Image recognition system and method for white space distribution experiment of hydrangea serrulata leaves
By combining color space segmentation, texture gradient features, morphological features and convolution recognition models in the silver edge hydrangea blade image processing, the problem of low recognition accuracy of white space distribution is solved, and higher recognition accuracy and accuracy are achieved.
Patent Information
- Application Number
- CN202510184777.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-02-19
- Publication Date
- 2025-06-06
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
When the prior art recognizes the white space distribution in silver-edged hydrangea leaves, it is affected by light and environmental conditions, resulting in blurred boundaries, making it difficult to accurately distinguish white space from green background, thereby affecting the recognition accuracy.
By collecting leaf experimental images, color space segmentation is performed, the edge color features of the white space area and the texture gradient features of the green background area are extracted, and the pixel topology structure and fractal gradient of edge contour are combined to determine the morphological characteristics and boundary significance of the white part. Finally, the convolutional recognition model is used to achieve accurate recognition of white space distribution.
The recognition accuracy of the white spatial distribution of silver-edged hydrangea leaves is improved, and the error segmentation caused by background interference is reduced, ensuring the accurate identification of white partial boundaries by the model.
Smart Images
Figure CN120107554A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of image processing technology, and more specifically, to an image recognition system and method for white space distribution experiment of silver-edged hydrangea leaves. Background Art
[0002] As an important ornamental plant, the white edges of the leaves of Hydrangea argentea usually have specific morphological characteristics and distribution patterns, which are of great significance for the evaluation of plant growth conditions, species identification and breeding research. In recent years, with the continuous advancement of image processing technology and computer vision, plant morphological analysis methods based on digital images have gradually emerged, providing new possibilities for efficient and automated analysis of the spatial distribution characteristics of plant leaves.
[0003] In the prior art, the boundary recognition of the white space of the silver-edged hydrangea leaves usually adopts simple methods such as color segmentation and edge detection. These methods rely on the color information of the image or the local brightness change to extract the edge of the object. However, due to the change of the direction, intensity or environmental conditions of the light source in the image, the brightness or color of the leaf surface and the background may change, thereby affecting the effect of color segmentation or edge detection. For example, strong illumination or shadow may cause the boundary between the white area and the background to be blurred, and then the simple color segmentation method cannot accurately distinguish the white space from the green background. Therefore, how to realize the convolution recognition of the white part of the silver-edged hydrangea leaves, so as to improve the recognition accuracy of the white space distribution, has become a difficult problem faced by the industry. Summary of the invention
[0004] The present application provides an image recognition system and method for a white space distribution experiment of silver-edged hydrangea leaves, which can realize convolution recognition of the white part of the silver-edged hydrangea leaves, thereby improving the recognition accuracy of the white space distribution.
[0005] In a first aspect, the present application provides an image recognition method for a white space distribution experiment of silver-edged hydrangea leaves, the image recognition method comprising the following steps:
[0006] Collect the experimental images of the leaves of silver-edged hydrangea in the white space distribution experiment;
[0007] Performing color space segmentation on the leaf experimental image to obtain the white space area and the green background area of the silver-edged hydrangea leaf, and then extracting the edge color features of the white space area and the texture gradient features of the green background area, and determining the boundary saliency of the white part in the leaf experimental image through the edge color features and the texture gradient features;
[0008] Determine the pixel connectivity of the white space region according to the pixel topological structure in the leaf experimental image, and determine the morphological characteristics of the white part in the leaf experimental image according to the pixel connectivity and the overall fractal gradient of the edge contour in the leaf experimental image;
[0009] The spatial distribution of the white part in the leaves of the silver-edged hydrangea is convoluted and recognized by using the morphological features and the boundary saliency, so as to obtain the spatial distribution features of the white part in the leaf experimental image.
[0010] In this embodiment, a hyperspectral camera is used to collect leaf experimental images of silver-edged hydrangea leaves in a white space distribution experiment.
[0011] In this embodiment, the leaf experimental image is segmented in color space to obtain the white space area and green background area of the silver-edged hydrangea leaf, which specifically include:
[0012] The leaf experimental image is converted from RGB color space to Lab color space to obtain the Lab image of the silver-edged hydrangea leaf;
[0013] Determine the color channel thresholds for the leaves of silver-edged hydrangea;
[0014] The color channel threshold is used to divide the spatial region of the Lab image into a white spatial region of the silver-edged hydrangea leaves and a green background region.
[0015] In this embodiment, extracting edge color features of the white space area and texture gradient features of the green background area specifically includes:
[0016] Perform edge detection on the white space area to obtain the edge color gradient direction and edge color gradient amplitude of the white space area;
[0017] Determine the edge color characteristics of the white space area by edge color gradient direction and edge color gradient amplitude;
[0018] The local contrast and uniformity of the texture in the green background area are calculated, and then the texture gradient characteristics of the green background area are determined through the local contrast and uniformity.
[0019] In this embodiment, determining the boundary saliency of the white part in the leaf experimental image by using the edge color feature and the texture gradient feature specifically includes:
[0020] Determine the color contrast of the boundary in the white space region by using the edge color feature;
[0021] Determine the local edge entropy of the boundary in the green background area according to the texture gradient feature;
[0022] The boundary saliency of the white part in the leaf experimental image is determined according to the local edge entropy and the color contrast.
[0023] In this embodiment, determining the pixel connectivity of the white space region according to the pixel topological structure in the leaf experimental image specifically includes:
[0024] Extract isolated white pixel blocks and continuous white pixel blocks in the white space area from the pixel topological structure in the leaf experimental image;
[0025] Determining isolated distribution characteristics of isolated white pixel blocks and continuous distribution characteristics of continuous white pixel blocks;
[0026] The isolated distribution features and the continuous distribution features are feature merged to obtain pixel connectivity of the white space area.
[0027] In this embodiment, determining the morphological features of the white part in the leaf experimental image based on the pixel connectivity and the overall fractal gradient of the edge contour in the leaf experimental image specifically includes:
[0028] The morphological complexity of the white part in the leaf experimental image is extracted from the overall fractal gradient of the edge contour in the leaf experimental image;
[0029] The morphological features of the white part in the leaf experimental image are determined according to the pixel connectivity and the morphological complexity.
[0030] In this embodiment, the overall fractal gradient is the fractal dimension change rate of the edge contour in the blade experimental image.
[0031] In this embodiment, the spatial distribution of the white part in the silver-edged hydrangea leaf is convolutionally identified by the morphological features and the boundary saliency, and the spatial distribution features of the white part in the leaf experimental image are obtained, which specifically include:
[0032] Initialize a convolutional recognition model based on a neural network;
[0033] Using the morphological features as input vectors of the white part morphology in the convolution recognition model;
[0034] Using the boundary saliency as an input feature of the white part boundary in a convolutional recognition model;
[0035] The convolutional recognition model is used to identify the spatial distribution of the white part in the leaves of Hydrangea serrulata, and the spatial distribution characteristics of the white part in the leaf experimental images are obtained.
[0036] In a second aspect, the present application provides an image recognition system for a white space distribution experiment of a silver-edged hydrangea leaf, which is used to perform an image recognition method for a white space distribution experiment of a silver-edged hydrangea leaf, and the image recognition system comprises:
[0037] The acquisition module is used to acquire the experimental images of the leaves of the silver-edged hydrangea in the white space distribution experiment;
[0038] A color feature module is used to perform color space segmentation on the leaf experimental image to obtain the white space area and the green background area of the silver-edged hydrangea leaf, and then extract the edge color features of the white space area and the texture gradient features of the green background area, and determine the boundary saliency of the white part in the leaf experimental image through the edge color features and the texture gradient features;
[0039] A morphological feature module is used to determine the pixel connectivity of the white space area according to the pixel topological structure in the leaf experimental image, and determine the morphological features of the white part in the leaf experimental image according to the pixel connectivity and the overall fractal gradient of the edge contour in the leaf experimental image;
[0040] The convolution recognition module is used to perform convolution recognition on the spatial distribution of the white part in the silver-edged hydrangea leaves through the morphological features and the boundary saliency to obtain the spatial distribution features of the white part in the leaf experimental image.
[0041] The technical solution provided by the embodiments disclosed in this application has the following beneficial effects:
[0042] The leaf experimental images of the silver-edged hydrangea leaves in the white space distribution experiment are collected; the leaf experimental images are segmented in color space to obtain the white space area and the green background area of the silver-edged hydrangea leaves, and then the edge color characteristics of the white space area and the texture gradient characteristics of the green background area are extracted, and the boundary salience of the white part in the leaf experimental image is determined by the edge color characteristics and the texture gradient characteristics; the pixel connectivity of the white space area is determined according to the pixel topological structure in the leaf experimental image, and the morphological characteristics of the white part in the leaf experimental image are determined according to the pixel connectivity and the overall fractal gradient of the edge contour in the leaf experimental image; the spatial distribution of the white part in the silver-edged hydrangea leaves is convoluted and recognized by the morphological characteristics and the boundary salience to obtain the spatial distribution characteristics of the white part in the leaf experimental image.
[0043] It can be seen that in the present application, firstly, the texture gradient features of the green background area provide contrast information between the background and the white area, further enhancing the clarity of the boundary between the white part and the background. The evaluation of the boundary saliency not only improves the prominent recognition of the edge, but also effectively reduces the erroneous segmentation caused by background interference, thereby helping to more accurately extract the spatial distribution characteristics of the white area. In the convolution recognition process, the boundary saliency is an important input, which can ensure that the model accurately recognizes the boundary of the white part, thereby improving the overall recognition accuracy; then, the overall fractal gradient can further capture the complexity and details of the morphology of the white part. By calculating the fractal dimension, the geometric complexity of the white area can be quantitatively described. The precise extraction of morphological features provides a clearer and more meaningful input for the convolutional neural network, so that the network can more accurately capture the spatial distribution characteristics of the white part, thereby effectively improving the recognition accuracy of the white spatial distribution.
[0044] In summary, the technical solution adopted in the present application can realize the convolution recognition of the white part in the silver-edged hydrangea leaves, thereby improving the recognition accuracy of the white space distribution. BRIEF DESCRIPTION OF THE DRAWINGS
[0045] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only the embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0046] Figure 1 It is a flow chart of the image recognition method for the white space distribution experiment of silver-edged hydrangea leaves provided in the present application;
[0047] Figure 2 is an exemplary flow chart for determining the morphological features of the white portion in the leaf experimental image provided by the present application;
[0048] Figure 3 This is a module structure diagram of an image recognition system for white space distribution experiment of silver-edged hydrangea leaves provided in this application. DETAILED DESCRIPTION
[0049] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0050] The embodiment of the present application provides an image recognition system and method for a white space distribution experiment of a silver-edged hydrangea leaf. The core of the system is to collect a leaf experimental image of a silver-edged hydrangea leaf in a white space distribution experiment; perform color space segmentation on the leaf experimental image to obtain a white space region and a green background region of the silver-edged hydrangea leaf, and then extract the edge color features of the white space region and the texture gradient features of the green background region, and determine the boundary salience of the white part in the leaf experimental image through the edge color features and the texture gradient features; determine the pixel connectivity of the white space region according to the pixel topological structure in the leaf experimental image, and determine the morphological features of the white part in the leaf experimental image according to the pixel connectivity and the overall fractal gradient of the edge contour in the leaf experimental image; perform convolution recognition on the spatial distribution of the white part in the silver-edged hydrangea leaf through the morphological features and the boundary salience to obtain the spatial distribution features of the white part in the leaf experimental image. The above scheme can realize the convolution recognition of the white part in the silver-edged hydrangea leaf, thereby improving the recognition accuracy of the white space distribution.
[0051] Embodiment 1: In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods. Figure 1 As shown in FIG. 1 , this figure is an exemplary flow chart of an image recognition method for a white space distribution experiment of silver-edged hydrangea leaves according to this embodiment of the present application, and the image recognition method comprises the following steps:
[0052] In step S1, a leaf experimental image of a silver-edged hydrangea leaf in a white space distribution experiment is collected.
[0053] It should be noted that, in the present application, the leaf experimental image represents the appearance image of the silver-edged hydrangea leaves in the white space distribution experiment; in specific implementation, a hyperspectral camera can be used to collect the leaf experimental image of the silver-edged hydrangea leaves in the white space distribution experiment.
[0054] In step S2, the leaf experimental image is segmented in color space to obtain the white space area and green background area of the silver-edged hydrangea leaf, and then the edge color features of the white space area and the texture gradient features of the green background area are extracted, and the boundary saliency of the white part in the leaf experimental image is determined by the edge color features and the texture gradient features.
[0055] In this embodiment, the color space segmentation of the leaf experimental image to obtain the white space area and green background area of the silver-edged hydrangea leaf can be achieved by the following steps:
[0056] The leaf experimental image is converted from RGB color space to Lab color space to obtain the Lab image of the silver-edged hydrangea leaf;
[0057] Determine the color channel thresholds for the leaves of silver-edged hydrangea;
[0058] The color channel threshold is used to divide the spatial region of the Lab image into a white spatial region of the silver-edged hydrangea leaves and a green background region.
[0059] It should be noted that in this application, the white space area refers to the area in the image that is mainly white, which is used to highlight a specific object or area. In this solution, it refers to the white part of the silver-edged hydrangea leaf; the green background area refers to the area in the image that is mainly green, usually the background or other parts of the leaf, which is used to distinguish from the target area (e.g., white space);
[0060] In the specific implementation, first, the leaf experimental image is converted from the RGB color space to the Lab color space, and the converted leaf experimental image can be used as the Lab image of the silver-edged hydrangea leaf; then, the color channel threshold of the silver-edged hydrangea leaf can be preset according to historical experience; finally, the Lab image can be segmented by color threshold judgment (for example: OpenCV's inRange() function) to generate a binary mask image, so as to classify the pixels into the white space area or the green background area, that is, for each pixel in the Lab image, it is determined whether the Lab value of the pixel is within the color channel threshold range. If the Lab value of the pixel is within the color channel threshold range, it is marked as a white area, otherwise it is marked as a green background area. The marking of each pixel in the Lab image can be completed in the above manner, and the white space area and green background area of the silver-edged hydrangea leaf can be obtained.
[0061] It should be noted that the perceptual uniformity and brightness-chromaticity separation characteristics of the Lab color space are used for more accurate color segmentation. In the Lab color space, the L channel represents brightness information, the a channel represents the red-green axis, and the b channel represents the yellow-blue axis. Since there is a significant difference in the color distribution of the white area of the silver-edged hydrangea leaves and the green background on the a and b channels, the white space area and the green background area of the silver-edged hydrangea leaves can be distinguished by the numerical threshold of the color channel.
[0062] In this embodiment, the extraction of edge color features of the white space area and texture gradient features of the green background area can be achieved by using the following steps:
[0063] Perform edge detection on the white space area to obtain the edge color gradient direction and edge color gradient amplitude of the white space area;
[0064] Determine the edge color characteristics of the white space area by edge color gradient direction and edge color gradient amplitude;
[0065] The local contrast and uniformity of the texture in the green background area are calculated, and then the texture gradient characteristics of the green background area are determined through the local contrast and uniformity.
[0066] It should be noted that, in the present application, the texture gradient feature represents the structural change of the local texture in the white space area and its spatial distribution; the local contrast represents the difference of the pixel intensity in the local area in the white space area, and the uniformity represents the balance of the pixel distribution in the local area in the white space area; the edge color feature represents the color change feature of the edge in the green background area; the edge color gradient direction represents the direction of the color change at the edge in the green background area, and the edge color gradient amplitude represents the intensity of the color change at the edge in the green background area.
[0067] In the specific implementation, first, the Sobel operator can be used to perform edge detection on the white space area, so as to extract the edge gradient amplitude and all gradient directions from the edge detection results, and the set of all gradient directions can be used as the edge color gradient direction of the white space area, and the gradient amplitude can be used as the edge color gradient amplitude of the white space area; then, the set of edge color gradient direction and edge color gradient amplitude can be used as the edge color feature of the white space area; finally, the grayscale co-occurrence matrix of the texture in the green background area can be used to generate the grayscale co-occurrence matrix, so as to use the contrast feature in the grayscale co-occurrence matrix as the local contrast, and use the standard deviation of all elements in the grayscale co-occurrence matrix as the uniformity, and the set of local contrast and uniformity can be used as the texture gradient feature of the green background area.
[0068] In this embodiment, the following steps can be used to determine the boundary saliency of the white part in the leaf test image by using the edge color feature and the texture gradient feature:
[0069] Determine the color contrast of the boundary in the white space region by using the edge color feature;
[0070] Determine the local edge entropy of the boundary in the green background area according to the texture gradient feature;
[0071] The boundary saliency of the white part in the leaf experimental image is determined according to the local edge entropy and the color contrast.
[0072] In the specific implementation, first, each edge color gradient direction in the edge color feature can be used as the direction of change of the edge color gradient amplitude to obtain the gradient vector of each edge color gradient direction, and the cosine similarity between each gradient vector can be calculated, so that the standard deviation of all cosine similarities can be used as the color contrast of the boundary in the white space area; then, the product of uniformity and local contrast in the texture gradient feature can be used as the local edge entropy of the boundary in the green background area; finally, the influence weights of local edge entropy and color contrast on the prominence of the white part in the leaf experimental image can be preset through experimental statistics, so that the weighted sum of local edge entropy and color contrast can be calculated using the two influence weights in combination with a weighted combination method as the boundary prominence of the white part in the leaf experimental image.
[0073] It should be noted that in this application, boundary saliency is an indicator used to measure the recognizability between the white part and the surrounding area in the leaf experimental image; color contrast represents the intensity of color difference between different color areas in the white space area; local edge entropy represents the uncertainty of the edge area in the green background area.
[0074] In step S3, the pixel connectivity of the white space area is determined according to the pixel topological structure in the leaf experimental image, and the morphological characteristics of the white part in the leaf experimental image are determined according to the pixel connectivity and the overall fractal gradient of the edge contour in the leaf experimental image.
[0075] In this embodiment, the pixel connectivity of the white space region is determined according to the pixel topological structure in the leaf experimental image by using the following steps:
[0076] Extract isolated white pixel blocks and continuous white pixel blocks in the white space area from the pixel topological structure in the leaf experimental image;
[0077] Determining isolated distribution characteristics of isolated white pixel blocks and continuous distribution characteristics of continuous white pixel blocks;
[0078] The isolated distribution features and the continuous distribution features are feature merged to obtain pixel connectivity of the white space area.
[0079] In specific implementation, first, the connected domain labeling algorithm can be used to extract isolated white pixel blocks and continuous white pixel blocks in the white space area from the pixel topological structure in the leaf experimental image, wherein the isolated white pixel block refers to an area composed of isolated pixels in the white space area, and the isolated pixels are not directly connected to other white pixels around them, while the continuous white pixel block refers to a group of white pixel blocks connected by adjacency (such as 4 neighborhoods or 8 neighborhoods); then, the pixel counting method can be used to calculate the area (i.e., the number of pixels in the pixel block) and the isolation (i.e., the distance from the surrounding background) of the isolated white pixel block, and the set of the area and the isolation can be used as the isolated distribution feature of the isolated white pixel block, and the pixel counting method can be used to calculate the shortest distance of the connected area, the area of the largest connected area, and the set of the boundary length in the continuous white pixel block as the continuous distribution feature of the continuous white pixel block; finally, the feature merging algorithm (e.g., a multi-scale feature fusion algorithm based on a channel attention mechanism) can be used to merge the isolated distribution feature and the continuous distribution feature, so that the quantized value after the feature merging is used as the pixel connectivity of the white space area.
[0080] It should be noted that, in the present application, pixel connectivity reflects the correlation between pixels within the white space region; isolated distribution features are features used to describe the independence of local regions; and continuous distribution features are features used to describe the coherence of pixel regions.
[0081] Preferably, in this embodiment, reference Figure 2 As shown in FIG. 1 , this figure is an exemplary flow chart for determining the morphological features of the white part in the leaf test image in an embodiment of the present application. In this embodiment, the morphological features of the white part in the leaf test image are determined based on the pixel connectivity and the overall fractal gradient of the edge contour in the leaf test image, which can be specifically implemented by the following steps:
[0082] First, in step S31, the morphological complexity of the white part in the leaf experimental image is extracted from the overall fractal gradient of the edge contour in the leaf experimental image;
[0083] Then, in step S32, the morphological features of the white part in the leaf experimental image are determined according to the pixel connectivity and the morphological complexity.
[0084] It should be noted that, in the present application, morphological features represent the geometric properties of the shape, structure and spatial arrangement of the white part in the leaf experimental image; morphological complexity represents the complexity of the morphology of the white part in the leaf experimental image; the overall fractal gradient is the rate of change of the fractal dimension of the edge contour in the leaf experimental image, and the fractal dimension of the edge contour can be calculated using the box counting method or the particle counting method, thereby calculating the difference between adjacent fractal dimensions as the rate of change of the fractal dimension, and finally the set of all fractal dimension change rates can be used as the overall fractal gradient of the edge contour in the leaf experimental image.
[0085] In the specific implementation, first, all fractal dimension change rates of the white part in the leaf experimental image are screened out from the overall fractal gradient of the edge contour in the leaf experimental image, and the standard deviation of all fractal dimension change rates is used as the morphological complexity of the white part in the leaf experimental image; then, a multivariate regression model based on a neural network is constructed to combine pixel connectivity and morphological complexity to calculate the morphological characteristics of the white part in the leaf experimental image, that is, morphological characteristics = a*pixel connectivity+b*morphological complexity+c, wherein a, b, c are parameters of the multivariate regression model, which are obtained by training with historical operation data. The morphological characteristics of the white part in the leaf experimental image can be obtained in the above manner.
[0086] In step S4, convolution recognition is performed on the spatial distribution of the white part in the leaves of Hydrangea argentea by using the morphological features and the boundary saliency to obtain the spatial distribution features of the white part in the leaf experimental image.
[0087] In this embodiment, the spatial distribution of the white part in the leaves of the silver-edged hydrangea is convoluted and recognized by using the morphological features and the boundary saliency, and the spatial distribution features of the white part in the leaf experimental image are obtained, which can be specifically achieved by the following steps:
[0088] Initialize a convolutional recognition model based on a neural network;
[0089] Using the morphological features as input vectors of the white part morphology in the convolution recognition model;
[0090] Using the boundary saliency as an input feature of the white part boundary in a convolutional recognition model;
[0091] The convolutional recognition model is used to identify the spatial distribution of the white part in the leaves of Hydrangea serrulata, and the spatial distribution characteristics of the white part in the leaf experimental images are obtained.
[0092] It should be noted that the convolution recognition model is a deep learning model built on a convolutional neural network, which is specifically used to process and analyze image data. The technical principle of the convolution recognition model mainly relies on the hierarchical structure of convolutional layers, pooling layers, and fully connected layers. The local features of the input image are automatically extracted through the convolution kernel (filter), and these features are combined layer by layer to finally identify and classify the global features. In the above process, the convolution recognition model first accepts the morphological features and boundary saliency as input, and undergoes a series of convolution operations and activation function processing in the convolution layer to extract the spatial structural features of the white part; then, the pooling layer is used to reduce the data dimension and computational complexity while retaining important features; finally, the high-level features are mapped to a spatially distributed output result through the fully connected layer to complete the recognition and classification of the spatial distribution features of the white part. The advantage of this convolution recognition model is that it can automatically learn and extract effective features from the original image data, reducing the reliance on manually designed features, and has high image processing accuracy and robustness.
[0093] It can be seen that in the present application, firstly, the texture gradient features of the green background area provide contrast information between the background and the white area, further enhancing the clarity of the boundary between the white part and the background. The evaluation of the boundary saliency not only improves the prominent recognition of the edge, but also effectively reduces the erroneous segmentation caused by background interference, thereby helping to more accurately extract the spatial distribution characteristics of the white area. In the convolution recognition process, the boundary saliency is an important input, which can ensure that the model accurately recognizes the boundary of the white part, thereby improving the overall recognition accuracy; then, the overall fractal gradient can further capture the complexity and details of the morphology of the white part. By calculating the fractal dimension, the geometric complexity of the white area can be quantitatively described. The precise extraction of morphological features provides a clearer and more meaningful input for the convolutional neural network, so that the network can more accurately capture the spatial distribution characteristics of the white part, thereby effectively improving the recognition accuracy of the white spatial distribution.
[0094] In summary, the technical solution adopted in the present application can realize the convolution recognition of the white part in the silver-edged hydrangea leaves, thereby improving the recognition accuracy of the white space distribution.
[0095] Embodiment 2: This application provides an image recognition system for the white space distribution experiment of silver-edged hydrangea leaves, referring to Figure 3 As shown, this figure is a schematic diagram of an image recognition system according to this embodiment of the present application, and the image recognition system includes:
[0096] The acquisition module 100 is used to acquire the leaf experimental images of the silver-edged hydrangea leaves in the white space distribution experiment;
[0097] The color feature module 200 is used to perform color space segmentation on the leaf experimental image to obtain the white space area and the green background area of the silver-edged hydrangea leaf, and then extract the edge color features of the white space area and the texture gradient features of the green background area, and determine the boundary saliency of the white part in the leaf experimental image through the edge color features and the texture gradient features;
[0098] The morphological feature module 300 is used to determine the pixel connectivity of the white space area according to the pixel topological structure in the leaf experimental image, and determine the morphological features of the white part in the leaf experimental image according to the pixel connectivity and the overall fractal gradient of the edge contour in the leaf experimental image;
[0099] The convolution recognition module 400 is used to perform convolution recognition on the spatial distribution of the white part in the leaves of the silver-edged hydrangea by using the morphological features and the boundary saliency to obtain the spatial distribution features of the white part in the leaf experimental image.
[0100] The present application is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0101] Those skilled in the art will appreciate that all or part of the steps in the various methods of the above embodiments may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, the storage medium including a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically-erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.
[0102] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.
Claims
1. An image recognition method for white space distribution experiment of silver-edged hydrangea leaves, characterized in that: The image recognition method comprises the following steps: Collect the experimental images of the leaves of silver-edged hydrangea in the white space distribution experiment; Performing color space segmentation on the leaf experimental image to obtain the white space area and the green background area of the silver-edged hydrangea leaf, and then extracting the edge color features of the white space area and the texture gradient features of the green background area, and determining the boundary saliency of the white part in the leaf experimental image through the edge color features and the texture gradient features; Determine the pixel connectivity of the white space region according to the pixel topological structure in the leaf experimental image, and determine the morphological characteristics of the white part in the leaf experimental image according to the pixel connectivity and the overall fractal gradient of the edge contour in the leaf experimental image; The spatial distribution of the white part in the leaves of the silver-edged hydrangea is convoluted and recognized by using the morphological features and the boundary saliency, so as to obtain the spatial distribution features of the white part in the leaf experimental image.
2. The image recognition method for the white space distribution experiment of silver-edged hydrangea leaves as claimed in claim 1, characterized in that: A hyperspectral camera was used to collect experimental images of leaves of silver-edged hydrangea in the white spatial distribution experiment.
3. The image recognition method for the white space distribution experiment of silver-edged hydrangea leaves as claimed in claim 1, characterized in that: The leaf experimental image is segmented in color space to obtain the white space area and green background area of the silver-edged hydrangea leaf, which specifically include: The leaf experimental image is converted from RGB color space to Lab color space to obtain the Lab image of the silver-edged hydrangea leaf; Determine the color channel thresholds for the leaves of silver-edged hydrangea; The color channel threshold is used to divide the spatial region of the Lab image into a white spatial region of the silver-edged hydrangea leaves and a green background region.
4. The image recognition method for the white space distribution experiment of silver-edged hydrangea leaves as claimed in claim 1, characterized in that: Extracting edge color features of white space areas and texture gradient features of green background areas specifically includes: Perform edge detection on the white space area to obtain the edge color gradient direction and edge color gradient amplitude of the white space area; Determine the edge color characteristics of the white space area by edge color gradient direction and edge color gradient amplitude; The local contrast and uniformity of the texture in the green background area are calculated, and then the texture gradient characteristics of the green background area are determined through the local contrast and uniformity.
5. The image recognition method for the white space distribution experiment of silver-edged hydrangea leaves as claimed in claim 1, characterized in that: Determining the boundary saliency of the white part in the leaf experimental image by using the edge color feature and the texture gradient feature specifically includes: Determine the color contrast of the boundary in the white space region by using the edge color feature; Determine the local edge entropy of the boundary in the green background area according to the texture gradient feature; The boundary saliency of the white part in the leaf experimental image is determined according to the local edge entropy and the color contrast.
6. The image recognition method for the white space distribution experiment of silver-edged hydrangea leaves as claimed in claim 1, characterized in that: Determining the pixel connectivity of the white space area based on the pixel topological structure in the leaf experimental image specifically includes: Extract isolated white pixel blocks and continuous white pixel blocks in the white space area from the pixel topological structure in the leaf experimental image; Determining isolated distribution characteristics of isolated white pixel blocks and continuous distribution characteristics of continuous white pixel blocks; The isolated distribution features and the continuous distribution features are feature merged to obtain pixel connectivity of the white space area.
7. The image recognition method for the white space distribution experiment of silver-edged hydrangea leaves as claimed in claim 1, characterized in that: Determining the morphological features of the white part in the leaf experimental image based on the pixel connectivity and the overall fractal gradient of the edge contour in the leaf experimental image specifically includes: The morphological complexity of the white part in the leaf experimental image is extracted from the overall fractal gradient of the edge contour in the leaf experimental image; The morphological features of the white part in the leaf experimental image are determined according to the pixel connectivity and the morphological complexity.
8. The image recognition method for white space distribution experiment of silver-edged hydrangea leaves as claimed in claim 1, characterized in that: The overall fractal gradient is the fractal dimension change rate of the edge contour in the blade experimental image.
9. The image recognition method for white space distribution experiment of silver-edged hydrangea leaves as claimed in claim 1, characterized in that: The spatial distribution of the white part in the silver-edged hydrangea leaf is convoluted and recognized by the morphological features and the boundary saliency, and the spatial distribution features of the white part in the leaf experimental image are obtained, which specifically include: Initialize a convolutional recognition model based on a neural network; Using the morphological features as an input vector of the white part morphology in a convolutional recognition model; Using the boundary saliency as an input feature of the white part boundary in a convolutional recognition model; The convolutional recognition model is used to identify the spatial distribution of the white part in the leaves of Hydrangea serrulata, and the spatial distribution characteristics of the white part in the leaf experimental images are obtained.
10. An image recognition system for a white space distribution experiment of a silver-edged hydrangea leaf, used to execute an image recognition method for a white space distribution experiment of a silver-edged hydrangea leaf as claimed in any one of claims 1 to 9, characterized in that: The image recognition system comprises: The acquisition module is used to acquire the experimental images of the leaves of the silver-edged hydrangea in the white space distribution experiment; A color feature module is used to perform color space segmentation on the leaf experimental image to obtain the white space area and the green background area of the silver-edged hydrangea leaf, and then extract the edge color features of the white space area and the texture gradient features of the green background area, and determine the boundary saliency of the white part in the leaf experimental image through the edge color features and the texture gradient features; A morphological feature module is used to determine the pixel connectivity of the white space area according to the pixel topological structure in the leaf experimental image, and determine the morphological features of the white part in the leaf experimental image according to the pixel connectivity and the overall fractal gradient of the edge contour in the leaf experimental image; The convolution recognition module is used to perform convolution recognition on the spatial distribution of the white part in the silver-edged hydrangea leaves through the morphological features and the boundary saliency to obtain the spatial distribution features of the white part in the leaf experimental image.