Municipal greening nursery stock growth prediction method and system based on big data

Abstract

The invention relates to the technical field of intelligent prediction, and particularly discloses a municipal greening nursery stock growth prediction method and system based on big data, and the method comprises the steps: extracting nursery stock growth state features from nursery stock growth state images of all days through employing an artificial intelligence technology based on deep learning; multi-dimensional growth state time domain change characteristic analysis is carried out on the growth state characteristics of the nursery stocks in each day, and the growth height of the nursery stocks after one week is predicted based on the growth state time domain multi-dimensional characteristics of the nursery stocks. Therefore, automatic prediction of the growth condition of the nursery stock can be realized, the labor cost is saved, and a decision basis is provided for growth planning management of the nursery stock.

Description

Technical Field

[0001] The present application relates to the field of intelligent prediction technology, and more specifically, to a municipal greening seedling growth prediction method and system based on big data. Background Art

[0002] With the acceleration of urbanization, municipal greening construction has become an important part of urban development. Municipal greening construction refers to greening projects carried out within the city, aiming to improve the urban environment, enhance the beauty and livability of the city, improve air quality, and thus improve the quality of life of residents. In municipal greening construction, the growth management of seedlings is a key link. Seedlings refer to small trees that have just been planted or transplanted to urban greening areas. In order to ensure that the seedlings can grow healthily and take root in the urban environment, they need to be properly cared for and managed.

[0003] In the traditional seedling growth management process, the seedling growth is usually predicted based on artificial experience, and then appropriate intervention measures are taken in time based on the prediction results to maintain the healthy growth of the seedlings. However, this seedling growth prediction method based on artificial experience judgment is usually subjective and uncertain, and cannot accurately predict the growth height of the seedlings. Therefore, a municipal greening seedling growth prediction method and system based on big data is expected. Summary of the invention

[0004] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a municipal greening seedling growth prediction method and system based on big data, which uses artificial intelligence technology based on deep learning to extract seedling growth status characteristics from the seedling growth status images of each day, and performs multi-dimensional growth status time domain change characteristics analysis on the seedling growth status characteristics of each day, and predicts the growth height of the seedlings one week later based on the multi-dimensional characteristics of the seedling growth status time domain. In this way, the automatic prediction of seedling growth can be achieved, labor costs can be saved, and a decision-making basis can be provided for the growth planning and management of seedlings.

[0005] Accordingly, according to one aspect of the present application, a method for predicting the growth of municipal greening seedlings based on big data is provided, which includes:

[0006] Acquire the growth status images of the seedlings collected by the camera for multiple days within a predetermined time period;

[0007] Extracting growth state features from the seedling growth state images of multiple days within the predetermined time period to obtain a sequence of seedling growth state semantic focus feature vectors;

[0008] Performing a multi-dimensional time-domain change feature analysis on the sequence of the seedling growth state semantic focus feature vector to obtain a multi-dimensional time-domain feature vector of the seedling growth state;

[0009] Based on the multi-dimensional feature vector of the seedling growth status in time domain, the growth height of the seedling one week later is predicted.

[0010] In the above-mentioned municipal greening seedling growth prediction method based on big data, growth status feature extraction is performed on the seedling growth status images of multiple days within the predetermined time period to obtain a sequence of seedling growth status semantic focus feature vectors, including: passing the seedling growth status images of multiple days within the predetermined time period through a seedling growth status feature extractor based on a convolutional neural network model to obtain a sequence of seedling growth status feature maps; passing the sequence of seedling growth status feature maps through a significant feature extractor to obtain a sequence of seedling growth status semantic focus feature vectors.

[0011] In the above-mentioned municipal greening seedling growth prediction method based on big data, the sequence of the seedling growth state feature map is passed through a significant feature extractor to obtain a sequence of the seedling growth state semantic focus feature vector, including: extracting significant features from the seedling growth state feature map using the following significant feature extraction formula to obtain the seedling growth state semantic focus feature vector; wherein the significant feature extraction formula is:

[0012] V output =GAP{ReLU(Conv 1×1 [Conv 3×3 (F input )])}

[0013] Among them, V output is the semantic focused feature vector of the seedling growth status, F input is the characteristic graph of the seedling growth status, GAP represents the pooling operation, ReLU represents the ReLU activation processing, Conv 1×1 (·) indicates a convolution operation based on a 1×1 convolution kernel. 3×3 (·) indicates a convolution operation based on a 3×3 convolution kernel.

[0014] In the above-mentioned municipal greening seedling growth prediction method based on big data, a multi-dimensional time domain change feature analysis is performed on the sequence of the seedling growth state semantic focus feature vectors to obtain the seedling growth state time domain multi-dimensional feature vector, including: calculating the growth fluctuation semantic measurement coefficient between each two adjacent seedling growth state semantic focus feature vectors in the sequence of the seedling growth state semantic focus feature vectors to obtain a seedling growth state time series change feature vector composed of multiple growth fluctuation semantic measurement coefficients; cascading the sequence of the seedling growth state semantic focus feature vectors to obtain a full time domain seedling growth state semantic feature vector; fusing the seedling growth state time series change feature vector and the full time domain seedling growth state semantic feature vector to obtain the seedling growth state time domain multi-dimensional feature vector.

[0015] In the above-mentioned municipal greening seedling growth prediction method based on big data, the growth fluctuation semantic measurement coefficient between each two adjacent seedling growth state semantic focus feature vectors in the sequence of the seedling growth state semantic focus feature vector is calculated to obtain a seedling growth state time series change feature vector composed of multiple growth fluctuation semantic measurement coefficients, including: using the following growth fluctuation semantic measurement formula to calculate the growth fluctuation semantic measurement coefficient between each two adjacent seedling growth state semantic focus feature vectors in the sequence of the seedling growth state semantic focus feature vector to obtain a seedling growth state time series change feature vector composed of multiple growth fluctuation semantic measurement coefficients;

[0016] The growth fluctuation semantic measurement formula is:

[0017]

[0018] in, is the semantic focused feature vector of the seedling growth status mentioned above, is the semantic focus feature vector of the latter seedling growth state, N is the dimension of each semantic focus feature vector of the seedling growth state, H p is the growth fluctuation semantic measurement coefficient, and log represents the logarithmic function operation with base 2.

[0019] In the above-mentioned municipal greening seedling growth prediction method based on big data, the seedling growth state temporal change feature vector and the full time domain seedling growth state semantic feature vector are fused to obtain the seedling growth state time domain multi-dimensional feature vector, including: using a projection layer to fuse the seedling growth state temporal change feature vector and the full time domain seedling growth state semantic feature vector to obtain the seedling growth state time domain multi-dimensional feature vector.

[0020] In the above-mentioned municipal greening seedling growth prediction method based on big data, a projection layer is used to fuse the seedling growth state time series change feature vector and the full time domain seedling growth state semantic feature vector to obtain the seedling growth state time domain multi-dimensional feature vector, including: using the projection layer to fuse the seedling growth state time series change feature vector and the full time domain seedling growth state semantic feature vector using the following projection feature fusion formula to obtain the seedling growth state time domain multi-dimensional feature vector; wherein, the projection feature fusion formula is:

[0021]

[0022] Among them, V1 is the temporal change feature vector of the seedling growth state, V2 is the semantic feature vector of the seedling growth state in the entire time domain, and V f is the multi-dimensional feature vector of the seedling growth state in time domain, Represents projection fusion processing.

[0023] In the above-mentioned municipal greening seedling growth prediction method based on big data, the growth height of the seedlings one week later is predicted based on the multi-dimensional feature vector of the seedling growth status in time domain, including: performing feature distribution correction on the multi-dimensional feature vector of the seedling growth status in time domain to obtain the corrected multi-dimensional feature vector of the seedling growth status in time domain; passing the corrected multi-dimensional feature vector of the seedling growth status in time domain through a decoder-based seedling growth predictor to obtain a predicted value, and the predicted value is used to represent the predicted value of the growth height of the seedlings one week later.

[0024] In the above-mentioned municipal greening seedling growth prediction method based on big data, the corrected seedling growth state time domain multi-dimensional feature vector is passed through a decoder-based seedling growth predictor to obtain a predicted value, including: using the decoder to decode and regress the corrected seedling growth state time domain multi-dimensional feature vector using the following decoding formula to obtain the growth height prediction value of the seedling one week later, wherein the decoding formula is: Among them, X is the multi-dimensional feature vector of the corrected seedling growth state in the time domain, Y is the predicted value, W is the weight matrix, Represents matrix multiplication.

[0025] According to another aspect of the present application, a municipal greening seedling growth prediction system based on big data is provided, which includes:

[0026] A seedling growth status monitoring module is used to obtain the seedling growth status images collected by the camera for multiple days within a predetermined time period;

[0027] A seedling growth state feature extraction module is used to extract growth state features from the seedling growth state images of multiple days within the predetermined time period to obtain a sequence of seedling growth state semantic focus feature vectors;

[0028] A seedling growth state temporal variation feature analysis module, used for performing a multi-dimensional temporal variation feature analysis on the sequence of the seedling growth state semantic focus feature vectors to obtain a seedling growth state temporal multi-dimensional feature vector;

[0029] The seedling growth height prediction module is used to predict the growth height of the seedlings one week later based on the multi-dimensional feature vector of the seedling growth state in the time domain.

[0030] Compared with the prior art, the municipal greening seedling growth prediction method and system based on big data provided by the present application uses artificial intelligence technology based on deep learning to extract seedling growth status features from the seedling growth status images of each day, and performs multi-dimensional growth status time domain change feature analysis on the seedling growth status features of each day, and predicts the growth height of the seedlings one week later based on the multi-dimensional time domain features of the seedling growth status. In this way, the automatic prediction of seedling growth can be achieved, labor costs can be saved, and a decision-making basis can be provided for the growth planning and management of seedlings. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0032] Figure 1 This is a flow chart of a municipal greening seedling growth prediction method based on big data according to an embodiment of the present application.

[0033] Figure 2 It is a schematic diagram of the architecture of the municipal greening seedling growth prediction method based on big data according to an embodiment of the present application.

[0034] Figure 3 This is a flowchart of extracting growth status features from seedling growth status images of multiple days within the predetermined time period in the municipal greening seedling growth prediction method based on big data according to an embodiment of the present application to obtain a sequence of seedling growth status semantic focus feature vectors.

[0035] Figure 4The present invention is a flowchart of performing a multi-dimensional time-domain change characteristic analysis on the sequence of the semantically focused feature vectors of the seedling growth status in the municipal greening seedling growth prediction method based on big data according to an embodiment of the present application to obtain a multi-dimensional time-domain feature vector of the seedling growth status.

[0036] Figure 5 This is a flowchart of predicting the growth height of seedlings one week later based on the multi-dimensional feature vector of the seedling growth status in the time domain in the municipal greening seedling growth prediction method based on big data according to an embodiment of the present application.

[0037] Figure 6 It is a block diagram of a municipal greening seedling growth prediction system based on big data according to an embodiment of the present application. DETAILED DESCRIPTION

[0038] To facilitate the understanding of the embodiments of the present invention, the specific embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the specific embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0039] Figure 1 This is a flow chart of a municipal greening seedling growth prediction method based on big data according to an embodiment of the present application. Figure 2 FIG. 1 is a schematic diagram of the architecture of a method for predicting the growth of municipal greening seedlings based on big data according to an embodiment of the present application. Figure 1 and Figure 2 As shown, according to the municipal greening seedling growth prediction method based on big data according to the embodiment of the present application, the steps include: S110, acquiring seedling growth status images of multiple days within a predetermined time period collected by a camera; S120, performing growth status feature extraction on the seedling growth status images of multiple days within the predetermined time period to obtain a sequence of seedling growth status semantic focus feature vectors; S130, performing multi-dimensional time domain change feature analysis on the sequence of seedling growth status semantic focus feature vectors to obtain a seedling growth status time domain multi-dimensional feature vector; S140, predicting the growth height of the seedlings one week later based on the seedling growth status time domain multi-dimensional feature vector.

[0040] As mentioned in the above background technology, in the municipal greening construction, the growth management of seedlings is a key link. By carrying out growth management on seedlings, it can be ensured that they grow healthily and luxuriantly, so as to play the best effect in greening projects. In the traditional seedling growth management process, the seedling growth situation is usually predicted based on artificial experience, and then appropriate intervention measures are taken in time based on the prediction results to maintain the healthy growth of seedlings. However, this seedling growth prediction method based on artificial experience judgment usually requires a large amount of manpower and time investment, which is less efficient for large-scale seedling growth management. In addition, since the prediction based on artificial experience is easily affected by individual experience and subjective judgment, the prediction results are usually subjective and uncertain, and the seedling growth height cannot be accurately predicted, and due to the inaccuracy of the prediction results, it may cause waste of resources and unnecessary cost expenditure.

[0041] In response to the above technical problems, the technical concept of this application is to use artificial intelligence technology based on deep learning to extract the characteristics of the growth status of seedlings from the growth status images of seedlings on each day, and perform multi-dimensional temporal change characteristics analysis on the growth status characteristics of seedlings on each day, and predict the growth height of seedlings one week later based on the multi-dimensional temporal characteristics of the growth status of seedlings. In this way, the automatic prediction of seedling growth can be achieved, labor costs can be saved, and a decision-making basis can be provided for the growth planning and management of seedlings.

[0042] In the above-mentioned municipal greening seedling growth prediction method based on big data, the step S110 obtains the seedling growth status images collected by the camera for multiple days within a predetermined time period. It should be understood that the seedling growth status images collected by the camera can reflect the external morphological characteristics of the seedlings, such as the growth of leaves, morphology, stems, etc., so as to obtain a description of the growth status of the seedlings. At the same time, considering that the growth of seedlings is a gradual process, by obtaining the seedling growth status images of multiple days within a continuous time period to observe and record the growth status of the seedlings at different time points, the dynamic change information of the seedling growth can be obtained, so as to have a more comprehensive understanding of the growth process of the seedlings, and then realize the comprehensive observation and analysis of the seedling growth status, and provide an accurate data basis for the prediction of seedling growth.

[0043] In the above-mentioned municipal greening seedling growth prediction method based on big data, the step S120 extracts growth state features from the seedling growth state images of multiple days within the predetermined time period to obtain a sequence of semantically focused feature vectors of the seedling growth state. It should be understood that by extracting growth state features from the seedling growth state images of multiple days within the predetermined time period, the seedling growth state features of each day, such as leaf color, branch density, etc., can be obtained, which helps to understand the growth trend and change pattern of the seedlings at different time points, and further provides data support for the analysis of the temporal change characteristics of the seedling growth state.

[0044] Figure 3 The flowchart is a sequence of seedling growth state semantic focus feature vectors for extracting growth state features from seedling growth state images of multiple days within the predetermined time period in the municipal greening seedling growth prediction method based on big data according to the embodiment of the present application. Figure 3 As shown, the step S120 includes: S121, respectively passing the seedling growth status images of multiple days within the predetermined time period through a seedling growth status feature extractor based on a convolutional neural network model to obtain a sequence of seedling growth status feature maps; S122, passing the sequence of seedling growth status feature maps through a significant feature extractor to obtain a sequence of seedling growth status semantic focus feature vectors.

[0045] Specifically, the step S121, the seedling growth state images of multiple days in the predetermined time period are respectively passed through the seedling growth state feature extractor based on the convolutional neural network model to obtain a sequence of seedling growth state feature maps. Those of ordinary skill in the art should know that convolutional neural network (CNN) is a deep learning model widely used in the field of computer vision. It has good feature extraction ability and can automatically extract useful features from raw image data. Compared with the traditional method of manually designing features, the convolutional neural network can automatically learn the feature representation suitable for the task according to the characteristics of the data and the task requirements, without manual intervention, thereby reducing the complexity of feature engineering and improving the effect of feature extraction. In the prediction of seedling growth state, by inputting the seedling growth state image into the feature extractor based on the convolutional neural network, the feature learning ability of the convolutional neural network can be used to extract the important features of the seedling growth state of each day, such as leaf morphology, growth height, stem thickness and other features. In addition, the convolutional neural network model is composed of multiple convolutional layers and pooling layers, and features of different abstract levels of the image can be extracted layer by layer. The lower convolutional layers can extract low-level features such as edges and textures of the image, while the higher convolutional layers can extract more abstract semantic features. This hierarchical feature extraction capability enables the seedling growth status feature extractor based on the convolutional neural network model to capture local details and global context information in the seedling growth status image, thereby better describing the growth status of the seedlings and providing a more accurate data basis for subsequent analysis and prediction.

[0046] Specifically, in step S122, the sequence of the seedling growth state feature map is passed through a significant feature extractor to obtain a sequence of the seedling growth state semantic focus feature vector. In the technical solution of the present application, considering that in the sequence of the seedling growth state feature map, each seedling growth state feature map may have background, noise and insignificant detail features, these irrelevant information may interfere with the subsequent seedling growth state analysis. Therefore, in order to reduce the interference of irrelevant noise and pay more attention to the growth state of the seedling itself, the sequence of the seedling growth state feature map is further passed through a significant feature extractor to obtain a sequence of the seedling growth state semantic focus feature vector. That is, further extract and focus on the features with the most semantic information and importance in the image, so as to better describe the growth state and changes of the seedling. It should be understandable that the significant feature extractor can calculate the significance or importance of each feature value in the seedling growth status feature map, and focus on the areas and features that are significant to the growth status, thereby filtering out irrelevant information and focusing on important parts of the seedlings such as leaves, branches, and morphology, capturing the important characteristics of the seedling growth status, and providing more accurate and targeted feature representation for the analysis and prediction of the seedling growth status.

[0047] In a specific example of the present application, the step S122 includes: extracting significant features from the seedling growth state feature map using the following significant feature extraction formula to obtain the seedling growth state semantic focus feature vector; wherein the significant feature extraction formula is:

[0048] V output =GAP{ReLU(Conv 1×1 [Con 3×3 (F input )])}

[0049] Among them, V output is the semantic focused feature vector of the seedling growth status, F input is the characteristic graph of the seedling growth status, GAP represents the pooling operation, ReLU represents the ReLU activation processing, Conv 1×1 (·) indicates a convolution operation based on a 1×1 convolution kernel. 3×3 (·) indicates a convolution operation based on a 3×3 convolution kernel.

[0050] Specifically, the salient feature extractor adds a 1×1 convolution kernel and a ReLU activation function after convolution. From the perspective of cross-channel pooling, this processing method is equivalent to implementing cascaded cross-channel weighted pooling on a convolution layer, so that the model can learn the relationship between channels and model local information more efficiently. In other words, the salient feature extractor performs local perception and modeling on each of the seedling growth status feature maps, thereby performing explicit processing on specific local areas, making the growth status features of these areas more prominent and obvious.

[0051] In the above-mentioned municipal greening seedling growth prediction method based on big data, the step S130 performs a multi-dimensional temporal change feature analysis on the sequence of the semantic focus feature vector of the seedling growth state to obtain a multi-dimensional feature vector of the seedling growth state in the temporal domain. It should be understood that the growth state of the seedling is a dynamic process. As time goes by, the morphology, color, density and other characteristics of the seedling will change. By analyzing the temporal change characteristics of the seedling growth state, the trend and pattern of seedling growth can be revealed, and important information such as the growth rate and growth cycle of the seedling can be understood, so as to predict the growth height of the seedling. In addition, by performing a multi-dimensional temporal change feature analysis on the growth state of the seedling, the temporal change characteristics of the growth state at different levels can be comprehensively considered to obtain a more comprehensive and accurate representation of the temporal change characteristics of the seedling growth state.

[0052] Figure 4 The present invention is a flowchart of performing a multi-dimensional temporal change feature analysis on the sequence of the semantic focus feature vector of the seedling growth state in the municipal greening seedling growth prediction method based on big data according to the embodiment of the present application to obtain the multi-dimensional feature vector of the seedling growth state in the temporal domain. Figure 4 As shown, the step S130 includes: S131, calculating the growth fluctuation semantic measurement coefficient between each two adjacent seedling growth state semantic focus feature vectors in the sequence of the seedling growth state semantic focus feature vectors to obtain a seedling growth state time series change feature vector composed of multiple growth fluctuation semantic measurement coefficients; S132, cascading the sequence of the seedling growth state semantic focus feature vectors to obtain the full time domain seedling growth state semantic feature vector; S133, fusing the seedling growth state time series change feature vector and the full time domain seedling growth state semantic feature vector to obtain the seedling growth state time domain multi-dimensional feature vector.

[0053] Specifically, the step S131 calculates the growth fluctuation semantic measurement coefficient between each two adjacent seedling growth state semantic focus feature vectors in the sequence of the seedling growth state semantic focus feature vector to obtain a seedling growth state time series change feature vector composed of multiple growth fluctuation semantic measurement coefficients. It should be understood that the growth state of the seedlings does not change monotonically and linearly, but has a certain volatility. These fluctuations may be caused by factors such as environmental conditions, maintenance measures, and growth stages. By calculating the growth fluctuation semantic measurement coefficient between two adjacent seedling growth state semantic focus feature vectors, the change trend and relationship between the seedling growth state semantic focus feature vectors at two adjacent time points can be captured, and the degree of fluctuation of the seedling growth state can be quantified, thereby more accurately describing the short-term fluctuation change characteristics of the seedling growth state.

[0054] In a specific example of the present application, step S131 includes: using the following growth fluctuation semantic measurement formula to calculate the growth fluctuation semantic measurement coefficient between each two adjacent seedling growth state semantic focus feature vectors in the sequence of the seedling growth state semantic focus feature vector to obtain a seedling growth state time series change feature vector composed of multiple growth fluctuation semantic measurement coefficients; wherein the growth fluctuation semantic measurement formula is:

[0055]

[0056] in, is the semantic focused feature vector of the seedling growth status mentioned above, is the semantic focus feature vector of the latter seedling growth state, N is the dimension of each semantic focus feature vector of the seedling growth state, H p is the growth fluctuation semantic measurement coefficient, and log represents the logarithmic function operation with base 2.

[0057] Specifically, the step S132, cascades the sequence of the seedling growth state semantic focus feature vector to obtain the full-time domain seedling growth state semantic feature vector. It should be understood that by cascading the individual seedling growth state semantic focus feature vectors in the sequence of the seedling growth state semantic focus feature vector, the seedling growth state information at all time points can be retained, the integrity of the timing information can be ensured, and the characteristic information of the seedling growth state in the full time domain can be captured, including long-term growth state changes, so as to more comprehensively understand the growth law and periodic changes of the seedling in the full time domain. In addition, the growth state of the seedling may have the characteristics of continuity and consistency at different time points. By cascading the sequence of the seedling growth state semantic focus feature vector, the features at adjacent time points can be connected to form a continuous feature representation, so as to better capture the continuity characteristics of the seedling growth state.

[0058] Specifically, the step S133 fuses the seedling growth state temporal change feature vector and the full time domain seedling growth state semantic feature vector to obtain the seedling growth state temporal multi-dimensional feature vector. In a specific example of the present application, the encoding method of fusing the seedling growth state temporal change feature vector and the full time domain seedling growth state semantic feature vector to obtain the seedling growth state temporal multi-dimensional feature vector is to use a projection layer to fuse the seedling growth state temporal change feature vector and the full time domain seedling growth state semantic feature vector to obtain the seedling growth state temporal multi-dimensional feature vector. It should be understood that the projection layer can map the input data to a new feature space and perform linear combination of the input data to achieve the fusion and interaction of input features. In the technical solution of the present application, a projection layer is used to fuse the seedling growth state temporal change feature vector and the full-time-domain seedling growth state semantic feature vector, which can comprehensively consider the short-term fluctuation change characteristics of the seedling growth state and the full-time-domain growth state change characteristics, and obtain a richer and more comprehensive seedling growth state temporal feature representation, thereby improving the accuracy of subsequent seedling growth predictions. Moreover, in the actual feature fusion process, due to the different encoding methods of the seedling growth state temporal change feature vector and the full-time-domain seedling growth state semantic feature vector, the two may have different dimensions. By using a projection layer to perform a fusion operation, the dimensions of the seedling growth state temporal change feature vector and the full-time-domain seedling growth state semantic feature vector can be unified to facilitate subsequent feature analysis and modeling.

[0059] In a specific example of the present application, a projection layer is used to fuse the seedling growth state temporal change feature vector and the full time domain seedling growth state semantic feature vector to obtain the seedling growth state temporal multi-dimensional feature vector, including: using the projection layer to fuse the seedling growth state temporal change feature vector and the full time domain seedling growth state semantic feature vector using the following projection feature fusion formula to obtain the seedling growth state temporal multi-dimensional feature vector; wherein the projection feature fusion formula is:

[0060]

[0061] Among them, V1 is the temporal change feature vector of the seedling growth state, V2 is the semantic feature vector of the seedling growth state in the entire time domain, and V f is the multi-dimensional feature vector of the seedling growth state in time domain, Represents projection fusion processing.

[0062] In the above-mentioned municipal greening seedling growth prediction method based on big data, the step S140 predicts the growth height of the seedlings one week later based on the multi-dimensional feature vector of the seedling growth state in the time domain. It should be understood that the multi-dimensional feature vector of the seedling growth state in the time domain contains information about the seedling growth state at different time points, reflecting the evolution process and trend of the seedling growth state. The multi-dimensional feature vector of the seedling growth state in the time domain is further processed by the machine model, and the time domain change characteristics of the seedling growth state in the multi-dimensional feature vector of the seedling growth state can be learned, and then the growth height of the seedlings in the future period of time can be predicted.

[0063] Figure 5 The flowchart is a method for predicting the growth height of a seedling one week later based on the multi-dimensional feature vector of the seedling growth state in the time domain according to the embodiment of the present application. Figure 5 As shown, the step S140 includes: S141, performing feature distribution correction on the time-domain multi-dimensional feature vector of the seedling growth state to obtain a corrected time-domain multi-dimensional feature vector of the seedling growth state; S142, passing the corrected time-domain multi-dimensional feature vector of the seedling growth state through a decoder-based seedling growth predictor to obtain a predicted value, and the predicted value is used to represent the predicted value of the growth height of the seedling one week later.

[0064] Specifically, the step S141 performs feature distribution correction on the time domain multi-dimensional feature vector of the seedling growth state to obtain the corrected time domain multi-dimensional feature vector of the seedling growth state. In the technical solution of the present application, each seedling growth state semantic focus feature vector in the sequence of the seedling growth state semantic focus feature vector expresses the image semantic features of the corresponding seedling growth state image, thereby, the seedling growth state temporal change feature vector expresses the temporal change features of the local time domain image semantic features under the full time domain, and the full time domain seedling growth state semantic feature vector expresses the temporal set features of the full time domain image semantic features, therefore, in addition to the feature expression dimension, there is also an expression difference in the image semantic features themselves between the seedling growth state temporal change feature vector and the full time domain seedling growth state semantic feature vector.

[0065] Therefore, considering the differences in image semantic feature expression between the seedling growth state temporal change feature vector and the full time domain seedling growth state semantic feature vector, it is expected that when using the projection layer to fuse the seedling growth state temporal change feature vector and the full time domain seedling growth state semantic feature vector, the high-dimensional geometric change differences between the seedling growth state temporal change feature vector and the full time domain seedling growth state semantic feature vector caused by the differences in image semantic feature expression can be suppressed based on the feature distribution directions corresponding to different geometric transformations in the high-dimensional feature space, thereby improving the fusion effect.

[0066] Based on this, in the technical solution of the present application, the image semantics of the seedling growth state temporal change feature vector and the full-time domain seedling growth state semantic feature vector are further fused and corrected by the following fusion correction formula to obtain a corrected feature vector; wherein the fusion correction formula is:

[0067]

[0068] V1 is the interpolation feature vector obtained by converting the seedling growth state time series change feature vector into the same length as the full-time seedling growth state semantic feature vector through linear interpolation, V2 is the full-time seedling growth state semantic feature vector, v 1i 、v 2i and v ci are the eigenvalues ​​of the interpolation feature vector V1, the full-time-domain seedling growth status semantic feature vector V2 and the correction feature vector, respectively. ||·||1 and ||·||2 are the 1-norm and 2-norm of the feature vector, respectively. L is the length of the feature vector, and α is the weight hyperparameter.

[0069] Specifically, in order to improve the perception and cognitive ability of features to spatial transformations corresponding to different feature distributions during fusion, the interpolated feature vector V1 after interpolation of the seedling growth state temporal change feature vector and the distance structure parameter of the full-time-domain seedling growth state semantic feature vector V2 are used to control the rotation of the feature distribution of the feature vector in different directions from the vector scale dimension, so as to retain the rotational invariance of the fused feature through relative rotational unwrapping, thereby avoiding the high-dimensional geometric change differences in the feature distribution caused by the geometric transformation in the high-dimensional feature space during feature fusion. In this way, the correction feature vector is fused with the seedling growth state time domain multi-dimensional feature vector, which can improve the fusion expression effect of the seedling growth state time domain multi-dimensional feature vector, thereby improving the accuracy of the decoded value obtained by the decoder.

[0070] Specifically, in step S142, the multi-dimensional feature vector of the time domain of the corrected seedling growth state is passed through a seedling growth predictor based on a decoder to obtain a predicted value, and the predicted value is used to represent the predicted value of the growth height of the seedling one week later. In the technical solution of the present application, the seedling growth predictor based on a decoder is a trained model for analyzing the evolution process and trend of the seedling growth state, learning the mapping relationship between the multi-dimensional feature vector of the time domain of the seedling growth state after correction and the growth height of the seedling one week later, so as to predict the future growth height. That is, by inputting the multi-dimensional feature vector of the time domain of the seedling growth state after correction into the seedling growth predictor based on a decoder, the prediction ability of the decoder model can be used to perform decoding regression, and the growth height of the seedling one week later can be predicted based on the learned feature patterns and laws, thereby providing a decision-making basis for the growth planning management of the seedlings.

[0071] In a specific example of the present application, step S142 includes: using the decoder to decode and regress the multi-dimensional feature vector of the corrected seedling growth state in the time domain using the following decoding formula to obtain the predicted value of the growth height of the seedling one week later, wherein the decoding formula is: Among them, X is the multi-dimensional feature vector of the corrected seedling growth state in the time domain, Y is the predicted value, W is the weight matrix, Represents matrix multiplication.

[0072] In summary, the municipal greening seedling growth prediction method based on big data according to the embodiment of the present application is explained, which uses artificial intelligence technology based on deep learning to extract seedling growth status features from the seedling growth status images of each day, and performs multi-dimensional growth status time domain change feature analysis on the seedling growth status features of each day, and predicts the growth height of the seedlings one week later based on the multi-dimensional characteristics of the seedling growth status time domain. In this way, the automatic prediction of seedling growth can be achieved, labor costs can be saved, and a decision-making basis can be provided for the growth planning and management of seedlings.

[0073] Figure 6 FIG. 1 is a block diagram of a municipal greening seedling growth prediction system based on big data according to an embodiment of the present application. Figure 6As shown, according to the municipal greening seedling growth prediction system 100 based on big data according to the embodiment of the present application, it includes: a seedling growth status monitoring module 110, which is used to obtain seedling growth status images of multiple days within a predetermined time period collected by a camera; a seedling growth status feature extraction module 120, which is used to extract growth status features of the seedling growth status images of multiple days within the predetermined time period respectively to obtain a sequence of seedling growth status semantic focus feature vectors; a seedling growth status time domain change feature analysis module 130, which is used to perform multi-dimensional time domain change feature analysis on the sequence of seedling growth status semantic focus feature vectors to obtain a seedling growth status time domain multi-dimensional feature vector; a seedling growth height prediction module 140, which is used to predict the growth height of the seedling one week later based on the seedling growth status time domain multi-dimensional feature vector.

[0074] Here, those skilled in the art can understand that the specific operations of each module in the above-mentioned municipal greening seedling growth prediction system based on big data have been referred to above. Figures 1 to 5 The description of the municipal greening seedling growth prediction method based on big data has been introduced in detail, and therefore, its repeated description will be omitted.

[0075] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, please refer to the relevant description of other embodiments. In the several embodiments provided by the present invention, it should be understood that the disclosed devices, systems and methods can be implemented in other ways. For example, the system embodiments described above are only schematic. For example, the module division is only a logical function division, and there may be other division methods in actual implementation. The modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0076] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0077] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference to a figure in a claim should not be considered as limiting the claim to which it relates.

[0078] In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. The multiple units stated in the system claims can also be implemented by one unit through software or hardware.

[0079] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution 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 solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A municipal greening seedling growth prediction method based on big data, characterized in that: include: Acquire the growth status images of the seedlings collected by the camera for multiple days within a predetermined time period; Extracting growth state features from the seedling growth state images of multiple days within the predetermined time period to obtain a sequence of seedling growth state semantic focus feature vectors; Performing a multi-dimensional time-domain change feature analysis on the sequence of the seedling growth state semantic focus feature vector to obtain a multi-dimensional time-domain feature vector of the seedling growth state; Based on the multi-dimensional feature vector of the seedling growth status in time domain, the growth height of the seedling one week later is predicted.

2. The municipal greening seedling growth prediction method based on big data according to claim 1 is characterized in that: Extracting growth state features from the seedling growth state images of multiple days within the predetermined time period to obtain a sequence of seedling growth state semantic focus feature vectors includes: The seedling growth state images of multiple days within the predetermined time period are respectively passed through a seedling growth state feature extractor based on a convolutional neural network model to obtain a sequence of seedling growth state feature maps; The sequence of the seedling growth state feature graphs is passed through a significant feature extractor to obtain a sequence of the seedling growth state semantic focus feature vectors.

3. The municipal greening seedling growth prediction method based on big data according to claim 2 is characterized in that: The sequence of the seedling growth state feature graphs is passed through a significant feature extractor to obtain a sequence of the seedling growth state semantic focus feature vectors, including: The salient features of the seedling growth state feature map are extracted using the following salient feature extraction formula to obtain the seedling growth state semantic focus feature vector; wherein the salient feature extraction formula is: V output =GAP{ReLU(Conv 1×1 [Conv. 3×3 (F input )])} Among them, V output is the semantic focused feature vector of the seedling growth status, F input is the characteristic graph of the seedling growth status, GAP represents the pooling operation, ReLU represents the ReLU activation processing, Conv 1×1 (·) indicates a convolution operation based on a 1×1 convolution kernel. 3×3 (·) indicates a convolution operation based on a 3×3 convolution kernel.

4. The municipal greening seedling growth prediction method based on big data according to claim 3 is characterized in that: A multi-dimensional temporal variation feature analysis is performed on the sequence of the semantic focus feature vectors of the seedling growth state to obtain a multi-dimensional temporal feature vector of the seedling growth state, including: Calculating the growth fluctuation semantic measurement coefficient between each two adjacent seedling growth state semantic focus feature vectors in the sequence of the seedling growth state semantic focus feature vectors to obtain a seedling growth state time series change feature vector composed of multiple growth fluctuation semantic measurement coefficients; Cascading the sequence of the seedling growth state semantic focus feature vectors to obtain the full-time domain seedling growth state semantic feature vector; The time-series change feature vector of the seedling growth state and the full-time-domain semantic feature vector of the seedling growth state are fused to obtain the time-domain multi-dimensional feature vector of the seedling growth state.

5. The municipal greening seedling growth prediction method based on big data according to claim 4 is characterized in that: Calculating the growth fluctuation semantic measurement coefficient between each two adjacent seedling growth state semantic focus feature vectors in the sequence of the seedling growth state semantic focus feature vectors to obtain a seedling growth state time series change feature vector composed of multiple growth fluctuation semantic measurement coefficients, including: The growth fluctuation semantic measurement coefficient between each two adjacent seedling growth state semantic focus feature vectors in the sequence of the seedling growth state semantic focus feature vectors is calculated using the following growth fluctuation semantic measurement formula to obtain a seedling growth state time series change feature vector composed of multiple growth fluctuation semantic measurement coefficients; The growth fluctuation semantic measurement formula is: in, is the semantic focused feature vector of the seedling growth status mentioned above, is the semantic focus feature vector of the latter seedling growth state, N is the dimension of each semantic focus feature vector of the seedling growth state, H p is the growth fluctuation semantic measurement coefficient, and log represents the logarithmic function operation with 2 as the base.

6. The municipal greening seedling growth prediction method based on big data according to claim 5 is characterized in that: The time-series change feature vector of the seedling growth state and the semantic feature vector of the seedling growth state in the entire time domain are integrated to obtain the time-domain multi-dimensional feature vector of the seedling growth state, including: A projection layer is used to fuse the temporal variation feature vector of the seedling growth state and the full-time-domain semantic feature vector of the seedling growth state to obtain the temporal multi-dimensional feature vector of the seedling growth state.

7. The municipal greening seedling growth prediction method based on big data according to claim 6 is characterized in that: Using the projection layer to fuse the seedling growth state temporal change feature vector and the full-time domain seedling growth state semantic feature vector to obtain the seedling growth state temporal multi-dimensional feature vector, including: The projection layer is used to fuse the seedling growth state temporal change feature vector and the full-time domain seedling growth state semantic feature vector using the following projection feature fusion formula to obtain the seedling growth state temporal multi-dimensional feature vector; wherein the projection feature fusion formula is: Among them, V1 is the temporal change feature vector of the seedling growth state, V2 is the semantic feature vector of the seedling growth state in the entire time domain, and V f is the multi-dimensional feature vector of the seedling growth state in time domain, Represents projection fusion processing.

8. The municipal greening seedling growth prediction method based on big data according to claim 7 is characterized in that: Based on the multi-dimensional feature vector of the seedling growth state in the time domain, the growth height of the seedlings one week later is predicted, including: Performing feature distribution correction on the time-domain multi-dimensional feature vector of the seedling growth state to obtain a corrected time-domain multi-dimensional feature vector of the seedling growth state; The corrected multi-dimensional feature vector of the seedling growth state in the time domain is passed through a decoder-based seedling growth predictor to obtain a predicted value, and the predicted value is used to represent the predicted value of the growth height of the seedling one week later.

9. The municipal greening seedling growth prediction method based on big data according to claim 8 is characterized in that: The corrected seedling growth state time domain multi-dimensional feature vector is passed through a decoder-based seedling growth predictor to obtain a predicted value, including: The decoder is used to decode and regress the corrected multi-dimensional feature vector of the seedling growth state in the time domain using the following decoding formula to obtain the predicted value of the growth height of the seedling one week later, wherein the decoding formula is: Among them, X is the multi-dimensional feature vector of the corrected seedling growth state in the time domain, Y is the predicted value, W is the weight matrix, Represents matrix multiplication.

10. A municipal greening seedling growth prediction system based on big data, characterized in that: include: A seedling growth status monitoring module is used to obtain the seedling growth status images collected by the camera for multiple days within a predetermined time period; A seedling growth state feature extraction module is used to extract growth state features from the seedling growth state images of multiple days within the predetermined time period to obtain a sequence of seedling growth state semantic focus feature vectors; A seedling growth state temporal variation feature analysis module, used for performing a multi-dimensional temporal variation feature analysis on the sequence of the seedling growth state semantic focus feature vectors to obtain a seedling growth state temporal multi-dimensional feature vector; The seedling growth height prediction module is used to predict the growth height of the seedlings one week later based on the multi-dimensional feature vector of the seedling growth state in the time domain.

Images