An image-based intelligent spreading method and system

Abstract

This invention discloses an image-based intelligent feeding method and system, comprising: S1: collecting feeding data and image data at different feeding times and preprocessing them to obtain preprocessed feeding data, livestock image data, and feeding image data; S2: extracting livestock features based on the preprocessed livestock image data, and calculating incremental features and livestock weight; S3: calculating the feeding amount based on the livestock weight, and correcting the feeding amount based on the incremental features to obtain the corrected feeding amount; S4: calculating the target speed of the feeding motor based on the corrected feeding amount, the preprocessed feeding data, and the preprocessed feeding image data; S5: controlling the feeding vehicle to feed according to the target speed of the feeding motor. This invention calculates the feeding amount based on the livestock growth status and integrates image features and sensor data to intelligently control the target speed of the feeding motor, solving the problems of easy feeding outside the trough and inappropriate feed amount in traditional feeding methods.

Description

Technical Field

[0001] This invention relates to the field of intelligent material spreading technology, and in particular to an image-based intelligent material spreading method and system. Background Technology

[0002] Intelligent feed spreading is a farming method that integrates sensor technology and automated control technology. It aims to achieve precise and efficient feed spreading through intelligent means, thereby improving farming efficiency and reducing feed waste. Traditional feed spreading methods usually include manual spreading and mechanical spreading vehicles. Manual spreading mainly relies on people holding or carrying feed on their backs to spread it. This method is labor-intensive, inefficient, and it is difficult to ensure the uniformity of spreading. With the development of mechanization, mechanically powered spreading vehicles have emerged. These spreading vehicles usually control the speed and direction of spreading through a mechanical transmission system, which improves efficiency and the uniformity of spreading compared to manual spreading.

[0003] However, for traditional mechanical spreader trucks, the speed of the spreader motor is usually fixed or needs to be manually adjusted. The motor speed cannot adapt. During the spreading process, if the speed of the spreader truck changes or the direction deviates, but the motor still maintains the original speed, the feed will be spilled outside the intended area, resulting in waste. Therefore, the driver needs to concentrate highly on driving the spreader truck to avoid this situation, which increases the difficulty of operation and labor intensity. On the other hand, livestock at different growth stages have different feed requirements. Traditional feed spreaders usually rely solely on human experience to control the amount of feed spread, which can easily lead to insufficient or excessive feed when farmers prepare feed, affecting the growth efficiency of livestock and the cost of raising them. Summary of the Invention

[0004] In view of this, the present invention proposes an image-based intelligent feeding method and system, which aims to solve the problems of traditional feeding methods that easily spill feed outside the feeding trough and the inappropriate amount of feed.

[0005] To achieve the above objectives, the present invention provides the following technical solution: An image-based intelligent material spreading method includes: S1: Collect feed spreading data and image data at different feed spreading times and preprocess them to obtain preprocessed feed spreading data, preprocessed livestock image data, and preprocessed feed spreading image data; S2: Based on the preprocessed livestock image data, extract livestock image features, calculate incremental features, and then calculate livestock weight based on the incremental features; S3: Calculate the amount of feed to be spread based on the weight of the livestock, and correct the amount of feed to be spread based on the incremental characteristics to obtain the corrected amount of feed; S4: Calculate the target speed of the spreading motor based on the corrected spreading amount, the pre-processed spreading data, and the pre-processed spreading image data; S5: Control the spreading vehicle to spread material according to the calculated target speed of the spreading motor.

[0006] Furthermore, in step S1, the feeding data and image data at different feeding times are collected and preprocessed to obtain preprocessed feeding data, preprocessed livestock image data, and preprocessed feeding image data, including: S11: Collect spreading data, which includes the spreading distance between the spreading vehicle and the feeding trough, and the spreading vehicle's moving speed; the spreading distance between the spreading vehicle and the feeding trough is obtained by measuring the distance between the vehicle tracks and the feeding trough, and the spreading vehicle's moving speed is obtained by an acceleration sensor; S12: Iterate through the moving speed of the spreading truck, remove outliers using the 3σ principle, and obtain the pre-processed moving speed of the spreading truck; then, for the spreading distance between the spreading truck and the feeding trough, use the mean interpolation method to interpolate every two data points once to obtain the pre-processed spreading distance. The pre-processed moving speed of the spreading truck and the pre-processed spreading distance constitute the pre-processed spreading data. S13: Collect image data at different spreading times using a video camera installed on the spreading truck at a frame rate of 30 frames per second; S14: Crop the image data at different feeding times, using the portion containing livestock as the livestock image data and the remaining portion as the feeding image data. Then, adjust the size of both the livestock image data and the feeding image data to 640 pixels. 640 pixels, to obtain preprocessed livestock image data and preprocessed feed spreading image data.

[0007] Furthermore, in step S2, based on the preprocessed livestock image data, livestock image features are extracted, incremental features are calculated, and then livestock weight is calculated based on the incremental features, including: First, based on the preprocessed livestock image data, livestock image features are extracted. Second, combined with the livestock image features, the contour features and geometric features of the livestock at different feeding times are extracted, and a gating mechanism is designed to calculate the incremental features. Finally, the livestock weight is calculated based on the incremental features.

[0008] Furthermore, step S2 includes: S21: Extract the livestock image features at time one of feeding, calculated as follows: ; ; in, The original features of the livestock image at the moment of feeding. For convolutional neural network operations, This is the preprocessed image data of livestock at the time of feeding. The image features of livestock at the moment of feeding are shown. For dimension reduction matrix operations, For splicing operations, This is a global pooling operation. For operations on convolutional neural networks with a kernel size greater than or equal to 64, Operations for convolutional neural networks with kernel size less than or equal to 32; Traditional feed distribution methods struggle to accurately track livestock's real-time weight and growth, leading to inaccurate feed amounts, potential feed waste, or malnutrition. Furthermore, traditional methods largely rely on human experience to estimate livestock weight, lacking scientific basis. This invention employs convolutional neural networks (CNNs) and global pooling to calculate livestock image features. It utilizes CNNs with varying kernel sizes, using large kernels to capture broader contextual information and small kernels to capture detailed information. Finally, a dimensionality reduction matrix is ​​used to reduce the dimensionality of the features, removing redundant information, and then these features are fused to obtain the livestock image features. These extracted features are used in subsequent steps to refine the livestock's outline, area, and other geometric features, thereby more accurately estimating livestock weight. S22: Combine the livestock image features at feed-spreading moment one Hierarchical feature extraction was performed on the preprocessed livestock image data at feed-spreading time one, extracting the first contour feature and the first geometric feature of the livestock. The calculation method was as follows: ; ; ; ; in, For the original contour features, For Canny edge detection operations, The first outline features of livestock These are the first perimeter characteristic and the first area characteristic of livestock, respectively. For DenseNet network operations, The first geometric characteristic of livestock; Traditional Canny edge detection algorithms often fail to accurately extract the contours of target objects when processing images with complex backgrounds or interference. In this scenario, the presence of feeding trough fences severely interferes with the extraction of livestock contours. Directly using the Canny edge detection algorithm results in extracted contour features containing significant errors, which in turn affects subsequent livestock weight estimation and feed amount calculation. This invention uses the results obtained from the Canny edge detection algorithm as the original contour features, and then combines them with the livestock image features calculated in step S21 to correct the edge detection errors caused by the feeding trough fences. This allows for more accurate extraction of livestock contour features, providing more accurate basic data for subsequent extraction of livestock geometric features (such as perimeter and area). The steps for extracting the first contour features and first geometric features of livestock described above are uniformly represented as follows: ; in, For feature extraction operations; S23: Preprocessed livestock image data based on feeding time two The livestock image features at the second feeding time are extracted using the method in step S21. Combined with the livestock image features at the second feeding moment. The livestock image data at the second feeding time are preprocessed using the method in step S22. Hierarchical feature extraction was performed to obtain the second contour features of the livestock. and the second geometric features of livestock The calculation method is as follows: ; S24: Design a gating mechanism to calculate incremental features. The calculation method is as follows: ; ; ; ; in, The primary characteristic of livestock, As a second characteristic of livestock, For incremental gates, For the sigmoid function, This is an incremental weight matrix. For incremental bias vectors, It is an incremental feature; To adapt to the changing characteristics of livestock weight and morphology over time, this invention combines livestock image data from different time points and calculates incremental features to represent these changes. The incremental gate is designed to adjust the calculation of the incremental features. When the difference in livestock characteristics between two time points is large, the Gate value is close to 1, resulting in a larger calculated incremental feature. Conversely, when the difference in livestock characteristics between two time points is small, the Gate value is close to 0, resulting in a smaller calculated incremental feature. This mechanism helps to give more attention to situations with large feature changes and reduce unnecessary interference when feature changes are small, making the entire intelligent feeding system more flexible and intelligent. S25: Calculate livestock weight based on incremental characteristics, using the following method: ; in, The weight of the livestock.

[0009] Furthermore, in step S3, the amount of feed to be spread is calculated based on the livestock's weight, and the amount of feed to be spread is corrected by combining incremental characteristics to obtain the corrected amount of feed to be spread, including: S31: Calculate the amount of feed to be distributed based on the weight of the livestock. The calculation method is as follows: ; in, This refers to the amount of material to be spread. For support vector regression model; S32: Correct the spreading amount based on the incremental characteristics to obtain the corrected spreading amount. The calculation method is as follows: ; in, This is the corrected amount of material to be spread; This invention uses a support vector regression (SVR) model to calculate the initial feed amount A based on the livestock's weight. Then, the SVR model is used again to correct the feed amount based on the incremental feature ΔF, resulting in the corrected feed amount A'. The incremental feature ΔF reflects the characteristic changes of the livestock at different feed times, and the SVR model can convert it into a correction value for the feed amount, thereby achieving dynamic adjustment of the feed amount.

[0010] Furthermore, in step S4, the target rotational speed of the spreading motor is calculated based on the corrected spreading amount, the pre-processed spreading data, and the pre-processed spreading image data, including: First, the reference speed of the spreading motor is calculated based on the corrected spreading amount and the pre-processed spreading vehicle speed. Second, the original spreading features are extracted from the pre-processed spreading image data, and the spreading deviation category distribution is calculated based on the original spreading features. Then, the spreading deviation features are calculated by combining the spreading distance between the spreading vehicle and the feeding trough and the spreading deviation category distribution. Finally, the target speed of the spreading motor is calculated by combining the reference speed of the spreading motor and the spreading deviation features.

[0011] Furthermore, step S4 includes: S41: Calculate the base speed of the spreading motor based on the corrected spreading amount and the pre-treated spreading vehicle moving speed. The calculation method is as follows: ; in, This is the reference speed of the spreading motor. This is a scaling factor used to convert the ratio of the material spreading amount to the spreading vehicle speed into motor speed. The speed at which the spreading truck moves after pretreatment. This is the reference offset, a reference value used to adjust the motor speed; S42: Extract the original spreading features from the preprocessed spreading image data, and calculate the spreading deviation category distribution based on the original spreading features; then, combine the spreading distance between the spreading truck and the feed trough with the spreading deviation category distribution to calculate the spreading deviation features. The calculation method is as follows: ; ; ; in, The original characteristics of spreading materials, For the downsampling weight matrix of the material spreading, This is the preprocessed image data of the material spreading process. Distribution of material spreading deviation categories (no deviation, some deviation, severe deviation). For the softmax function, This refers to the spreading distance after pretreatment. This indicates a deviation in material spreading. Step S42 aims to calculate the feed spreading deviation characteristics to assess whether the feed accurately falls into the trough during the spreading process and the degree of deviation. This invention proposes to directly calculate the feed spreading deviation characteristics by analyzing the feed spreading image data and adjust the motor speed accordingly. First, a convolutional neural network is used to extract global, major, and local features from the preprocessed feed spreading image data. Then, these features are input into a multilayer perceptron to calculate the feed spreading deviation category distribution p, where p represents the degree of deviation of the feed falling into the trough during the spreading process, divided into three categories: no deviation, deviation, and severe deviation. Finally, the original feed spreading features and the feed spreading deviation category distribution p are concatenated to obtain the feed spreading deviation characteristics, which are used in the subsequent step of calculating the target speed of the feed spreading motor to provide information for adjusting the motor speed. S43: Calculate the target speed of the spreading motor by combining the reference speed of the spreading motor and the spreading deviation characteristics. The calculation method is as follows: ; in, This is the target speed of the material spreading motor.

[0012] Furthermore, in step S5, controlling the spreading vehicle to spread material according to the calculated target speed of the spreading motor includes: S51: Input the calculated target speed of the spreading motor into the spreading vehicle motor control system, and convert the target speed of the spreading motor into a digital signal of the spreading vehicle motor; S52: Employs a PID control algorithm to adjust the control signal based on the deviation between the actual speed of the spreading motor and the target speed of the spreading motor; S53: Real-time monitoring of the actual speed of the spreading motor and feedback to the PID control algorithm to adjust the target speed of the spreading motor in real time.

[0013] This invention also discloses an image-based intelligent material spreading system, comprising: Data acquisition and preprocessing module: collects feed spreading data and image data at different feeding times and preprocesses them to obtain preprocessed feed spreading data, preprocessed livestock image data, and preprocessed feed spreading image data; Livestock feature extraction module: Based on the preprocessed livestock image data, extract livestock image features, calculate incremental features, and then calculate livestock weight based on the incremental features; Feed Calculation Module: Calculates the feed amount based on the livestock's weight and corrects the feed amount by incorporating incremental characteristics to obtain the corrected feed amount; The target speed calculation module for the spreading motor calculates the target speed of the spreading motor based on the corrected spreading amount, the pre-processed spreading data, and the pre-processed spreading image data. Material spreading control module: controls the material spreading vehicle to spread material according to the calculated target speed of the material spreading motor.

[0014] Compared with the prior art, the advantages of the present invention are as follows: (1) The present invention first extracts the contour features and geometric features of livestock based on image data to provide basic data for subsequent feed amount calculation; secondly, by comparing the livestock features at different feed times, incremental features are calculated, and the feed amount is adjusted according to the incremental features and livestock weight to ensure that the feed amount always matches the actual needs of the livestock; finally, during the feed process, the present invention detects feed deviation in real time based on images and adjusts the speed of the feed motor according to the degree of deviation to ensure that the feed can fall accurately into the feed trough and reduce the situation of feed being spilled outside the feed trough. (2) In view of the problem that the feeding trough fence interferes with the extraction of the livestock outline, the present invention uses the results obtained by the Canny edge detection algorithm as the original outline features, and then corrects the edge detection error caused by the feeding trough fence by combining the calculated livestock image features, providing more accurate basic data for the subsequent extraction of the geometric features of the livestock (such as perimeter, area, etc.). (3) The present invention proposes an incremental gate design, which calculates the gate control value through the sigmoid function. When the difference between the livestock characteristics at two times is large, the incremental gate value is close to 1, and the calculation result of the incremental feature is large; when the difference is small, the incremental gate value is close to 0, and the calculation result of the incremental feature is small. This helps to give more attention when the livestock characteristics change significantly, and reduce unnecessary interference when the livestock characteristics change slightly, making the entire intelligent feeding system more flexible and intelligent. (4) The present invention introduces incremental characteristics and livestock weight as the basis for calculating feed amount, realizing dynamic adjustment of feed amount. It can adjust feed amount according to the actual growth of livestock, avoid feed waste and shortage, and improve feed utilization rate. (5) Traditional material spreading deviation detection usually relies on expensive distance sensors. However, this invention proposes a low-cost image-based material spreading deviation detection method. By analyzing material spreading image data, the material spreading deviation characteristics are calculated, and the motor speed is adjusted accordingly, thereby reducing the cost of the system. Attached Figure Description

[0015] Figure 1 A schematic flowchart of the image-based intelligent material spreading method provided by the present invention; Figure 2 A schematic diagram of the algorithm flow for extracting livestock outline features and geometric features provided by the present invention; Figure 3 This is a schematic diagram of the algorithm flow for incremental feature extraction provided by the present invention. Detailed Implementation

[0016] The present invention will be further described below with reference to the accompanying drawings, but this is not intended to limit the present invention in any way. Any modifications or substitutions made based on the teachings of the present invention shall fall within the protection scope of the present invention.

[0017] Example 1: An image-based intelligent material spreading method, such as Figure 1 As shown, it includes the following steps: S1: Collect feed spreading data and image data at different feed spreading times and preprocess them to obtain preprocessed feed spreading data, preprocessed livestock image data, and preprocessed feed spreading image data; S11: Collect spreading data, which includes the spreading distance between the spreading vehicle and the feeding trough, and the spreading vehicle's moving speed; the spreading distance between the spreading vehicle and the feeding trough is obtained by measuring the distance between the vehicle tracks and the feeding trough, and the spreading vehicle's moving speed is obtained by an acceleration sensor; S12: Iterate through the moving speed of the spreading truck, remove outliers using the 3σ principle, and obtain the pre-processed moving speed of the spreading truck; then, for the spreading distance between the spreading truck and the feeding trough, use the mean interpolation method to interpolate every two data points once to obtain the pre-processed spreading distance. The pre-processed moving speed of the spreading truck and the pre-processed spreading distance constitute the pre-processed spreading data. S13: Collect image data at different spreading times using a video camera installed on the spreading truck at a frame rate of 30 frames per second; S14: Crop the image data at different feeding times, using the portion containing livestock as the livestock image data and the remaining portion as the feeding image data. Then, adjust the size of both the livestock image data and the feeding image data to 640 pixels. 640 pixels, to obtain preprocessed livestock image data and preprocessed feed spreading image data.

[0018] S2: Based on the preprocessed livestock image data, extract livestock image features, calculate incremental features, and then calculate livestock weight based on the incremental features; S21: Extract the livestock image features at time one of feeding, calculated as follows: ; ; in, The original features of the livestock image at the moment of feeding. For convolutional neural network operations, This is the preprocessed image data of livestock at the time of feeding. The image features of livestock at the moment of feeding are shown. For dimension reduction matrix operations, For splicing operations, This is a global pooling operation. For operations on convolutional neural networks with a kernel size greater than or equal to 64, Operations for convolutional neural networks with kernel size less than or equal to 32; Traditional feed distribution methods struggle to accurately track livestock's real-time weight and growth, leading to inaccurate feed amounts, potential feed waste, or malnutrition. Furthermore, traditional methods largely rely on human experience to estimate livestock weight, lacking scientific basis. This invention employs convolutional neural networks (CNNs) and global pooling to calculate livestock image features. It utilizes CNNs with varying kernel sizes, using large kernels to capture broader contextual information and small kernels to capture detailed information. Finally, a dimensionality reduction matrix is ​​used to reduce the dimensionality of the features, removing redundant information, and then these features are fused to obtain the livestock image features. These extracted features are used in subsequent steps to refine the livestock's outline, area, and other geometric features, thereby more accurately estimating livestock weight. S22: Combine the livestock image features at feed-spreading moment one Hierarchical feature extraction was performed on the preprocessed livestock image data at feed-spreading time one, extracting the first contour features and first geometric features of the livestock, such as... Figure 2 As shown, the calculation method is as follows: ; ; ; ; in, For the original contour features, For Canny edge detection operations, The first outline features of livestock These are the first perimeter characteristic and the first area characteristic of livestock, respectively. For DenseNet network operations, The first geometric characteristic of livestock; Traditional Canny edge detection algorithms often fail to accurately extract the contours of target objects when processing images with complex backgrounds or interference. In this scenario, the presence of feeding trough fences severely interferes with the extraction of livestock contours. Directly using the Canny edge detection algorithm results in extracted contour features containing significant errors, which in turn affects subsequent livestock weight estimation and feed amount calculation. This invention uses the results obtained from the Canny edge detection algorithm as the original contour features, and then combines them with the livestock image features calculated in step S21 to correct the edge detection errors caused by the feeding trough fences. This allows for more accurate extraction of livestock contour features, providing more accurate basic data for subsequent extraction of livestock geometric features (such as perimeter and area). The steps for extracting the first contour features and first geometric features of livestock described above are uniformly represented as follows: ; in, For feature extraction operations; S23: Preprocessed livestock image data based on feeding time two The livestock image features at the second feeding time are extracted using the method in step S21. Combined with the livestock image features at the second feeding moment. The livestock image data at the second feeding time are preprocessed using the method in step S22. Hierarchical feature extraction was performed to obtain the second contour features of the livestock. and the second geometric features of livestock The calculation method is as follows: ; S24: Design a gating mechanism to calculate incremental features, such as Figure 3 As shown, the calculation method is as follows: ; ; ; ; in, The primary characteristic of livestock, As a second characteristic of livestock, For incremental gates, For the sigmoid function, This is an incremental weight matrix. For incremental bias vectors, It is an incremental feature; To adapt to the changing characteristics of livestock weight and morphology over time, this invention combines livestock image data from different time points and calculates incremental features to represent these changes. The incremental gate is designed to adjust the calculation of the incremental features. When the difference in livestock characteristics between two time points is large, the Gate value is close to 1, resulting in a larger calculated incremental feature. Conversely, when the difference in livestock characteristics between two time points is small, the Gate value is close to 0, resulting in a smaller calculated incremental feature. This mechanism helps to give more attention to situations with large feature changes and reduce unnecessary interference when feature changes are small, making the entire intelligent feeding system more flexible and intelligent. S25: Calculate livestock weight based on incremental characteristics, using the following method: ; in, The weight of the livestock.

[0019] S3: Calculate the amount of feed to be spread based on the weight of the livestock, and correct the amount of feed to be spread based on the incremental characteristics to obtain the corrected amount of feed; S31: Calculate the amount of feed to be distributed based on the weight of the livestock. The calculation method is as follows: ; in, This refers to the amount of material to be spread. For support vector regression model; S32: Correct the spreading amount based on the incremental characteristics to obtain the corrected spreading amount. The calculation method is as follows: ; in, This is the corrected amount of material to be spread; For example, suppose in a farm, the weight calculated at time 1 Weight calculated at time 2 The calculated incremental features ; Weight at time 2 The amount of material to be spread is calculated using the SVR model as input. Using incremental features The correction amount is calculated using the SVR model as input. =0.5kg, then the corrected spreading amount ; This invention uses a support vector regression (SVR) model to calculate the initial feed amount A based on the livestock's weight. Then, the SVR model is used again to correct the feed amount based on the incremental feature ΔF, resulting in the corrected feed amount A'. The incremental feature ΔF reflects the characteristic changes of the livestock at different feed times, and the SVR model can convert it into a correction value for the feed amount, thereby achieving dynamic adjustment of the feed amount.

[0020] S4: Calculate the target speed of the spreading motor based on the corrected spreading amount, the pre-processed spreading data, and the pre-processed spreading image data; S41: Calculate the base speed of the spreading motor based on the corrected spreading amount and the pre-treated spreading vehicle moving speed. The calculation method is as follows: ; in, This is the reference speed of the spreading motor. This is a scaling factor used to convert the ratio of the material spreading amount to the spreading vehicle speed into motor speed. The speed at which the spreading truck moves after pretreatment. This is the reference offset, a reference value used to adjust the motor speed; S42: Extract the original spreading features from the preprocessed spreading image data, and calculate the spreading deviation category distribution based on the original spreading features; then, combine the spreading distance between the spreading truck and the feed trough with the spreading deviation category distribution to calculate the spreading deviation features. The calculation method is as follows: ; ; ; in, The original characteristics of spreading materials, For the downsampling weight matrix of the material spreading, This is the preprocessed image data of the material spreading process. For the distribution of material spreading deviation categories, For the softmax function, This refers to the spreading distance after pretreatment. This indicates a deviation in material spreading. Furthermore, when the feed truck moves further and further away from the trough, resulting in multiple consecutive serious deviations or alternating instances of "some deviation" and "serious deviation," this system can automatically detect such situations and issue an alarm to the driver. The calculation method is as follows: ; ; in, This is a deviation alarm feature. This is a gated loop unit operation with a length of 5. The original material spreading characteristics at time t, Let be the distribution of material spreading deviation categories at time t. This represents the probability of a deviation alarm. This is the deviation alarm matrix. This is the deviation alarm bias vector; A gated loop unit of length 5 performs feature extraction in groups of 5 time units, while simultaneously performing a sliding window operation in groups of 2 time units; when a detection is made... When a deviation alarm is detected, a deviation alarm is issued. At any time, move the sliding window once to perform deviation alarm detection for the next time unit; For example, there is data with 9 time units. First, using data entries 1, 2, 3, 4, and 5 as input, if detected... If a deviation alarm is detected, a deviation alarm will be issued. If the condition is met, a sliding window operation is performed, and the 3rd, 4th, 5th, 6th, and 7th data items are used as input for the next round of detection; Step S42 aims to calculate the feed spreading deviation characteristics to assess whether the feed accurately falls into the trough during the spreading process and the degree of deviation. This invention proposes to directly calculate the feed spreading deviation characteristics by analyzing the feed spreading image data and adjust the motor speed accordingly. First, a convolutional neural network is used to extract global, major, and local features from the preprocessed feed spreading image data. Then, these features are input into a multilayer perceptron to calculate the feed spreading deviation category distribution p, where p represents the degree of deviation of the feed falling into the trough during the spreading process, divided into three categories: no deviation, deviation, and severe deviation. Finally, the original feed spreading features and the feed spreading deviation category distribution p are concatenated to obtain the feed spreading deviation characteristics, which are used in the subsequent step of calculating the target speed of the feed spreading motor to provide information for adjusting the motor speed. S43: Calculate the target speed of the spreading motor by combining the reference speed of the spreading motor and the spreading deviation characteristics. The calculation method is as follows: ; in, The target speed for the material spreading motor; For example, the speed of the pre-treated spreading truck Corrected amount of material to be spread , scaling factor Reference offset The spreading distance between the spreading truck and the feeding trough: ,but ; The calculated distribution p of the material spreading deviation categories is [0.1, 0.7, 0.2], which means that the probability of no deviation is 0.1, the probability of deviation is 0.7, and the probability of severe deviation is 0.2. The calculation result is , Therefore, the corrected target speed of the spreading motor is: .

[0021] S5: Control the spreading vehicle to spread material according to the calculated target speed of the spreading motor; S51: Input the calculated target speed of the spreading motor into the spreading vehicle motor control system, and convert the target speed of the spreading motor into a digital signal of the spreading vehicle motor; S52: Employs a PID control algorithm to adjust the control signal based on the deviation between the actual speed of the spreading motor and the target speed of the spreading motor; S53: Real-time monitoring of the actual speed of the spreading motor and feedback to the PID control algorithm to adjust the target speed of the spreading motor in real time.

[0022] Example 2: The present invention also discloses an image-based intelligent material spreading system, comprising: Data acquisition and preprocessing module: collects feed spreading data and image data at different feeding times and preprocesses them to obtain preprocessed feed spreading data, preprocessed livestock image data, and preprocessed feed spreading image data; Livestock feature extraction module: Based on the preprocessed livestock image data, extract livestock image features, calculate incremental features, and then calculate livestock weight based on the incremental features; Feed Calculation Module: Calculates the feed amount based on the livestock's weight and corrects the feed amount by incorporating incremental characteristics to obtain the corrected feed amount; The target speed calculation module for the spreading motor calculates the target speed of the spreading motor based on the corrected spreading amount, the pre-processed spreading data, and the pre-processed spreading image data. Material spreading control module: controls the material spreading vehicle to spread material according to the calculated target speed of the material spreading motor.

[0023] It should be noted that the sequence numbers of the above embodiments of the present invention are merely for descriptive purposes and do not represent the superiority or inferiority of the embodiments. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, apparatus, article, or method. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0024] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0025] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. An image-based intelligent spreading method, characterized in that, Includes the following steps: S1: Collect feed spreading data and image data at different feed spreading times and preprocess them to obtain preprocessed feed spreading data, preprocessed livestock image data, and preprocessed feed spreading image data; Step S1 includes: S11: Collect spreading data, which includes the spreading distance between the spreading vehicle and the feeding trough, and the spreading vehicle's moving speed; the spreading distance between the spreading vehicle and the feeding trough is obtained by measuring the distance between the vehicle tracks and the feeding trough, and the spreading vehicle's moving speed is obtained by an acceleration sensor; S12: Iterate through the moving speed of the spreading truck, remove outliers using the 3σ principle, and obtain the pre-processed moving speed of the spreading truck; then, for the spreading distance between the spreading truck and the feeding trough, use the mean interpolation method to interpolate every two data points once to obtain the pre-processed spreading distance. The pre-processed moving speed of the spreading truck and the pre-processed spreading distance constitute the pre-processed spreading data. S13: Collect image data at different spreading times using a video camera installed on the spreading truck at a frame rate of 30 frames per second; S14: Crop the image data at different feeding times, using the portion containing livestock as the livestock image data and the remaining portion as the feeding image data. Then, adjust the size of both the livestock image data and the feeding image data to 640 pixels. 640 pixels, to obtain preprocessed livestock image data and preprocessed feed spreading image data; S2: Based on the preprocessed livestock image data, extract livestock image features, calculate incremental features, and then calculate livestock weight based on the incremental features; S3: Calculate the amount of feed to be spread based on the weight of the livestock, and correct the amount of feed to be spread based on the incremental characteristics to obtain the corrected amount of feed; S4: Calculate the target speed of the spreading motor based on the corrected spreading amount, the pre-processed spreading data, and the pre-processed spreading image data; Step S4 includes: First, the reference speed of the spreading motor is calculated based on the corrected spreading amount and the pre-processed spreading vehicle speed. Second, the original spreading features are extracted from the pre-processed spreading image data, and the spreading deviation category distribution is calculated based on the original spreading features. Then, the spreading deviation features are calculated by combining the spreading distance between the spreading vehicle and the feeding trough and the spreading deviation category distribution. Finally, the target speed of the spreading motor is calculated by combining the reference speed of the spreading motor and the spreading deviation features. S5: Control the spreading vehicle to spread material according to the calculated target speed of the spreading motor.

2. The image-based intelligent material spreading method according to claim 1, characterized in that, Step S2 includes: First, based on the preprocessed livestock image data, livestock image features are extracted. Second, combined with the livestock image features, the contour features and geometric features of the livestock at different feeding times are extracted, and a gating mechanism is designed to calculate the incremental features. Finally, the livestock weight is calculated based on the incremental features.

3. The image-based intelligent material spreading method according to claim 2, characterized in that, Step S2 includes: S21: Extract the livestock image features at time one of feeding, calculated as follows: ； ； in, The original features of the livestock image at the moment of feeding. For convolutional neural network operations, This is the preprocessed image data of livestock at the time of feeding. The image features of livestock at the moment of feeding are shown. For dimension reduction matrix operations, For splicing operations, This is a global pooling operation. For operations on convolutional neural networks with a kernel size greater than or equal to 64, Operations for convolutional neural networks with kernel size less than or equal to 32; S22: Combine the livestock image features at feed-spreading moment one Hierarchical feature extraction was performed on the preprocessed livestock image data at feed-spreading time one, extracting the first contour feature and the first geometric feature of the livestock. The calculation method was as follows: ； ； ； ； in, For the original contour features, For Canny edge detection operations, The first outline features of livestock These are the first perimeter characteristic and the first area characteristic of livestock, respectively. For DenseNet network operations, The first geometric characteristic of livestock; The steps for extracting the first contour features and first geometric features of livestock described above are uniformly represented as follows: ； in, For feature extraction operations; S23: Preprocessed livestock image data based on feeding time two The livestock image features at the second feeding time are extracted using the method in step S21. Combined with the livestock image features at the second feeding moment. The livestock image data at the second feeding time are preprocessed using the method in step S22. Hierarchical feature extraction was performed to obtain the second contour features of the livestock. and the second geometric features of livestock The calculation method is as follows: ； S24: Design a gating mechanism to calculate incremental features. The calculation method is as follows: ； ； ； ； in, The primary characteristic of livestock, As a second characteristic of livestock, For incremental gates, For the sigmoid function, This is an incremental weight matrix. For incremental bias vectors, It is an incremental feature; S25: Calculate livestock weight based on incremental characteristics, using the following method: ； in, The weight of the livestock.

4. The image-based intelligent material spreading method according to claim 3, characterized in that, Step S3 includes: S31: Calculate the amount of feed to be distributed based on the weight of the livestock. The calculation method is as follows: ； in, This refers to the amount of material to be spread. For support vector regression model; S32: Correct the spreading amount based on the incremental characteristics to obtain the corrected spreading amount. The calculation method is as follows: ； in, This is the corrected amount of material to be spread.

5. The image-based intelligent material spreading method according to claim 4, characterized in that, Step S4 includes: S41: Calculate the base speed of the spreading motor based on the corrected spreading amount and the pre-treated spreading vehicle moving speed. The calculation method is as follows: ； in, This is the reference speed of the spreading motor. As a scaling factor, The speed at which the spreading truck moves after pretreatment. This is the reference offset; S42: Extract the original spreading features from the preprocessed spreading image data, and calculate the spreading deviation category distribution based on the original spreading features; then, combine the spreading distance between the spreading truck and the feed trough with the spreading deviation category distribution to calculate the spreading deviation features; the calculation method is as follows: ； ； ； in, The original characteristics of spreading materials, For the downsampling weight matrix of the material spreading, This is the preprocessed image data of the material spreading process. For the distribution of material spreading deviation categories, For the softmax function, The spreading distance after pretreatment. This indicates a deviation in material spreading. S43: Calculate the target speed of the spreading motor by combining the reference speed of the spreading motor and the spreading deviation characteristics. The calculation method is as follows: ； in, This is the target speed of the material spreading motor.

6. The image-based intelligent material spreading method according to claim 1, characterized in that, Step S5 includes: S51: Input the calculated target speed of the spreading motor into the spreading vehicle motor control system, and convert the target speed of the spreading motor into a digital signal of the spreading vehicle motor; S52: Employs a PID control algorithm to adjust the control signal based on the deviation between the actual speed of the spreading motor and the target speed of the spreading motor; S53: Real-time monitoring of the actual speed of the spreading motor and feedback to the PID control algorithm to adjust the target speed of the spreading motor in real time.

7. An image-based intelligent material spreading system, characterized in that, include: Data acquisition and preprocessing module: collects feed spreading data and image data at different feeding times and preprocesses them to obtain preprocessed feed spreading data, preprocessed livestock image data, and preprocessed feed spreading image data; Livestock feature extraction module: Based on the preprocessed livestock image data, extract livestock image features, calculate incremental features, and then calculate livestock weight based on the incremental features; Feed Calculation Module: Calculates the feed amount based on the livestock's weight and corrects the feed amount by incorporating incremental characteristics to obtain the corrected feed amount; The target speed calculation module for the spreading motor calculates the target speed of the spreading motor based on the corrected spreading amount, the pre-processed spreading data, and the pre-processed spreading image data. The material spreading control module controls the material spreading vehicle to spread material according to the calculated target speed of the spreading motor; To achieve an image-based intelligent material spreading method as described in any one of claims 1-6.

Images