A method for intelligent plant identification

Abstract

The present invention discloses a method for intelligent plant identification, which specifically relates to the technical field of plant identification. The method comprises the following steps: first, setting a plurality of plant identification points along the direction of plant growth characteristics in a plant image, and collecting data on plant shape characteristics, wherein the plurality of plant identification points are set at non-equidistant distances along the plant growth direction, and the density of the plant identification points gradually increases along the plant growth direction with the plant growth density; second, obtaining characteristic data on each flower and fruit in the plant image within the same plant identification time period, analyzing the flower and fruit characteristic data through a component analysis method, and obtaining the most critical data features using a discrete model. The present invention processes the flower and fruit characteristic data through a discrete model to obtain the most critical data features, thereby reducing the amount of data processing, alleviating the computational pressure of a flower and fruit classification process, and obtaining faster plant identification capabilities.

Description

Technical Field

[0001] The present invention relates to the technical field of plant identification, and more particularly to a method for intelligent plant identification. Background Art

[0002] There are rich and diverse life forms on the earth, among which the number of plant species is countless. In addition to common agricultural plants and potted plants, there are also many wild plants, which play the most important role in natural resources and environmental resources. The diversity of wild plants and people's lack of understanding of them have led to the over-picking of wild plants. Therefore, plant classification is of great significance to the protection of biodiversity, gardening and biosafety. In order to identify plants, people take pictures of plants and identify the photos to obtain plant classification information.

[0003] At present, the main method for plant identification is through plant leaf identification. Plant leaf identification uses the flat leaf structure of the plant to perform classification and identification using photo processing and plant feature extraction. However, due to the many similarities between plant species, the plant identification algorithm is inaccurate, which increases the difficulty of plant identification. Therefore, a new method for intelligent plant identification is provided. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a method for intelligent plant identification to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solution: a method for intelligent plant identification, comprising the following steps:

[0006] Step 101: In a plant image, a plurality of plant identification points are set along the plant growth characteristic direction to collect data on the plant shape characteristics, wherein the plurality of plant identification points are set at non-equidistant distances along the plant growth direction, and the density of the plant identification points gradually increases along the plant growth direction as the plant growth density increases.

[0007] Step 201: Acquire feature data of each flower and fruit in the plant image using the same plant recognition time period, analyze the feature data of the flowers and fruits using component analysis, and obtain the most critical data features using a discrete model;

[0008] Step 301: Obtain a comparison function, obtain a plant feature database based on the same plant growth cycle, perform comparison using the comparison function, modify and correct the comparison function when using it, and calculate the actual category of the plant identified this time using neural network technology after determining the modification and correction;

[0009] Step 401: Obtain several plant shape characteristics and complete the comparison function correction based on the plant feature database. If the plant feature database does not have the plant data shape characteristics, the data record is saved and the plant feature database is improved through the online plant expert mode to prepare for the next plant identification.

[0010] The specific formula of the discrete model is as follows: Where L is a discrete value, h i is the flower and fruit sample data, i and n are discrete values containing i n-dimensional vector flower and fruit samples, that is, i flower and fruit samples, each flower and fruit sample contains n-dimensional vector information. The flower and fruit are reduced in dimensionality by principal component analysis, and then the discrete model is used to extract the most critical data features from the data, reducing the amount of data processing and alleviating the computational pressure of the plant flower and fruit classification process. The smaller the discrete value, the more compact the flower and fruit data feature data.

[0011] In a preferred embodiment, the plant image is processed based on size normalization and grayscale through an automatic encoder to use the original plant image data value as the primary image feature, and the plant shape features are extracted by using the plant image dimensionality reduction method to obtain plant image preprocessing data, and then the probabilistic graphical model is used to detect the main plant shape features of the plant image data. The plant shape main feature data is unsupervisedly pre-trained based on the constraint conditions to obtain high-order plant shape main feature data, so that the local information of the plant shape main features is trained layer by layer to reduce the analysis error of the plant shape feature data. The constraint condition is that the plant visible layer and the plant hidden layer constitute a bipartite graph, and the plant shape features include the shape, color, thickness and length of plant roots, stems, leaves, petioles, leaf buds and stipules.

[0012] In a preferred embodiment, the flower and fruit feature data analysis is to divide the flowers and fruits in the plant image into blocks, and then use the image complete local binary algorithm to extract the flower feature data, obtain the texture features of local differences, the flower and fruit amplitude value feature grayscale values, and obtain three flower and fruit histograms, and then connect all the blocked flower and fruit images as the subsequent classification total features of the flower and fruit images, and then use the principal component analysis method to perform texture feature dimensionality reduction on the subsequent classification total features of the flower and fruit images and perform discrete model calculation to remove the flower and fruit features with the smallest discrete contribution.

[0013] In a preferred embodiment, the comparison function is based on the comparison of plant shape feature data and flower and fruit feature data with a plant feature database, that is, a comparison type of a custom nonlinear model is constructed, the identification requirement time is set as an independent variable, and then the comparison type of the plant shape feature data and the flower and fruit feature data is obtained, and the comparison is created and a comparison model is drawn using the comparison type. The comparison model is as follows: F = Z*x+G*sin(x), where F is the comparison model item, Z is the plant shape feature data value, G is the flower and fruit feature data value, and x is the comparison difference coefficient. The smaller the comparison difference coefficient, the higher the comparison completion and the more accurate the plant identification.

[0014] In a preferred embodiment, the contrast function has a good contrast effect on the plant identification that has been generated. After completing the preliminary comparison, the plant shape feature data, flower and fruit data and the contrast function are used to modify and correct them, and new contrast functions are continuously determined. Then, the accurate plant category is obtained based on the overall data characteristics of the plant.

[0015] In a preferred embodiment, the comparison function is tested by SW to determine whether the comparison function conforms to the simulated flower and fruit data. If not, the SW test is continuously performed until the comparison function passes. When performing function comparison, the flower and fruit feature data are subjected to noise reduction processing.

[0016] In a preferred embodiment, the plant feature database is a plant remote identification database based on a combination of cloud storage and mobile terminals. The plant image is uploaded to the cloud server through the mobile terminal, and compared with the plant shape feature data, flower and fruit feature data and the plant feature database. If the plant feature database does not have the identified plant, all the feature information of the plant and the geographical location where the plant was found are stored. Then, online authoritative experts connected to the cloud server observe, analyze, and experiment on the plant, or conduct investigations and research based on the geographical location to supplement the plant information and improve the plant feature database.

[0017] The technical effects and advantages of the present invention are as follows:

[0018] The present invention specifically sets plant identification points to enhance the collection of plant growth characteristic data, uses plant shape characteristics for unsupervised training, obtains main plant shape characteristic data, and reduces plant shape characteristic data analysis errors; processes flower and fruit characteristic data through a discrete model to obtain the most critical data features, reduces the data processing volume, alleviates the calculation pressure of the flower and fruit classification process, obtains faster plant identification capabilities, and improves the accuracy of plant identification; plant shape characteristic data and flower and fruit characteristic data are compared and fitted with a plant characteristic database through a comparison model, further improving the accuracy of plant identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 Flow chart of the method of the present invention.

[0020] Figure 2 This is a structural diagram of the method of the present invention. DETAILED DESCRIPTION

[0021] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0022] Example

[0023] See also Figure 1-2 This embodiment provides a method for intelligent plant identification, which specifically includes the following steps:

[0024] Step 101: Set a number of plant identification points along the plant growth characteristic direction in the plant image to collect data on the plant shape characteristics, wherein the number of plant identification points are set at non-equidistant intervals along the plant growth direction. Along the plant growth direction, the density of the plant identification points gradually increases with the plant growth density. By setting the plant identification points, the collection of plant growth characteristic data is enhanced;

[0025] In this embodiment, it should be specifically explained that the plant image uses an automatic encoder to take the original plant image data value as the primary image feature based on size normalization and grayscale processing, and uses the plant image dimensionality reduction method to extract plant shape features to obtain plant image preprocessing data, and then uses a probabilistic graph model to detect the main plant shape features of the plant image data. Based on the constraint conditions, the plant shape main feature data is unsupervisedly pre-trained to obtain high-order plant shape main feature data, so that the local information of the plant shape main features is trained layer by layer, reducing the error in plant shape feature data analysis. The constraint conditions are that the plant visible layer and the plant hidden layer constitute a bipartite graph. The plant shape features include the shape, color, thickness and length of plant roots, stems, leaves, petioles, leaf buds and stipules, which are not limited here.

[0026] Step 201: Acquire feature data of each flower and fruit in the plant image using the same plant recognition time period, analyze the feature data of the flowers and fruits using component analysis, and obtain the most critical data features using a discrete model;

[0027] In this embodiment, it should be specifically explained that the flower and fruit feature data analysis is to divide the flowers and fruits in the plant image into blocks, and then use the image complete local binary algorithm to extract the flower feature data, obtain the texture features of local differences, the flower and fruit amplitude value feature grayscale values, and obtain three flower and fruit histograms. Then, all the divided flower and fruit images are connected as the total features for subsequent classification of the flower and fruit images. Then, the principal component analysis method is used to perform texture feature dimensionality reduction on the total features for subsequent classification of the flower and fruit images, and then a discrete model calculation is performed. The flower and fruit features with the smallest discrete contribution are removed to obtain the most critical data features. The specific formula of the discrete model is as follows: Where L is a discrete value, h i is the flower and fruit sample data, i and n are discrete values containing i n-dimensional vector flower and fruit samples, that is, i flower and fruit samples, each flower and fruit sample contains n-dimensional vector information, the flower and fruit are reduced in dimensionality by principal component analysis, and then the most critical data features are extracted from the data using discrete model calculation, which reduces the amount of data processing and alleviates the computational pressure of the plant flower and fruit classification process. The smaller the discrete value, the more compact the flower and fruit data feature data, thereby improving the accuracy of flower and fruit recognition in plant images.

[0028] Step 301: Obtain a comparison function, obtain a plant feature database based on the same plant growth cycle, perform comparison using the comparison function, modify and correct the comparison function when using it, and calculate the actual category of this plant identification based on the neural network technology after determining the modification and correction.

[0029] In this embodiment, it should be specifically explained that the comparison function is based on the comparison of plant shape feature data and flower and fruit feature data with the plant feature database, that is, the comparison type of the custom nonlinear model is constructed, the recognition requirement time is set as the independent variable, and then the comparison type of the plant shape feature data and the flower and fruit feature data is obtained, and the comparison is created and the comparison model is drawn using the comparison type. The comparison model is as follows: F = Z*x+G*sin(x), F is the comparison model item, Z is the plant shape feature data value, G is the flower and fruit feature data value, and x is the comparison difference coefficient. The smaller the comparison difference coefficient, the higher the comparison completion degree and the more accurate the plant identification.

[0030] In this embodiment, it should be specifically explained that the contrast function has a good contrast effect on the plant identification that has been generated. After completing the preliminary comparison, the plant shape feature data, flower and fruit data and the contrast function are used to modify and correct them, and new contrast functions are continuously determined. Then, the accurate plant category is obtained based on the overall data characteristics of the plant.

[0031] Among them, it should be specifically explained that the comparison function is judged by SW test whether the comparison function conforms to the simulated flower and fruit data. If not, the SW test is continuously performed until the comparison function passes. When performing function comparison, the flower and fruit characteristic data are denoised to reduce the interference of abnormal data and improve the accuracy of comparison function data processing.

[0032] Step 401: Obtain several plant shape characteristics and complete the comparison function correction based on the plant feature database. If the plant feature database does not have the plant data shape characteristics, the data record is saved and the plant feature database is improved through the online plant expert mode to prepare for the next plant identification.

[0033] In this embodiment, it should be specifically explained that the plant feature database is a plant remote identification database based on a combination of cloud storage and mobile terminals. The plant image is uploaded to the cloud server through the mobile terminal, and compared with the plant shape feature data, flower and fruit feature data and the plant feature database. If the plant feature database does not have the plant identified this time, all the feature information of the plant and the geographical location where the plant was found are stored. Then, the online authoritative experts connected to the cloud server observe, analyze, and experiment on the plants, or conduct investigations and research based on the geographical location to supplement the plant information, improve the plant feature database, and provide a data basis for the next plant identification.

[0034] The above-described embodiments can be applied to a computer system / server that can operate in conjunction with numerous other general-purpose or special-purpose computing system environments or configurations. Examples of well-known computing systems, environments, and / or configurations suitable for use with the computer system / server include, but are not limited to, personal computer systems, server computer systems, thin clients, thick clients, handheld or laptop devices, microprocessor-based systems, set-top boxes, programmable consumer electronics, network personal computers, minicomputer systems, mainframe computer systems, and distributed cloud computing technology environments that include any of the above.

[0035] Computer systems / servers may be described in the general context of computer system-executable instructions, such as program modules, executed by a computer system. Generally, program modules may include routines, programs, objects, components, logic, data structures, and the like, that perform specific tasks or implement specific abstract data types. Computer systems / servers may be implemented in a distributed cloud computing environment where tasks are performed by remote processing devices linked through a communications network. In a distributed cloud computing environment, program modules may be located on local or remote computer system storage media, including storage devices.

[0036] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for intelligent plant identification, characterized by: The specific steps include: Step 101: In a plant image, a plurality of plant identification points are set along the plant growth characteristic direction to collect data on the plant shape characteristics, wherein the plurality of plant identification points are set at non-equidistant distances along the plant growth direction, and the density of the plant identification points gradually increases along the plant growth direction as the plant growth density increases. Step 201: Acquire feature data of each flower and fruit in the plant image using the same plant recognition time period, analyze the feature data of the flowers and fruits using component analysis, and obtain the most critical data features using a discrete model; Step 301: Obtain a comparison function, obtain a plant feature database based on the same plant growth cycle, perform comparison using the comparison function, modify and correct the comparison function when using it, and calculate the actual category of the plant identified this time using neural network technology after determining the modification and correction; Step 401: Obtain several plant shape characteristics and perform comparison function correction based on the plant feature database. If the plant feature database does not have the plant data shape characteristics, save the data and improve the plant feature database through the online plant expert mode to prepare for the next plant identification. The specific formula of the discrete model is as follows: Where L is a discrete value, h i is the flower and fruit sample data, i and n are discrete values containing i n-dimensional vector flower and fruit samples, that is, i flower and fruit samples, each flower and fruit sample contains n-dimensional vector information, the flower and fruit are reduced in dimension by principal component analysis, and then the most critical data features are extracted from the data using discrete model calculation. The comparison function is based on the comparison of plant shape feature data and flower and fruit feature data with the plant feature database, that is, constructing a comparison type of a custom nonlinear model, setting the recognition requirement time as an independent variable, and then obtaining the comparison type of the plant shape feature data and the flower and fruit feature data. The comparison is created using the comparison type and a comparison model is drawn. The comparison model is as follows: F = Z*x+G*sin(x), where F is a comparison model item, Z is a plant shape feature data value, G is a flower and fruit feature data value, and x is a comparison difference coefficient. The smaller the comparison difference coefficient, the higher the comparison completion degree and the more accurate the plant identification.

2. The method for intelligent plant identification according to claim 1, characterized in that: The plant image is based on size normalization and grayscale processing. The original plant image data value is used as the primary image feature through an automatic encoder, and the plant shape feature is extracted by using the plant image dimensionality reduction method to obtain plant image preprocessing data. Then, the main plant shape features of the plant image data are detected using a probabilistic graphical model. The main plant shape feature data is unsupervisedly pre-trained based on the constraint conditions to obtain high-order plant shape main feature data. The constraint conditions are that the plant visible layer and the plant hidden layer constitute a bipartite graph. The plant shape features include the shape, color, thickness and length of plant roots, stems, leaves, petioles, leaf buds and stipules.

3. The method for intelligent plant identification according to claim 1, characterized in that: The flower and fruit feature data analysis is to divide the flowers and fruits in the plant image into blocks, then use the image complete local binary algorithm to extract the flower feature data, obtain the texture features of local differences and the grayscale values of the flower and fruit amplitude value features, and obtain three flower and fruit histograms. Then, all the divided flower and fruit images are connected as the total features for subsequent classification of the flower and fruit images. Then, the principal component analysis method is used to reduce the texture feature dimensionality of the total features for subsequent classification of the flower and fruit images, and then a discrete model calculation is performed to remove the flower and fruit features with the smallest discrete contribution.

4. The method for intelligent plant identification according to claim 1, wherein: The comparison function has a good comparison effect on the plant identification that has been produced. After completing the preliminary comparison, the plant shape feature data, flower and fruit data and the comparison function are used to modify and correct them, and new comparison functions are continuously determined. Then, the accurate plant category is obtained based on the overall plant data characteristics.

5. The method for intelligent plant identification according to claim 1, characterized in that: The comparison function is judged by SW test whether the comparison function conforms to the simulated flower and fruit data. If not, the SW test is continuously performed until the comparison function passes. When performing the function comparison, the flower and fruit feature data are subjected to noise reduction processing.

6. The method for intelligent plant identification according to claim 1, characterized in that: The plant feature database is a plant remote identification database based on a combination of cloud storage and mobile terminals. The plant image is uploaded to the cloud server through the mobile terminal, and compared with the plant shape feature data, flower and fruit feature data and the plant feature database. If the plant feature database does not have the identified plant, all the feature information of the plant and the geographical location where the plant was found are stored. Then, online authoritative experts connected to the cloud server observe, analyze, and experiment on the plant, or conduct investigations and research based on the geographical location to supplement the plant information and improve the plant feature database.

Images