images of the second seed standard image are marked,

[0033] If all adjacent standard images are marked as 2, the range of adjacent standard images is expanded until all standard images are marked;

[0034] If the labels of the adjacent standard images are 0 and 2,

[0035] Obtain an adjacent standard image of a standard image marked as 2, and the adjacent standard image is not marked as 2,

[0036] Comparing the adjacent standard images with the second seed standard images by category, and repeating the above process until all standard images are marked;

[0037] Acquire a standard image marked as 2, and integrate the standard image into a second sub-region image;

[0038] Repeat the above process to obtain S sub-region images.

[0039] Preferably, the image segmentation model is established based on an artificial intelligence model, and the specific process includes:

[0040] Obtain standard training images from the image processing module;

[0041] The artificial intelligence model is trained using standard training images, and the trained artificial intelligence model is marked as an image segmentation model;

[0042] The standard training images include several groups of input images and corresponding region labels, and the content attributes of the input images and the sub-region images are consistent;

[0043] The artificial intelligence model includes a deep convolutional neural network model or a RBF neural network model.

[0044] Preferably, the region division module receives the original image, the sub-region image and the sub-region image label, and divides the original image into regions according to the sub-region image and the sub-region image label. The specific process includes:

[0045] The region division module receives the sub-region image and the sub-region image label;

[0046] The region division module merges the original image, the sub-region image and the sub-region image label;

[0047] Obtaining the original image with region differentiation;

[0048] Different sub-regions on the original image are marked with different colors.

[0049] Preferably, the image acquisition module is in communication and / or electrical connection with the image processing module;

[0050] The image acquisition module is in communication and / or electrically connected with the area division module;

[0051] The image processing module is in communication and / or electrically connected with the area division module.

[0052] A method for image recognition by self-dividing regions, the specific method comprising the following steps:

[0053] Step 1: The camera device obtains the original image of the target area and sends the original image to the image processing module;

[0054] Step 2: The image processing module receives the original image and divides the original image into regions to obtain S sub-region images;

[0055] Step 3: Obtain sub-region image labels according to the sub-region images and the image recognition model; and send the sub-region images and sub-region image labels to the region division module;

[0056] Step 4: The region division module receives the original image, the sub-region image and the sub-region image label, and divides the original image into regions according to the sub-region image and the sub-region image label.

[0057] Compared with the prior art, the present invention has the following beneficial effects:

[0058] The present invention obtains the original image of the target area through a camera device, and sends the original image to an image processing module, which divides the original image into regions and obtains S sub-region images; thus, the region division of the original image is realized, and the accuracy of the region division is improved;

[0059] The sub-region image label is obtained according to the sub-region image and the image recognition model; and the sub-region image and the sub-region image label are sent to the region division module; the region division module divides the original image into regions according to the sub-region image and the sub-region image label; and a label can be obtained for each sub-region, so that the user can understand the sub-region characteristics. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 It is the principle diagram of the present invention;

[0061] Figure 2 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0062] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0063] like Figure 1 As shown, an image recognition system with self-regional division includes an image acquisition module, an image processing module and a region division module; each module exchanges information based on data signals;

[0064] The image acquisition module is used to acquire the original image of the target area and send the original image to the image processing module and the area division module;

[0065] The image processing module is used to receive the original image, and divide the original image into regions to obtain S sub-region images;

[0066] Acquire a sub-region image label according to the sub-region image and an image recognition model; wherein the image recognition model is established based on an artificial intelligence model;

[0067] and sending the sub-region image and the sub-region image label to the region division module;

[0068] The region division module is used to receive the original image, the sub-region image and the sub-region image label, and divide the original image into regions according to the sub-region image and the sub-region image label.

[0069] The image acquisition module includes a camera device; wherein the camera device includes an intelligent photographing device such as a camera;

[0070] The camera device acquires an original image of a target area and sends the original image to the image processing module.

[0071] The image processing module receives the original image and divides the original image into regions to obtain S sub-region images. The specific process includes:

[0072] The image processing module receives the original image;

[0073] Divide the original image into M rows×N columns to obtain M×N standard images;

[0074] Select any standard image and mark it as the first seed standard image, which is marked as 1. It should be noted that the seed standard image usually does not select the standard image at the edge, but selects the standard image at the middle. The reason is that if the seed standard image is at the edge of the image, there are fewer adjacent standard images. For example, if the standard image at the vertex is selected as the seed standard image, the number of adjacent standard images of the seed standard image is 3, if the standard image at the side of the image is selected as the seed standard image, the number of adjacent standard images of the seed standard image is 5, and if the standard image at the middle of the image is selected as the seed standard image, the number of adjacent standard images of the seed standard image is 8. Therefore, try not to select the standard image at the edge as the seed standard image.

[0075] Compare the first seed standard image with the adjacent standard images by category,

[0076] The adjacent standard image and the first seed standard image belong to the same category, and the adjacent standard image is marked as 1;

[0077] The adjacent standard image and the first seed standard image do not belong to the same category, and the adjacent standard image is marked as 0;

[0078] After the adjacent standard images of the first seed standard image are marked,

[0079] If all adjacent standard images are marked as 1, the range of adjacent standard images is expanded until all standard images are marked;

[0080] If the labels of adjacent standard images are 0 and 1,

[0081] Obtain an adjacent standard image of a standard image marked as 1, and the adjacent standard image is not marked,

[0082] Comparing the standard image with the first seed standard image by category, and repeating the above process until all standard images are marked;

[0083] Acquire a standard image marked as 1, and integrate the standard image into a first sub-region image;

[0084] Select any standard image marked as 0 and set it as the second seed standard image. The second seed standard image is marked as 2. It should be noted that the seed standard image should be selected as the standard image in the middle position as much as possible;

[0085] Compare the second seed standard image with the adjacent standard image by category,

[0086] The adjacent standard image and the second seed standard image belong to the same category, and the adjacent standard image is marked as 2;

[0087] The adjacent standard image and the second seed standard image do not belong to the same category, and the adjacent standard image is marked as 0;

[0088] After the adjacent standard images of the second seed standard image are marked,

[0089] If all adjacent standard images are marked as 2, the range of adjacent standard images is expanded until all standard images are marked;

[0090] If the labels of the adjacent standard images are 0 and 2,

[0091] Obtain an adjacent standard image of a standard image marked as 2, and the adjacent standard image is not marked as 2,

[0092] Comparing the adjacent standard images with the second seed standard images by category, and repeating the above process until all standard images are marked;

[0093] Acquire a standard image marked as 2, and integrate the standard image into a second sub-region image;

[0094] Repeat the above process to obtain S sub-region images.

[0095] In this embodiment, the image segmentation model is established based on an artificial intelligence model, and the specific process includes:

[0096] Obtain standard training images from the image processing module;

[0097] The artificial intelligence model is trained using standard training images, and the trained artificial intelligence model is marked as an image segmentation model.

[0098] In this embodiment, the standard training image includes several groups of input images and corresponding region labels, and the content attributes of the input images and the sub-region images are consistent;

[0099] In this embodiment, the artificial intelligence model includes a deep convolutional neural network model or an RBF neural network model and other models with powerful nonlinear fitting capabilities.

[0100] The region division module receives the original image, the sub-region image and the sub-region image label, and divides the original image into regions according to the sub-region image and the sub-region image label. The specific process includes:

[0101] The region division module receives the sub-region image and the sub-region image label;

[0102] The region division module merges the original image, the sub-region image and the sub-region image label;

[0103] Obtaining the original image with region differentiation;

[0104] Different sub-regions on the original image are marked with different colors.

[0105] It should be further explained that marking different areas with different colors can help users distinguish different areas in the image more clearly.

[0106] For example:

[0107] The region division module receives four sub-region images and corresponding sub-region image labels;

[0108] The region division module divides the original image according to the sub-region image labels and the sub-region images;

[0109] After the original image is divided, 4 different sub-areas are displayed;

[0110] The first sub-region is marked in red;

[0111] The second sub-region is marked in yellow;

[0112] The third sub-region is marked in blue;

[0113] The fourth sub-region is marked in green.

[0114] In this embodiment, the image acquisition module is in communication and / or electrically connected with the image processing module;

[0115] The image acquisition module is in communication and / or electrically connected with the area division module;

[0116] The image processing module is in communication and / or electrically connected with the region division module.

[0117] like Figure 2 As shown, a method for image recognition by self-dividing regions comprises the following steps:

[0118] Step 1: The camera device obtains the original image of the target area and sends the original image to the image processing module;

[0119] Step 2: The image processing module receives the original image and divides the original image into regions to obtain S sub-region images;

[0120] Step 3: Obtain sub-region image labels according to the sub-region images and the image recognition model; and send the sub-region images and sub-region image labels to the region division module;

[0121] Step 4: The region division module receives the original image, the sub-region image and the sub-region image label, and divides the original image into regions according to the sub-region image and the sub-region image label.

[0122] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. An image recognition system with self-regional division, characterized in that: It includes image acquisition module, image processing module and area division module; each module exchanges information based on data signals; The image acquisition module is used to acquire the original image of the target area and send the original image to the image processing module and the area division module; The image processing module is used to receive the original image, and divide the original image into regions to obtain S sub-region images; Acquire a sub-region image label according to the sub-region image and an image recognition model; wherein the image recognition model is established based on an artificial intelligence model; and sending the sub-region image and the sub-region image label to the region division module; The region division module is used to receive the original image, the sub-region image and the sub-region image label, and divide the original image into regions according to the sub-region image and the sub-region image label; The image processing module receives the original image and divides the original image into regions to obtain S sub-region images. The specific process includes: The image processing module receives the original image; Divide the original image into M rows N columns, get M N standard images; Select any standard image and mark it as a first seed standard image, where the first seed standard image is marked as 1; Comparing the first seed standard image with adjacent standard images in terms of category; The adjacent standard image and the first seed standard image belong to the same category, and the adjacent standard image is marked as 1; The adjacent standard image and the first seed standard image do not belong to the same category, and the adjacent standard image is marked as 0; After the adjacent standard images of the first seed standard image are marked; If all adjacent standard images are marked as 1, the range of adjacent standard images is expanded until all standard images are marked; If the labels of the adjacent standard images are 0 and 1; Obtain an adjacent standard image of a standard image marked as 1, where the adjacent standard image is not marked; Comparing the standard image with the first seed standard image by category, and repeating the above process until all standard images are marked; Acquire a standard image marked as 1, and integrate the standard image into a first sub-region image; Select any standard image marked as 0 and set it as the second seed standard image, and the second seed standard image is marked as 2; Comparing the second seed standard image with adjacent standard images in terms of category; The adjacent standard image and the second seed standard image belong to the same category, and the adjacent standard image is marked as 2; The adjacent standard image and the second seed standard image do not belong to the same category, and the adjacent standard image is marked as 0; After the adjacent standard images of the second sub-standard image are marked; If all adjacent standard images are marked as 2, the range of adjacent standard images is expanded until all standard images are marked; If the labels of the adjacent standard images are 0 and 2; Acquire an adjacent standard image of a standard image marked as 2, and the adjacent standard image is not marked as 2; Comparing the adjacent standard images with the second seed standard images by category, and repeating the above process until all standard images are marked; Acquire a standard image marked as 2, and integrate the standard image into a second sub-region image; Repeat the above process to obtain S sub-region images; The image segmentation model is established based on an artificial intelligence model, and the specific process includes: Obtain standard training images from the image processing module; The artificial intelligence model is trained using standard training images, and the trained artificial intelligence model is marked as an image segmentation model; The standard training images include several groups of input images and corresponding region labels, and the content attributes of the input images and the sub-region images are consistent; The artificial intelligence model includes a deep convolutional neural network model or a RBF neural network model.

2. The image recognition system with self-regional division according to claim 1, characterized in that: The image acquisition module includes a camera device.

3. The image recognition system with self-regional division according to claim 2, characterized in that: The camera device acquires an original image of a target area and sends the original image to the image processing module.

4. The image recognition system with self-regional division according to claim 1, characterized in that: The region division module receives the original image, the sub-region image and the sub-region image label, and divides the original image into regions according to the sub-region image and the sub-region image label. The specific process includes: The region division module receives the sub-region image and the sub-region image label; The region division module merges the original image, the sub-region image and the sub-region image label; Obtaining the original image with region differentiation; Different sub-regions on the original image are marked with different colors.

5. The image recognition system with self-regional division according to claim 4, characterized in that: The image acquisition module is in communication and / or electrically connected with the image processing module; The image acquisition module is in communication and / or electrically connected with the area division module; The image processing module is in communication and / or electrically connected with the region division module.

6. A method for image recognition by self-regional segmentation, characterized in that: The image recognition system for self-dividing regions according to any one of claims 1 to 5, wherein the specific method comprises the following steps: Step 1: The camera device obtains the original image of the target area and sends the original image to the image processing module; Step 2: The image processing module receives the original image and divides the original image into regions to obtain S sub-region images; Step 3: Obtain sub-region image labels according to the sub-region images and the image recognition model; and send the sub-region images and sub-region image labels to the region division module; Step 4: The region division module receives the original image, the sub-region image and the sub-region image label, and divides the original image into regions according to the sub-region image and the sub-region image label.

Images