A forage harvesting system based on a growth state prediction model

Abstract

The present invention relates to the technical field of data processing, and in particular to a forage harvesting system based on a growth state prediction model. The system includes: a collection module, a preliminary screening module, a determination module, a determination module, an adjustment module, a prediction module, a selection module, and a harvesting module. By collecting key data, the present invention screens out high-quality forage areas; through preliminary screening of plant height and lesion area, it quickly locks in areas with heights meeting requirements and fewer pests and diseases; then, it combines the lesion area and tiller number to determine healthy and well-growing marked squares; further, it determines the attention squares with high biomass and excellent quality according to the tiller number and leaf area; adjusts the square width according to the aggregation degree of the attention squares in time and space; finally, combines the attention squares and the predicted key squares to accurately select the harvesting area, effectively solving the problems of uneven forage quality and low harvesting efficiency caused by blind harvesting and over-reliance on static models.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly to a forage harvesting system based on a growth state prediction model. Background Art

[0002] With the continuous growth of the global population and the continuous improvement of living standards, the demand for livestock products is increasing day by day, and the importance of the livestock industry is becoming more and more prominent. As the basis of the livestock industry, the demand for high-quality forage has also increased significantly. High-quality forage can not only significantly improve the growth rate of livestock and the quality of livestock products such as meat and milk, but also enhance the disease resistance of livestock, reduce the occurrence of diseases, lower the breeding cost, and improve the breeding efficiency. In modern livestock breeding, large-scale and intensive breeding models are gradually popularized, which puts forward higher requirements for the stable supply and precise management of high-quality forage. At the same time, consumers' attention to the quality and safety of livestock products is also constantly increasing, requiring that forage be effectively monitored throughout the growth and harvesting process to ensure that its quality meets the standards, thereby guaranteeing the safety and health of the final livestock products.

[0003] The patent document with the publication number CN116952853A discloses an agricultural monitoring system for forage quality monitoring. The system includes: an edge system, including one or more sensors for collecting environmental data and / or forage image data; a side system, including a side controller configured to perform real-time monitoring and analysis of forage quality based on the environmental data and / or forage image data collected by the edge system to obtain forage quality data; and a cloud system, including a data aggregation center for storing and managing the environmental data and / or the forage image data, the forage quality data, edge system data and side system data.

[0004] It can be seen that the agricultural monitoring system for forage quality monitoring has the following problems: when the monitoring area is large and the data volume is huge, it is impossible to ensure the timely update of forage quality data; although there are sensors, the accurate recognition and real-time adjustment capabilities in complex environments need to be improved; the in-depth analysis and application of data are insufficient, and the data value cannot be fully exploited to provide more valuable decision-making support for forage planting and harvesting; during the harvesting process, the quality indicators of forage cannot be monitored in real time, and the harvesting parameters cannot be dynamically adjusted according to these indicators, making it difficult to guarantee the quality of the harvested forage and unable to meet the personalized needs of different users for high-quality forage. Summary of the Invention

[0005] Therefore, the present invention provides a forage harvesting system based on a growth state prediction model to overcome the problems of uneven forage quality and low harvesting efficiency caused by blind harvesting and excessive reliance on static models in the prior art through real-time multi-dimensional data, a growth state prediction model, and a dynamic adjustment mechanism.

[0006] To achieve the above object, the present invention provides a forage harvesting system based on a growth state prediction model, comprising:

[0007] A collection module is used to obtain in real time the plant height, diseased spot area, tiller number and leaf area of the forage grass in each square to be tested based on a preset square width;

[0008] A primary screening module, which is connected to the collection module and is used to obtain a number of temporary grids by primary screening according to the plant height, the diseased spot area and a preset plant height range;

[0009] A determination module, which is connected to the collection module and the primary screening module respectively, and is used to determine a number of marked squares according to the diseased spot area and the tillering number of each temporary square;

[0010] A determination module, which is connected to the acquisition module and the determination module respectively, and is used to determine a number of focus squares according to the tillering number and the leaf area of each marked square;

[0011] An adjustment module, connected to the determination module, for adjusting the preset square width according to the position of each of the focus squares to obtain an adjusted square width;

[0012] A prediction module, which is connected to the collection module and the adjustment module respectively, and is used to input the plant height, the diseased spot area, the tiller number and the leaf area collected based on the adjusted grid width into a preset growth state prediction model to predict a number of key grids;

[0013] A selection module, which is connected to the determination module and the prediction module respectively, and is used to select a number of harvesting squares according to the key squares and the concerned squares determined based on the adjustment square width;

[0014] A harvesting module is connected to the selection module and is used for harvesting the grass in each harvesting grid.

[0015] Furthermore, the determination module includes:

[0016] A normalization processing unit, used to perform normalization processing on all the diseased spot areas within a preset determination time to obtain a diseased spot normalized data set, and to perform normalization processing on all the tillering numbers within a preset determination time to obtain a tillering normalized data set;

[0017] A determination unit is connected to the normalization processing unit and is used to determine a number of the marked squares according to the lesion normalization data set and the tillering normalization data set.

[0018] Furthermore, the determination unit includes:

[0019] A relevance calculation subunit, which is used to calculate the correlation coefficient between the diseased spot normalization dataset and the tiller normalization dataset to obtain a change relevance;

[0020] A determination subunit, which is connected to the relevance calculation subunit and is used to determine the temporary grid as the marked grid when the change relevance is less than a preset relevance threshold, so as to determine a number of marked grids.

[0021] Further, the determination module includes:

[0022] A tiller fluctuation value calculation unit, which is used to calculate the standard deviation of the tiller numbers at each moment from the initial moment to a preset determination duration to obtain a number of tiller fluctuation values;

[0023] A leaf area fluctuation value calculation unit, which is used to calculate the standard deviation of the leaf areas at each moment from the initial moment to the preset determination duration to obtain a number of leaf area fluctuation values;

[0024] A determination unit, which is respectively connected to the tiller fluctuation value calculation unit and the leaf area fluctuation value calculation unit, and is used to determine a number of concerned grids according to all the tiller fluctuation values and all the leaf area fluctuation values.

[0025] Further, the determination unit includes:

[0026] A tiller change curve drawing subunit, which is used to draw a curve of the change of all the tiller fluctuation values with time within the preset determination duration to obtain a tiller change curve;

[0027] A leaf area change curve drawing subunit, which is used to draw a curve of the change of all the leaf area fluctuation values with time within the preset determination duration to obtain a leaf area change curve;

[0028] A synchronization degree calculation subunit, which is respectively connected to the tiller change curve drawing subunit and the leaf area change curve drawing subunit, and is used to calculate the cosine similarity between the tiller change curve and the leaf area change curve to obtain a change synchronization degree;

[0029] A determination subunit, which is connected to the synchronization degree calculation subunit and is used to determine the marked grid as the concerned grid when the change synchronization degree is greater than a preset synchronization degree threshold, so as to determine a number of concerned grids.

[0030] Further, the adjustment module includes:

[0031] A dispersion calculation unit, which is used to calculate the standard deviation of the positions of each concerned grid to obtain a dispersion;

[0032] A dispersion deviation calculation unit, which is connected to the dispersion calculation unit and is used to calculate the relative deviation between the dispersion and a preset dispersion threshold when the dispersion is greater than the preset dispersion threshold, so as to obtain a dispersion deviation;

[0033] An adjustment unit, which is connected to the dispersion deviation calculation unit and is used to adjust the preset grid width according to the dispersion deviation and the preset dispersion deviation threshold when the dispersion deviation is greater than the preset dispersion deviation threshold, so as to obtain an adjusted grid width.

[0034] Further, the adjustment unit includes:

[0035] A uniformity index calculation sub-unit, which is used to mark all timestamps when the dispersion deviation is greater than the preset dispersion deviation threshold within a preset adjustment duration. When the number of all timestamps is greater than a preset number threshold, calculate the time intervals between all adjacent timestamps, and calculate the ratio of the standard deviation to the mean of the time intervals to obtain a uniformity index;

[0036] An adjustment factor calculation sub-unit, which is connected to the uniformity index calculation sub-unit and is used to calculate the relative deviation between the dispersion deviation and the preset dispersion deviation threshold when the uniformity index is less than the preset uniformity index threshold, so as to obtain an adjustment factor;

[0037] An adjustment sub-unit, which is connected to the adjustment factor calculation sub-unit and is used to increase the preset grid width according to the adjustment factor and a preset adjustment coefficient to obtain an adjusted grid width.

[0038] Further, the selection module includes:

[0039] A ratio calculation unit, which is used to calculate the ratio of the number of the key grids to the number of the concerned grids to obtain a quantity ratio;

[0040] A coincidence degree calculation unit, which is connected to the ratio calculation unit and is used to count the number of overlapping grids among all the key grids and all the concerned grids to obtain an overlapping quantity when the quantity ratio is greater than the minimum value of a preset ratio range and less than the maximum value of the preset ratio range. When the number of the key grids is greater than the number of the concerned grids, calculate the ratio of the overlapping quantity to the number of the key grids to obtain a coincidence degree, or when the number of the key grids is less than the number of the concerned grids, calculate the ratio of the overlapping quantity to the number of the concerned grids to obtain a coincidence degree;

[0041] A selection unit, which is connected to the coincidence degree calculation unit and is used to determine that the key grid is the harvesting grid when the coincidence degree is greater than the preset coincidence degree threshold, so as to select several harvesting grids.

[0042] Further, the preliminary screening module includes:

[0043] A height comparison unit for comparing the plant height and the preset plant height range to obtain a height comparison result;

[0044] A primary screening unit, connected to the height comparison unit, for initially screening a number of temporary squares according to the lesion area when the height comparison result is that the plant height is greater than the minimum value of the preset plant height range and less than the maximum value of the preset plant height range.

[0045] Further, the primary screening unit includes:

[0046] A lesion comparison unit for comparing the lesion area and a preset lesion area threshold to obtain a lesion comparison result;

[0047] A primary screening sub-unit, connected to the lesion comparison unit, for determining the square to be tested as the temporary square when the lesion comparison result is that the lesion area is less than the preset lesion area threshold, so as to initially screen a number of temporary squares.

[0048] Compared with the prior art, the beneficial effect of the present invention is to screen out high-quality forage areas by collecting key data; first, initially screen through plant height and lesion area to quickly lock in areas with height meeting requirements and fewer pests and diseases; then, combine lesion area and tiller number to determine healthy and well-growing marked squares. Next, determine the attention squares with high biomass and good quality according to tiller number and leaf area; adjust the square width according to the aggregation degree of attention squares in time and space to optimize the harvesting plan; finally, accurately select the harvesting area by combining the key squares and attention squares predicted by the growth state prediction model, effectively solving the problems of uneven forage quality and low harvesting efficiency caused by blind harvesting and over-reliance on static models.

[0049] Further, by calculating the correlation coefficient of the two normalized data sets, their correlation can be quantified. When determining the marked squares based on the correlation coefficient, only the areas with weak correlation between lesion area and tiller number are retained, which can screen out areas with stable forage growth and little influence of lesions, and can effectively screen out the marked squares meeting the requirements to ensure forage quality.

[0050] Further, by calculating the correlation coefficient of the normalized lesion data set and tiller data set, the strength of the linear relationship between the two can be quantified. When the change correlation degree is less than the preset correlation degree threshold, it indicates that there is a weak correlation between the lesion area and the tiller number. Such a logical association means that the appearance of lesions has little influence on tillering, so the temporary square can be determined as the marked square, which can effectively screen out areas with stable forage growth and less influence of lesions.

[0051] Furthermore, by calculating the standard deviation of the tiller number and leaf area within a preset determination duration, the fluctuations of the tiller number and leaf area can be quantified. The fluctuation of the tiller number reflects the stability of the tillering growth of the forage grass, while the fluctuation of the leaf area reflects the change in the photosynthetic capacity of the forage grass. Both of them reflect the health status and growth trend of the forage grass during the growth process. Combining these two fluctuation values can more comprehensively evaluate the growth state of the forage grass.

[0052] Furthermore, by plotting the curves of the tiller fluctuation value and leaf area fluctuation value over time and calculating the cosine similarity of the two curves, the synchronization degree between the tiller change and leaf area change can be quantified. The tiller number and leaf area are both key indicators for measuring the growth status of the forage grass. Their synchronous changes usually reflect the coordination in aspects such as nutrient utilization efficiency and photosynthetic capacity during the growth process of the forage grass. When the change synchronization degree is greater than the preset synchronization degree threshold, it indicates that the forage grass in the marked grid has good consistency in tillering growth and leaf expansion. This often means that the forage grass grows healthily and has excellent quality, meeting the growth characteristics of high-quality forage grass. Therefore, it is determined as a concerned grid.

[0053] Furthermore, the dispersion degree is obtained by calculating the standard deviation of the positions of the concerned grids. The smaller the standard deviation, the more concentrated the concerned grids; conversely, the larger the standard deviation, the more dispersed they are. When the dispersion degree is greater than the preset dispersion degree threshold, then calculate the relative deviation between the dispersion degree and the preset dispersion degree threshold. When the dispersion degree deviation is greater than the preset dispersion degree deviation threshold, it indicates that the concerned grids are relatively dispersed. At this time, adjust the preset grid width and increase the width to more finely divide the area and improve the accuracy of the harvesting plan.

[0054] Furthermore, by marking the relevant timestamps within a preset adjustment duration and calculating the uniformity index of the time intervals when the quantity reaches the standard, the time regularity of the occurrence of the dispersion degree deviation can be accurately evaluated. When the uniformity index is less than the preset threshold, it indicates that the time intervals are relatively uniform, meaning that the occurrence time distribution of the situation where the dispersion degree deviation exceeds the preset threshold within the preset adjustment duration is regular. This implies that the non-uniformity of the forage grass distribution shows a certain stability in time and at the same time reflects the stability of the system's monitoring of the forage grass distribution, indicating that the system can reliably capture the signal of the forage grass distribution change. Calculating the relative deviation between the dispersion degree deviation and the preset threshold as the adjustment factor can quantify the adjustment amplitude. Increasing the preset grid width according to the adjustment factor and the preset adjustment coefficient can expand the harvesting range, make the forage grass distribution in each grid more uniform, and reduce the harvesting problems caused by uneven distribution.

[0055] Furthermore, by comprehensively considering the quantity ratio and the degree of overlap, first judge whether the quantity ratio is within a reasonable range. If it exceeds the range, directly determine that the concerned grid is a harvesting grid, which can quickly screen out the situations that obviously do not meet the quantity requirements. If the quantity ratio is within a reasonable range, then further judge by the degree of overlap. When the number of key grids and concerned grids is different, different calculation methods are adopted to ensure the accuracy of the degree of overlap, thus more objectively reflecting the overlapping situation between the two, improving the accuracy and reliability of determining the harvesting grid, and avoiding misjudgment. It reduces human intervention and errors, improves the efficiency and quality of selecting the harvesting grid, ensures the objectivity and accuracy of the data, and provides strong support for subsequent harvesting operations.

[0056] Furthermore, through a rough screening by comparing the plant heights, the plant height reflects the growth stage and development of the forage grass. Plants that are too tall or too short may have problems of over-ripeness or poor growth respectively, affecting the quality and nutritional value. The diseased spot area reflects the health status of the forage grass. Too many diseased spots will reduce the quality and yield. Therefore, first conduct a preliminary screening by height, and then conduct a secondary screening in combination with the diseased spot area, which can accurately lock in the forage grass area that meets the requirements, improve the harvesting efficiency, and ensure the forage grass quality.

[0057] Furthermore, by screening the diseased spot area, the diseased spot area is a key indicator for evaluating the health status of the forage grass. An overly large diseased spot area usually indicates that the forage grass is invaded by diseases, reducing the forage grass quality. When the diseased spot area is less than the preset diseased spot area threshold, the forage grass significantly affected by diseases can be effectively excluded, and the forage grass area that meets the height and health requirements can be efficiently screened out, improving the harvesting efficiency and the forage grass quality. Description of the Drawings

[0058] Figure 1 It is a schematic diagram of the forage grass harvesting system based on the growth state prediction model in this embodiment;

[0059] Figure 2 It is the determination logic diagram of the determination unit for determining the marked grid in this embodiment;

[0060] Figure 3 It is the determination logic diagram for determining the concerned grid in this embodiment;

[0061] Figure 4 It is the determination logic diagram of the preliminary screening unit for obtaining the temporary grid through preliminary screening in this embodiment. Detailed Embodiment

[0062] In order to make the purpose and advantages of the present invention clearer, the present invention will be further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0063] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0064] See also Figure 1 As shown, it is a schematic diagram of a forage harvesting system based on a growth state prediction model of this embodiment;

[0065] This embodiment provides a forage harvesting system based on a growth state prediction model, including:

[0066] A collection module is used to obtain in real time the plant height, diseased spot area, tiller number and leaf area of the forage grass in each square to be tested based on a preset square width;

[0067] A primary screening module, which is connected to the collection module and is used to obtain a number of temporary grids by primary screening according to the plant height, the diseased spot area and a preset plant height range;

[0068] A determination module, which is connected to the collection module and the primary screening module respectively, and is used to determine a number of marked squares according to the diseased spot area and the tillering number of each temporary square;

[0069] A determination module, which is connected to the acquisition module and the determination module respectively, and is used to determine a number of focus squares according to the tillering number and the leaf area of each marked square;

[0070] An adjustment module, connected to the determination module, for adjusting the preset square width according to the position of each of the focus squares to obtain an adjusted square width;

[0071] A prediction module, which is connected to the collection module and the adjustment module respectively, and is used to input the plant height, the diseased spot area, the tiller number and the leaf area collected based on the adjusted grid width into a preset growth state prediction model to predict a number of key grids;

[0072] A selection module, which is connected to the determination module and the prediction module respectively, and is used to select a number of harvesting squares according to the key squares and the concerned squares determined based on the adjustment square width;

[0073] A harvesting module is connected to the selection module and is used for harvesting the grass in each harvesting grid.

[0074] The acquisition module obtains data in the forage field operation scenario in real time. "Each square to be measured" refers to multiple square areas divided in the pasture land based on a preset square width. The forage data within each area will be collected and analyzed separately. These squares are partitions of the entire pasture land, enabling a more detailed understanding of the growth conditions of forage in different areas and facilitating subsequent precise management and harvesting decisions. The plant height reflects the growth stage and maturity of the forage, which is obtained through a laser ranging sensor; the lesion area shows the degree of damage to the forage by diseases, which is identified using image recognition technology; the tiller number reflects the growth density and growth trend of the forage, which is counted with the help of computer vision algorithms; the leaf area refers to the area size of the forage leaves, which reflects the photosynthesis ability, growth potential, and potential biomass accumulation of the forage, and is identified using image recognition technology.

[0075] The preset square width refers to the width of each square preset when dividing the squares in the pasture land. It depends on the forage variety, the area and terrain of the pasture land, and is usually set between 1 meter and 10 meters. In this embodiment, it is set to 5 meters, which can not only ensure the representativeness of data collection but also facilitate actual operation and management. At the same time, it can also adapt to the growth characteristics and field management requirements of most forages.

[0076] The preset plant height range refers to the height range used to screen forages that meet the harvesting conditions. It depends on the forage variety, growth stage, and harvesting purpose, and is usually set between 30 centimeters and 80 centimeters. In this embodiment, it is set to 40 centimeters to 70 centimeters to ensure that the forage has reached an appropriate maturity, while avoiding affecting the quality and harvesting efficiency due to the plant being too tall or too short.

[0077] The preset growth state prediction model is constructed by using a hybrid neural network architecture that combines a long short-term memory network (LSTM) and a convolutional neural network (CNN). The CNN is used to extract the spatial features of multi-dimensional data such as the plant height, lesion area, tiller number, and leaf area of the forage plants, and then the feature sequence extracted by the CNN is input into the LSTM layer. The LSTM captures the dynamic change trend of the data according to the time series characteristics, such as analyzing the increase and decrease law of the tiller number over time. Finally, the features output by the LSTM are integrated through a fully connected layer and mapped to the classification space of the key squares.

[0078] 1. Initial parameters

[0079] 1.1 CNN layer:

[0080] Convolution kernel size: Initially set to 3×3;

[0081] Stride: 1;

[0082] Padding method: "same", so as to keep the relative stability of the input and output sizes when extracting features;

[0083] Initial learning rate: 0.001;

[0084] The weights of the convolutional kernels are initialized using the Xavier normal distribution with random initialization;

[0085] The initial value of the bias term is set to: 0.

[0086] 1.2 LSTM layer:

[0087] The initial number of hidden layer units is: 64, which can balance the fitting ability and computational complexity of the model;

[0088] The weight matrix is initialized using the Xavier normal distribution with random initialization;

[0089] Initial learning rate: 0.001;

[0090] The initial value of the bias term of the forget gate is: 1, which helps the LSTM to better retain memory information in the initial stage of training.

[0091] 1.3 Fully connected layer:

[0092] The initial value of the number of neurons is: 128;

[0093] Initial learning rate: 0.001;

[0094] The weight matrix uses the He normal distribution with random initialization;

[0095] The initial value of the bias term; 0.

[0096] 2. Training process

[0097] Data preparation: Collect historical forage growth data, construct a training set with data such as plant height, lesion area, tiller number, and leaf area collected according to the adjusted grid width, label whether the sample is a key grid according to the forage growth law and expert experience, and divide the training set, validation set, and test set according to the ratio of 7:2:1.

[0098] Forward propagation and loss calculation: Input the training data into the model. After the CNN extracts the spatial features, the feature sequence enters the LSTM to capture the time features, and finally the fully connected layer outputs the prediction probability of the key grid. The cross-entropy loss function is used to measure the difference between the prediction probability and the true label, and the loss value is calculated.

[0099] Backward propagation and optimization: Use the Adam optimization algorithm to update the model parameters by backward propagation according to the calculated loss value. During the training cycle (100 epochs), adjust the learning rate according to the loss change of the validation set. If the validation set loss does not decrease for 5 consecutive epochs, multiply the learning rate by 0.1 to promote the model to gradually converge.

[0100] Regularization and early stopping: To prevent overfitting, L2 regularization terms are added to the CNN and LSTM layers, and the regularization coefficient is set to 0.0001. At the same time, an early stopping mechanism is set. When the validation set loss does not decrease for 10 consecutive epochs, training is stopped and the model parameters with the best performance on the validation set are saved.

[0101] 3. Training the model

[0102] After training is completed, the model solidifies the weights and bias parameters of each layer.

[0103] The CNN layer can accurately extract the spatial features of data such as plant height. For example, it can identify the edge features of lesions in lesion area images.

[0104] The LSTM layer grasps the dynamic characteristics of the data in the time dimension, such as knowing that the number of tillers first increases and then stabilizes over time;

[0105] The fully connected layer can accurately map these fused features to the classification results of the key squares. For input data such as plant height, it can stably output the probability value of each square being a key square.

[0106] 4. Final output

[0107] For the plant height, spot area, tiller number and leaf area data collected based on adjusting the grid width, the model ultimately outputs the probability value of each square to be tested as a key square. By setting the probability threshold (0.5), the squares with a probability greater than the threshold are judged as key squares, thereby achieving accurate prediction of the key squares.

[0108] By acquiring the plant height, disease spot area, tiller number and leaf area of the forage grass in real time. Then, a preliminary screening is performed based on the collected plant height and the preset plant height range to obtain several temporary grids. According to the disease spot area and tiller number of the temporary grids, several marked grids are further determined. Combined with the tiller number and leaf area of the marked grids, several focus grids are determined. According to the position of the focus grid, the preset grid width is adjusted to obtain the adjusted grid width. Based on the plant height, disease spot area, tiller number and leaf area collected based on the adjusted grid width, the preset growth state prediction model is input to predict several key grids. According to the key grids and the focus grids determined based on the adjusted grid width, several harvesting grids are selected. Finally, the forage grass in the harvesting grids is harvested.

[0109] By collecting key data, high-quality forage areas are screened out. First, through a preliminary screening based on plant height and lesion area, areas that meet the height requirements and have fewer pests and diseases are quickly locked in. Then, by combining the lesion area and the tiller number, marked squares that are healthy and growing well are determined. Next, squares with high biomass and excellent quality are identified based on the tiller number and leaf area. The width of the squares is adjusted according to the degree of aggregation of the squares in terms of time and space, optimizing the harvesting plan. Finally, by combining the key squares and the squares of concern predicted by the growth state prediction model, the harvesting area is accurately selected, effectively solving the problems of uneven forage quality and low harvesting efficiency caused by blind harvesting and over-reliance on static models.

[0110] Specifically, the determination module includes:

[0111] A normalization processing unit for normalizing all the lesion areas within a preset determination duration to obtain a lesion normalization data set, and for normalizing all the tiller numbers within the preset determination duration to obtain a tiller normalization data set;

[0112] A determination unit, which is connected to the normalization processing unit, for determining a number of the marked squares according to the lesion normalization data set and the tiller normalization data set.

[0113] The preset determination duration is a time period preset for determining the growth state of the forage, depending on the growth speed, growth cycle of the forage, as well as the frequency and precision requirements of the harvesting operation, and is usually set between 24 hours and 7 days. In this embodiment, it is set to 3 days, which can not only ensure that enough data samples are collected to reflect the change trend of the forage growth state, but also will not miss the critical period of forage growth due to too long a duration.

[0114] By separately normalizing all the lesion areas and all the tiller numbers within the preset determination duration, a lesion normalization data set and a tiller normalization data set are obtained, that is, the lesion area and the tiller number are respectively scaled between 0 and 1 to eliminate the dimensional and order-of-magnitude differences between different data, which is prior art and will not be elaborated here. Then, a number of marked squares are determined based on these two normalization data sets.

[0115] By calculating the correlation coefficient of the two normalization data sets, their correlation can be quantified. When determining the marked squares based on the correlation coefficient, only the areas with weak correlation between the lesion area and the tiller number are retained, which can screen out areas where the forage grows stably and the impact of lesions is small, and can effectively screen out the marked squares that meet the requirements to ensure the forage quality.

[0116] Please continue to refer to Figure 2 as shown, which is the determination logic diagram for the determination unit of this embodiment to determine the marked squares;

[0117] The determination unit includes:

[0118] A relevance calculation subunit, which is used to calculate the correlation coefficient between the diseased spot normalized dataset and the tiller normalized dataset to obtain a change relevance;

[0119] A determination subunit, which is connected to the relevance calculation subunit and is used to determine that the temporary grid is the marked grid when the change relevance is less than a preset relevance threshold, so as to determine a number of marked grids.

[0120] The preset relevance threshold is a reference value for judging the correlation between the diseased spot area and the tiller number, which depends on factors such as the growth characteristics of the forage variety and the law of disease influence, and is usually set between -1 and 1. In this embodiment, it is set to 0.5, which can effectively screen out the temporary grids with weak correlation between the diseased spot area and the tiller number, and ensure that the growth conditions of the forage in the marked grids are more stable and reliable.

[0121] By calculating the correlation coefficient between the diseased spot normalized dataset and the tiller normalized dataset, the change relevance is obtained. If the change relevance is less than the preset relevance threshold, it is determined that the corresponding temporary grid is the marked grid, so as to gradually screen out a number of marked grids that meet the requirements.

[0122] By calculating the correlation coefficient of the normalized diseased spot dataset and the tiller dataset, the strength of the linear relationship between the two can be quantified. When the change relevance is less than the preset relevance threshold, it indicates that there is a weak correlation between the diseased spot area and the tiller number. Such a logical association means that the appearance of the diseased spot has little impact on the tillering, so it can be determined that the temporary grid is the marked grid, which can effectively screen out the areas where the forage grows stably and is less affected by the diseased spot.

[0123] Specifically, the determination module includes:

[0124] A tiller fluctuation value calculation unit, which is used to calculate the standard deviation of the tiller numbers at each moment from the initial moment to the preset determination duration to obtain a number of tiller fluctuation values;

[0125] A leaf area fluctuation value calculation unit, which is used to calculate the standard deviation of the leaf areas at each moment from the initial moment to the preset determination duration to obtain a number of leaf area fluctuation values;

[0126] A determination unit, which is respectively connected to the tiller fluctuation value calculation unit and the leaf area fluctuation value calculation unit, and is used to determine a number of concerned grids according to all the tiller fluctuation values and all the leaf area fluctuation values.

[0127] The preset determination duration is the time period used to calculate the change between the tiller number and the leaf area, which depends on the stability of forage growth and the precision requirements of the harvesting operation. It is usually set between 3 days and 10 days. In this embodiment, it is set to 5 days, which can ensure the representativeness of the data while timely reflecting the fluctuations in forage growth.

[0128] By calculating the standard deviation of the tiller numbers at each moment from the initial moment to the preset determination duration, a number of tiller fluctuation values are obtained; then, by calculating the standard deviation of the leaf areas at each moment from the initial moment to the preset determination duration, a number of leaf area fluctuation values are obtained; finally, a number of concerned grids are determined based on all the tiller fluctuation values and leaf area fluctuation values.

[0129] By calculating the standard deviation of the tiller number and the leaf area within the preset determination duration, the fluctuations of the tiller number and the leaf area can be quantified. The fluctuation of the tiller number reflects the stability of forage tillering growth, while the fluctuation of the leaf area reflects the change in the photosynthesis ability of the forage. Both reflect the health status and growth trend of the forage during the growth process. Combining these two fluctuation values can more comprehensively evaluate the growth state of the forage.

[0130] Please continue to refer to Figure 3 as shown, which is the decision logic diagram for the determination unit in this embodiment to determine the concerned grids;

[0131] The determination unit includes:

[0132] A tiller change curve drawing subunit, used to draw the curve of the change of all the tiller fluctuation values within the preset determination duration over time, to obtain a tiller change curve;

[0133] A leaf area change curve drawing subunit, used to draw the curve of the change of all the leaf area fluctuation values within the preset determination duration over time, to obtain a leaf area change curve;

[0134] A synchronization degree calculation subunit, which is respectively connected to the tiller change curve drawing subunit and the leaf area change curve drawing subunit, used to calculate the cosine similarity of the tiller change curve and the leaf area change curve to obtain a change synchronization degree;

[0135] A determination subunit, which is connected to the synchronization degree calculation subunit, used to determine the marked grid as the concerned grid when the change synchronization degree is greater than the preset synchronization degree threshold, so as to determine a number of concerned grids.

[0136] The preset synchronization degree threshold is a standard value used to determine whether the synchronization of the tiller change curve and the leaf area change curve meets the standard. It depends on the growth characteristics of forage grass, the data collection accuracy, and the requirements for synchronization in the harvesting decision, and is usually set between 0.7 and 0.9. In this embodiment, it is set to 0.8, which can effectively screen out the grids with a relatively high synchronization of tiller number and leaf area change as the concerned grids while ensuring the judgment accuracy.

[0137] By relying on all the tiller fluctuation values and their corresponding time points within the preset determined time period, a curve that can intuitively show the trend of tiller fluctuation over time is drawn, that is, the tiller change curve; similarly, the leaf area change curve is drawn. Subsequently, vectorization processing is respectively performed on these two curves to obtain the corresponding vector forms, which is prior art and will not be elaborated here; then, the cosine similarity between the tiller change curve vector and the leaf area change curve vector is calculated to obtain the change synchronization degree; finally, when the change synchronization degree exceeds the preset synchronization degree threshold, it is determined that this marked grid is a concerned grid to determine several concerned grids.

[0138] By drawing the curves of tiller fluctuation values and leaf area fluctuation values over time and calculating the cosine similarity of the two curves, the synchronization degree between tiller change and leaf area change is quantified. Both the tiller number and the leaf area are key indicators for measuring the growth status of forage grass, and their synchronous changes usually reflect the coordination of forage grass in aspects such as nutrient utilization efficiency and photosynthesis ability during the growth process. When the change synchronization degree is greater than the preset synchronization degree threshold, it indicates that the forage grass in this marked grid has good consistency in tiller growth and leaf expansion, which often means that the forage grass grows healthily and has excellent quality, meeting the growth characteristics of high-quality forage grass, so it is determined as a concerned grid.

[0139] Specifically, the adjustment module includes:

[0140] A dispersion calculation unit for calculating the standard deviation of the positions of each of the concerned grids to obtain the dispersion;

[0141] A dispersion deviation calculation unit, which is connected to the dispersion calculation unit, and is used to calculate the relative deviation between the dispersion and the preset dispersion threshold when the dispersion is greater than the preset dispersion threshold to obtain the dispersion deviation;

[0142] An adjustment unit, which is connected to the dispersion deviation calculation unit, and is used to adjust the preset grid width according to the dispersion deviation and the preset dispersion deviation threshold when the dispersion deviation is greater than the preset dispersion deviation threshold to obtain the adjusted grid width.

[0143] The preset dispersion threshold is a standard value used to measure the dispersion degree of the concerned grids, which depends on the distribution characteristics of the forage in the field, the uniformity of forage growth, and the requirements for the forage coverage rate in the harvesting operation. It is usually set between 1.0 and 3.0. In this embodiment, it is set to 1.5, which can better distinguish whether the distribution of the concerned grids is too dispersed. When the dispersion degree exceeds this value, it indicates that the grid distribution may be too sparse, and the grid width needs to be adjusted to optimize the distribution.

[0144] The preset dispersion deviation threshold is a standard value used to determine whether the relative deviation between the dispersion degree and the preset dispersion threshold is acceptable. It depends on the stability of forage growth, the accuracy requirements of the harvesting operation, and the adaptability of the system to the changes in forage distribution, etc. It is usually set between 10% and 30%. In this embodiment, it is set to 20%, which can provide a certain degree of flexibility to cope with the natural fluctuations and uncertainties of forage distribution while ensuring the accuracy of the harvesting operation.

[0145] The dispersion degree is obtained by calculating the standard deviation of the positions of each concerned grid. When the dispersion degree is greater than the preset dispersion threshold, the relative deviation between the dispersion degree and the preset dispersion threshold is then calculated to obtain the dispersion deviation. Then, when the dispersion deviation is less than the preset dispersion deviation threshold, the preset grid width is adjusted according to the dispersion deviation and the preset dispersion deviation threshold, and finally the adjusted grid width is obtained.

[0146] The dispersion degree is obtained by calculating the standard deviation of the positions of the concerned grids. The smaller the standard deviation, the more concentrated the concerned grids are; on the contrary, the larger the standard deviation, the more dispersed they are. When the dispersion degree is greater than the preset dispersion threshold, the relative deviation between the dispersion degree and the preset dispersion threshold is then calculated. When the dispersion deviation is greater than the preset dispersion deviation threshold, it indicates that the concerned grids are relatively dispersed. At this time, the preset grid width is adjusted to increase the width to divide the area more finely and improve the accuracy of the harvesting plan.

[0147] Specifically, the adjustment unit includes:

[0148] A uniform index calculation sub-unit is used to mark all the timestamps during the preset adjustment duration when the dispersion deviation is greater than the preset dispersion deviation threshold. When the number of all timestamps is greater than the preset number threshold, the time intervals between all adjacent timestamps are calculated, and the ratio of the standard deviation to the mean value of the time intervals is calculated to obtain the uniform index;

[0149] An adjustment factor calculation sub-unit, which is connected to the uniform index calculation sub-unit, is used to calculate the relative deviation between the dispersion deviation and the preset dispersion deviation threshold to obtain the adjustment factor when the uniform index is less than the preset uniform index threshold;

[0150] An adjustment subunit, which is connected to the adjustment factor calculation subunit, is used to increase the preset grid width according to the adjustment factor and a preset adjustment coefficient to obtain an adjusted grid width.

[0151] The preset adjustment duration is a time period used to count and evaluate the occurrence of dispersion deviation, depending on factors such as the dynamic characteristics of forage growth, the frequency of harvesting operations, and the system's requirement for the response speed to changes in forage distribution. It is usually set between 1 day and 7 days. In this embodiment, it is set to 3 days, which can timely capture the change trend of forage distribution density.

[0152] The preset quantity threshold is a reference value used to determine whether the number of occurrences of dispersion deviation greater than the preset dispersion deviation threshold in the preset market is normal, depending on the frequency of possible uneven distribution during forage growth and the system's requirement for the data sample size. It is usually set between 3 and 10. In this embodiment, it is set to 5, which can evaluate the frequency of dispersion deviation.

[0153] The preset uniformity index threshold is a reference value used to judge the uniformity of time intervals, depending on the time distribution characteristics of dispersion deviation during forage growth and the requirement of harvesting operations for the uniformity of time intervals. It is usually set between 0.1 and 0.5. In this embodiment, it is set to 0.3, which can effectively distinguish the uniformity and non-uniformity of time intervals. When the uniformity index is less than 0.3, it indicates that the time intervals are relatively uniform.

[0154] The preset adjustment coefficient is a coefficient used to adjust the preset grid width, depending on the spatial characteristics of forage distribution and the requirement of harvesting operations for accuracy. It is usually set between 0.1 and 0.5. In this embodiment, it is set to 0.3, which can ensure the adjustment effect while avoiding excessive changes in the grid width caused by over-adjustment.

[0155] By marking all timestamps with dispersion deviation greater than the preset dispersion deviation threshold within the preset adjustment duration. When the number of these timestamps exceeds the preset quantity threshold, calculate the time intervals between all adjacent timestamps, and further calculate the ratio of the standard deviation to the mean of these time intervals to obtain the uniformity index. If the uniformity index is less than the preset uniformity index threshold, calculate the relative deviation of the dispersion deviation from the preset dispersion deviation threshold to obtain the adjustment factor. Finally, increase the preset grid width according to the adjustment factor and the preset adjustment coefficient to obtain the adjusted grid width.

[0156] By marking relevant timestamps within a preset adjusted duration and calculating the uniformity index of the time intervals when the quantity meets the standard, the time regularity of the occurrence of the dispersion deviation can be accurately evaluated. When the uniformity index is less than the preset threshold, it indicates that the time intervals are relatively uniform, meaning that the occurrence time distribution of the dispersion deviation exceeding the preset threshold within the preset adjusted duration is regular, indicating that the non-uniformity of the forage distribution shows a certain stability in time, and at the same time reflecting the stability of the system's monitoring of the forage distribution, indicating that the system can reliably capture the signal of the change in the forage distribution. Calculating the relative deviation between the dispersion deviation and the preset threshold as an adjustment factor can quantify the adjustment amplitude. According to the adjustment factor and the preset adjustment coefficient, increasing the width of the preset grid can expand the harvesting range, make the forage distribution in each grid more uniform, and reduce the harvesting problems caused by uneven distribution.

[0157] Specifically, the selection module includes:

[0158] A ratio calculation unit for calculating the ratio of the number of the key grids to the number of the concerned grids to obtain a quantity ratio;

[0159] A coincidence degree calculation unit connected to the ratio calculation unit for, when the quantity ratio is greater than the minimum value of the preset ratio range and less than the maximum value of the preset ratio range, counting the number of overlapping grids among all the key grids and all the concerned grids to obtain an overlapping quantity, and when the number of key grids is greater than the number of concerned grids, calculating the ratio of the overlapping quantity to the number of key grids to obtain a coincidence degree, or, when the number of key grids is less than the number of concerned grids, calculating the ratio of the overlapping quantity to the number of concerned grids to obtain a coincidence degree;

[0160] A selection unit connected to the coincidence degree calculation unit for, when the coincidence degree is greater than the preset coincidence degree threshold, determining the key grids as the harvesting grids to select a number of harvesting grids.

[0161] The preset ratio range is an interval used to determine whether the ratio of the number of key grids to the number of concerned grids is reasonable, depending on the growth distribution characteristics of the forage, the efficiency requirements of the harvesting operation, and the consistency requirements of the forage quality, and is usually set between 0.5 and 2.0. In this embodiment, it is set to [0.8, 1.2] to ensure that the number of key grids is not too large or too small relative to the number of concerned grids.

[0162] The preset coincidence degree threshold is a reference value used to determine whether to select key grids or concerned grids as harvesting grids, depending on the consistency requirements of the forage growth state, the accuracy of the harvesting operation, and the sensitivity of the forage quality to the coincidence degree, and is usually set between 0.6 and 0.9. In this embodiment, it is set to 0.7, which can provide a certain degree of flexibility while ensuring the accuracy of the selection of the harvesting grids.

[0163] By calculating the ratio of the number of key squares to the number of concerned squares, the quantity ratio is obtained. If the quantity ratio is less than the minimum value of the preset ratio range or greater than the maximum value of the preset ratio range, the concerned square is determined as a harvesting square. However, if the ratio comparison result shows that the quantity ratio is between the minimum and maximum values of the preset ratio range, the number of overlapping squares among all the key squares and all the concerned squares is counted to obtain the overlapping quantity. If the number of key squares is greater than the number of concerned squares, the ratio of the overlapping quantity to the number of key squares is calculated to obtain the overlapping degree; otherwise, the ratio of the overlapping quantity to the number of concerned squares is calculated to obtain the overlapping degree. Then, the overlapping degree is calculated. When the overlapping degree is greater than the preset overlapping degree threshold, the key square is determined as a harvesting square, and when the overlapping degree is less than the preset overlapping degree threshold, the concerned square is determined as a harvesting square.

[0164] By comprehensively considering the quantity ratio and the overlapping degree, first judge whether the quantity ratio is within a reasonable range. If it exceeds the range, directly determine that the concerned square is a harvesting square, which can quickly screen out the situations that obviously do not meet the quantity requirements. If the quantity ratio is within a reasonable range, then further judge through the overlapping degree. When the number of key squares and the number of concerned squares are different, different calculation methods are adopted to ensure the accuracy of the overlapping degree, thus more objectively reflecting the overlapping situation between the two, improving the accuracy and reliability of determining the harvesting square, and avoiding misjudgment. It reduces human intervention and errors, improves the efficiency and quality of selecting the harvesting square, ensures the objectivity and accuracy of the data, and provides strong support for subsequent harvesting operations.

[0165] Specifically, the preliminary screening module includes:

[0166] A height comparison unit for comparing the plant height with the preset plant height range to obtain a height comparison result;

[0167] A preliminary screening unit connected to the height comparison unit for preliminarily screening a number of temporary squares according to the lesion area when the height comparison result is that the plant height is greater than the minimum value of the preset plant height range and less than the maximum value of the preset plant height range.

[0168] By comparing the plant height with the preset plant height range, a height comparison result is obtained. If the result shows that the plant height is within the preset range, then a preliminary screening is carried out in combination with the lesion area to determine a number of temporary squares.

[0169] Coarse screening is carried out by comparing the plant height, which reflects the growth stage and development of forage grass. Plants that are too tall or too short may have problems of over-ripeness or poor growth respectively, affecting the quality and nutritional value. The lesion area reflects the health status of forage grass, and too many lesions will reduce the quality and yield. Therefore, by first conducting a preliminary screening based on height and then combining with the lesion area for a secondary screening, the forage grass areas that meet the requirements can be accurately locked, improving the harvesting efficiency and ensuring the forage grass quality.

[0170] Please continue to refer to Figure 4 as shown, which is the determination logic diagram of the temporary grids preliminarily screened by the preliminary screening unit in this embodiment;

[0171] The preliminary screening unit includes:

[0172] A lesion comparison unit for comparing the lesion area and a preset lesion area threshold to obtain a lesion comparison result;

[0173] A preliminary screening sub-unit, which is connected to the lesion comparison unit, and is used to determine the grid to be measured as the temporary grid when the lesion comparison result is that the lesion area is less than the preset lesion area threshold, so as to preliminarily screen out a number of temporary grids.

[0174] The preset lesion area threshold refers to the reference value for judging whether the forage grass is healthy, which depends on the disease resistance, growth stage and use of the forage grass variety, and is usually set between 5% and 15%. In this embodiment, it is set to 10%, which can, while ensuring the forage grass quality, minimize the misjudgment of being unqualified due to a small lesion area as much as possible and improve the accuracy of screening.

[0175] After the plant height meets the harvesting requirements, the lesion area and the preset lesion area threshold are compared. If the lesion area is less than the preset threshold, the grid to be measured is finally determined as the temporary grid, so as to gradually preliminarily screen out a number of temporary grids.

[0176] By screening the lesion area, the lesion area is a key indicator for evaluating the health status of forage grass. An overly large lesion area usually indicates that the forage grass is invaded by diseases, reducing the forage grass quality. When the lesion area is less than the preset lesion area threshold, the forage grass significantly affected by diseases can be effectively excluded, and the forage grass areas that meet the height and health requirements can be efficiently screened, improving the harvesting efficiency and forage grass quality.

[0177] The above are only the preferred embodiments of the present invention and are not used to limit the present invention; for those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A forage harvesting system based on a growth state prediction model, characterized in that, Including: A collection module for obtaining in real time the plant height, lesion area, tiller number, and leaf area of forage grass in each to-be-detected square constructed based on a preset square width; A primary screening module connected to the collection module for primarily screening a number of temporary squares according to the plant height, the lesion area, and a preset plant height range; A determination module respectively connected to the collection module and the primary screening module for determining a number of marked squares according to the lesion area and the tiller number of each of the temporary squares; A determination module respectively connected to the collection module and the determination module for determining a number of concerned squares according to the tiller number and the leaf area of each of the marked squares; An adjustment module connected to the determination module for adjusting the preset square width according to the positions of each of the concerned squares to obtain an adjusted square width; A prediction module respectively connected to the collection module and the adjustment module for inputting the plant height, the lesion area, the tiller number, and the leaf area collected based on the adjusted square width into a preset growth state prediction model to predict a number of key squares; A selection module respectively connected to the determination module and the prediction module for selecting a number of harvesting squares according to the key squares and the concerned squares determined based on the adjusted square width; A harvesting module connected to the selection module for harvesting the forage grass in each of the harvesting squares.

2. The forage harvesting system based on the growth state prediction model according to claim 1, characterized in that, The determination module includes: A normalization processing unit for performing normalization processing on all the lesion areas within a preset determination duration to obtain a lesion normalization data set, and performing normalization processing on all the tiller numbers within the preset determination duration to obtain a tiller normalization data set; A determination unit connected to the normalization processing unit for determining a number of the marked squares according to the lesion normalization data set and the tiller normalization data set.

3. The forage harvesting system based on the growth state prediction model according to claim 2, wherein, The determination unit includes: A correlation calculation sub-unit for calculating the correlation coefficient between the lesion normalization data set and the tiller normalization data set to obtain a change correlation degree; A determination sub-unit connected to the correlation calculation sub-unit for determining the temporary square as the marked square when the change correlation degree is less than a preset correlation degree threshold to determine a number of marked squares.

4. The forage harvesting system based on the growth state prediction model according to claim 3, characterized in that, The determination module includes: A tiller fluctuation value calculation unit for calculating the standard deviation of the tiller numbers at each moment from an initial moment to a preset determination duration to obtain a number of tiller fluctuation values; A leaf area fluctuation value calculation unit for calculating the standard deviation of the leaf areas at each moment from the initial moment to the preset determination duration to obtain a number of leaf area fluctuation values; A determination unit respectively connected to the tiller fluctuation value calculation unit and the leaf area fluctuation value calculation unit for determining a number of concerned squares according to all the tiller fluctuation values and all the leaf area fluctuation values.

5. The forage harvesting system based on the growth state prediction model according to claim 4, wherein The determination unit includes: A tiller change curve drawing sub-unit for drawing a curve of the change of all the tiller fluctuation values with time within the preset determination duration to obtain a tiller change curve; The leaf area change curve drawing subunit is used to draw a curve of the fluctuation values of all the leaf areas within the preset determination duration over time, so as to obtain a leaf area change curve; The synchronization degree calculation subunit, which is respectively connected to the tillering change curve drawing subunit and the leaf area change curve drawing subunit, is used to calculate the cosine similarity between the tillering change curve and the leaf area change curve, so as to obtain a change synchronization degree; The determination subunit, which is connected to the synchronization degree calculation subunit, is used to determine that the marked square is the concerned square when the change synchronization degree is greater than the preset synchronization degree threshold, so as to determine a number of concerned squares.

6. The forage harvesting system based on the growth state prediction model according to claim 5, wherein The adjustment module includes: The dispersion calculation unit is used to calculate the standard deviation of the positions of all the concerned squares, so as to obtain a dispersion; The dispersion deviation calculation unit, which is connected to the dispersion calculation unit, is used to calculate the relative deviation between the dispersion and the preset dispersion threshold when the dispersion is greater than the preset dispersion threshold, so as to obtain a dispersion deviation; The adjustment unit, which is connected to the dispersion deviation calculation unit, is used to adjust the preset square width according to the dispersion deviation and the preset dispersion deviation threshold when the dispersion deviation is greater than the preset dispersion deviation threshold, so as to obtain an adjusted square width.

7. The forage harvesting system based on the growth state prediction model according to claim 6, characterized in that, The adjustment unit includes: The uniformity index calculation subunit is used to mark all the timestamps when the dispersion deviation is greater than the preset dispersion deviation threshold within the preset adjustment duration. When the number of all the timestamps is greater than the preset number threshold, calculate the time intervals between all adjacent timestamps, and calculate the ratio of the standard deviation to the mean of the time intervals, so as to obtain a uniformity index; The adjustment factor calculation subunit, which is connected to the uniformity index calculation subunit, is used to calculate the relative deviation between the dispersion deviation and the preset dispersion deviation threshold when the uniformity index is less than the preset uniformity index threshold, so as to obtain an adjustment factor; The adjustment subunit, which is connected to the adjustment factor calculation subunit, is used to increase the preset square width according to the adjustment factor and the preset adjustment coefficient, so as to obtain an adjusted square width.

8. The forage harvesting system based on the growth state prediction model according to claim 7, characterized in that, The selection module includes: The ratio calculation unit is used to calculate the ratio of the number of the key squares to the number of the concerned squares, so as to obtain a quantity ratio; The coincidence degree calculation unit, which is connected to the ratio calculation unit, is used to count the number of overlapping squares among all the key squares and all the concerned squares to obtain an overlapping quantity when the quantity ratio is greater than the minimum value of the preset ratio range and less than the maximum value of the preset ratio range. When the number of the key squares is greater than the number of the concerned squares, calculate the ratio of the overlapping quantity to the number of the key squares to obtain a coincidence degree, or when the number of the key squares is less than the number of the concerned squares, calculate the ratio of the overlapping quantity to the number of the concerned squares to obtain a coincidence degree; The selection unit, which is connected to the coincidence degree calculation unit, is used to determine that the key square is the harvesting square when the coincidence degree is greater than the preset coincidence degree threshold, so as to select a number of harvesting squares.

9. The forage harvesting system based on the growth state prediction model according to claim 8, wherein The preliminary screening module includes: A height comparison unit for comparing the plant height and the preset plant height range to obtain a height comparison result; A primary screening unit connected to the height comparison unit for, when the height comparison result is that the plant height is greater than the minimum value of the preset plant height range and less than the maximum value of the preset plant height range, primarily screening a number of temporary squares according to the lesion area.

10. The forage harvesting system based on the growth state prediction model according to claim 9, wherein The primary screening unit includes: A lesion comparison unit for comparing the lesion area and a preset lesion area threshold to obtain a lesion comparison result; A primary screening sub-unit connected to the lesion comparison unit for, when the lesion comparison result is that the lesion area is less than the preset lesion area threshold, determining the square to be measured as the temporary square to primarily screen a number of temporary squares.

Images