Juncao adaptive harvesting depth control system

Abstract

The present invention relates to the field of harvesting control technology, and in particular to a Juncao adaptive harvesting depth control system, which includes the following steps: a camera monitoring module: capturing initial image data of the harvesting situation in front of a tool group, and recording the movement trajectory of the Juncao stems; a rebound effect recognition module: recognizing the movement characteristics of the Juncao stems through a convolutional neural network, and judging whether the Juncao stems have a rebound effect; a rebound effect scoring module: scoring the rebound effect according to the number of rebounds and the rebound frequency; and a depth adaptive adjustment module: dynamically adjusting the harvesting depth of the tool group according to the rebound effect scoring result. The present invention improves the ability to understand and judge the Juncao movement pattern, thereby ensuring the accuracy and stability of the harvesting operation.

Description

Technical Field

[0001] The present invention relates to the technical field of a Juncao adaptive harvesting depth control system, and in particular to a Juncao adaptive harvesting depth control system. Background Art

[0002] With the development of large-scale mechanized harvesting technology, how to improve the efficiency and accuracy of Juncao harvesting has become an urgent problem to be solved. Existing Juncao harvesting equipment generally has insufficient adaptability to complex growth environments, especially when facing the rebound effect when the Juncao stems come into contact with the harvester tool group. It is often difficult to achieve accurate harvesting depth control, resulting in repeated jams, missed harvesting, excessive cutting or incomplete cutting during harvesting operations.

[0003] When the mushroom grass stems come into contact with the tool group, a rebound effect often occurs due to their elasticity, causing the originally cut mushroom grass to re-enter the area in front of the tool group, but cannot enter the tool group to complete the crushing. The occurrence of this rebound effect makes it difficult for existing harvesting equipment to effectively handle the mushroom grass stems, especially under high density or specific environmental conditions. Traditional fixed-depth tools are difficult to adapt to this complex movement behavior, which in turn affects the overall harvesting efficiency. Summary of the invention

[0004] The invention provides a mushroom grass self-adaptive harvesting depth control system.

[0005] The Juncao adaptive harvesting depth control system includes the following steps:

[0006] Camera monitoring module: The camera is installed on the side wall of the cutter group protective cover to capture the initial image data of the harvesting situation in front of the cutter group in real time and record the movement trajectory of the Juncao stems;

[0007] Rebound effect recognition module: Based on the motion trajectory recorded in the initial image data, the convolutional neural network is used to identify the motion characteristics of the Juncao stems to determine whether the Juncao stems have a rebound effect;

[0008] Rebound effect scoring module: Based on the results of convolutional neural network recognition, the number of rebounds and the frequency of rebounds are recorded, and the rebound effect is scored according to the number of rebounds and the frequency of rebounds. The higher the score, the more stems of the rebounding grass are, and the more serious the rebound effect is.

[0009] Depth adaptive adjustment module: dynamically adjusts the harvesting depth of the tool group according to the rebound effect scoring result, specifically including:

[0010] When the rebound effect score exceeds the preset score threshold, the cutter group performs bidirectional depth adjustment, that is, repeatedly adjusts between deep and shallow depths to ensure that the rebounded grass stems enter the cutter group smoothly;

[0011] When the rebound effect score is lower than the preset score threshold, the tool group only performs depth adjustment in one direction, saving energy and improving efficiency, until the rebound effect mushroom grass stems enter the tool group and return to the original depth.

[0012] Optionally, the camera is an industrial camera, which is installed on the side wall of the tool group protective cover through a fixed bracket. The camera records the movement trajectory of the mushroom grass stems when approaching the tool group by sampling images multiple times and combining with a motion tracking algorithm.

[0013] Optionally, when recording the motion trajectory of the Juncao stems when they approach the tool group, the motion tracking algorithm uses an optical flow method, which calculates the object's motion speed and direction through the motion changes of pixels between consecutive image frames, and tracks the Juncao stems' motion trajectory in real time;

[0014] In the present invention, the optical flow method captures continuous image frames through a camera to record the movement trajectory of the grass stems in front of the tool set. The pixel points of the grass stems in each frame of the image will be tracked in real time. and The displacement between , and obtain the motion trajectory of the stems during the harvesting process.

[0015] Optionally, the identifying the movement pattern of the stems of Juncao by using a convolutional neural network specifically includes:

[0016] Displacement obtained based on optical flow method Get the movement trajectory of the Juncao stem , as the input of the convolutional neural network;

[0017] The convolutional neural network extracts patterns from the input motion trajectory data through multiple convolutional layers, pooling layers, and fully connected layers. Each frame of image data and the motion changes therein are processed by multiple convolution kernels to extract the motion features of the Juncao stems. The motion features include:

[0018] Changes in the contact frequency and position between the Juncao stems and the cutter set;

[0019] The track of the repeated movement of the stems of the Juncao;

[0020] The changing trend of the speed and direction of the movement of the Juncao stems;

[0021] The feature map after convolution and pooling is used to identify complex motion patterns, and finally outputs a high-dimensional feature vector F, which represents the motion pattern of the stems of Juncao.

[0022] In convolutional neural networks, motion trajectory data can be discretized into frame data input. Assume that the motion trajectory segment captured by the camera at time t is a set of continuous frame sequences. , where each frame of the image contains the location information of the Juncao.

[0023] Optionally, the method of determining whether a rebound effect occurs in the stems of the mushroom grass includes determining whether a rebound effect occurs by using a double standard based on motion features extracted by a convolutional neural network.

[0024] Optionally, the dual standards include that the mushroom grass stems contact the cutter group multiple times without being completely cut, and that the mushroom grass stems move repeatedly within a fixed range in front of the cutter group.

[0025] Optionally, the criteria for the Juncao stems to not be completely cut after multiple contacts with the cutting tool set include:

[0026] The motion characteristics analyzed by the convolutional neural network are used to detect whether the stems of the mushroom grass have frequent short-term contact behavior in front of the tool group. If the motion characteristics show that the mushroom grass contacts a certain area of ​​the tool group multiple times in a short period of time but is not cut by the tool group, it is regarded as a rebound effect. The number of contacts, that is, the number of rebounds N, is calculated: N>Nthreshold. If the number of rebounds exceeds the threshold Nthreshold, it is judged that the mushroom grass has not been cut and there is a rebound.

[0027] Optionally, the criteria for the repeated movement of the stems of the grass within a fixed range in front of the cutter group include:

[0028] The motion trajectory extracted by the convolutional neural network is used to detect whether the stems of the Juncao move repeatedly within a fixed spatial range. This behavior is usually related to the rebound effect. The rebound is judged by the periodic changes and displacement data of the motion trajectory. If the trajectory moves back and forth within a fixed range, it is a rebound, and the rebound frequency M exceeds the threshold Mthreshold:

[0029] M>Mthreshold and |R(t1,t2)|<Rthreshold, it is judged that the Juncao has a rebound effect.

[0030] R(t1, t2) represents the displacement (motion trajectory range) of the Juncao stem between time t1 and t2. In other words, it represents the motion range or moving distance of the Juncao in front of the tool group during this period of time. |R(t1, t2)| is the absolute value of the Juncao motion range R(t1, t2).

[0031] Optionally, the rebound effect scoring module is based on the number of rebounds. and rebound frequency Conduct a rebound effect score, including:

[0032] By detecting the rebound contact times of the stems of the grass in front of the cutting tool group and the rebound frequency per unit time , calculate the rebound effect score , the scoring formula is: ,in, and are weight parameters used to adjust the impact of rebound times and rebound frequency on the score; score The higher the score, the more rebounding grass stems there are and the more severe the rebound effect is. The working depth of the tool set is adjusted according to the score to cope with different degrees of rebound effect.

[0033] Optionally, the motion trajectory calculation is expressed as:

[0034] ,in, Indicates that the stems of Juncao are The two-dimensional spatial position at the time of the movement, describing the movement trajectory of the Juncao stem from the initial position to the current moment, and It's time The speed of movement at any moment, and Respectively represent the time The initial position coordinates at time ; and Respectively represent the time Always Direction and Directional velocity component; and Respectively represent the time Internal and The cumulative displacement in the direction, Indicates that Juncao Displacement in direction, Indicates that Juncao Displacement in direction.

[0035] Beneficial effects of the present invention:

[0036] The present invention, through the combination of industrial cameras and convolutional neural networks, can accurately detect the rebound effect of the mushroom grass stems in front of the tool group, and use the optical flow method to record the movement trajectory of the mushroom grass stems. Combined with the high-level motion features extracted by the convolutional neural network, it can effectively identify complex rebound behaviors, such as multiple contacts with the tool group in a short period of time without cutting, or repeated movement of the mushroom grass within a fixed range in front of the tool group. This detection mechanism has not yet been applied in traditional harvesting equipment, which improves the understanding and judgment ability of the mushroom grass movement pattern, thereby ensuring the accuracy and stability of the harvesting operation.

[0037] The present invention introduces a rebound effect scoring mechanism, and dynamically adjusts the depth of the tool group according to the number of rebounds and the rebound frequency of the mushroom grass. Through scoring, it can automatically determine whether to perform unidirectional depth adjustment or bidirectional depth cycle adjustment according to the severity of the rebound effect, thereby ensuring that the rebounded mushroom grass can smoothly enter the tool group for crushing processing. This intelligent adjustment strategy ensures that the system can adaptively adjust according to the actual operating conditions, improves the efficiency and processing capacity of the operation, and avoids the problems of repeated jams and miscutting in traditional harvesting systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0039] Figure 1 A schematic diagram of system function modules according to an embodiment of the present invention;

[0040] Figure 2 The figure is a schematic diagram of the system operation logic of the embodiment of the present invention. DETAILED DESCRIPTION

[0041] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0042] It should be noted that the references to "one embodiment", "embodiment", "exemplary embodiments", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0043] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0044] like Figure 1-Figure 2 As shown, the Juncao adaptive harvesting depth control system includes the following steps:

[0045] Camera monitoring module: The camera is installed on the side wall of the cutter group protective cover to capture the initial image data of the harvesting situation in front of the cutter group in real time and record the movement trajectory of the Juncao stems;

[0046] Rebound effect recognition module: Based on the motion trajectory recorded in the initial image data, the convolutional neural network is used to identify the motion characteristics of the Juncao stems, detect the repeated collision, deviation or inability of the Juncao stems to enter the cutter group, and determine whether the Juncao stems have a rebound effect;

[0047] Rebound effect scoring module: Based on the results of convolutional neural network recognition, the number of rebounds and the frequency of rebounds are recorded, and the rebound effect is scored according to the number of rebounds and the frequency of rebounds. The higher the score, the more stems of the rebounding grass are, and the more serious the rebound effect is.

[0048] Depth adaptive adjustment module: When the rebound effect is detected, the working depth of the cutter group is automatically adjusted, and the contact point between the cutter group and the stem of the grass is changed to ensure that the grass after rebounding smoothly enters the cutter group for crushing. According to the rebound effect scoring results, the harvesting depth of the cutter group is dynamically adjusted, including:

[0049] When the rebound effect score exceeds the preset score threshold, the cutter group performs bidirectional depth adjustment, that is, repeatedly adjusts between deep and shallow depths to ensure that the rebounded grass stems enter the cutter group smoothly;

[0050] When the rebound effect score is lower than the preset score threshold, the tool group only performs depth adjustment in one direction, saving energy and improving efficiency, until the rebound effect mushroom grass stems enter the tool group and return to the original depth.

[0051] The camera uses an industrial camera and is installed on the side wall of the tool group protective cover through a fixed bracket. The fixed bracket has an adjustable function to ensure that the camera can cover the harvesting area in front of the tool group. The camera is equipped with high frame rate imaging technology and can capture real-time image data in front of the tool group at a speed of at least 30 frames per second during operation. The camera samples images multiple times and combines motion tracking algorithms to record the movement trajectory of the mushroom grass stems when they approach the tool group.

[0052] When recording the motion trajectory of the Juncao stems approaching the tool group, the motion tracking algorithm uses the optical flow method. The optical flow method calculates the object's motion speed and direction through the motion changes of pixels between consecutive image frames, and tracks the motion trajectory of the Juncao stems in real time.

[0053] The optical flow assumes that the pixel brightness remains unchanged in two consecutive frames of images, and the following optical flow equation is derived:

[0054] ,in, Indicates at time When , the image coordinates are The pixel brightness at For pixels in The displacement within a time interval, For time, is the time step;

[0055] By Taylor expansion and ignoring higher-order terms, we can obtain:

[0056] ,in:

[0057] The pixel is Speed ​​in direction;

[0058] The pixel is Speed ​​in direction.

[0059] In the present invention, the optical flow method uses a camera to capture continuous image frames and record the movement trajectory of the grass stems in front of the tool set. The pixel points of the grass stems in each frame of the image will be tracked in real time. and The displacement between , and obtain the motion trajectory of the stems during the harvesting process.

[0060] The specific motion patterns of Juncao stems identified by convolutional neural network include:

[0061] Displacement obtained based on optical flow method Get the movement trajectory of the Juncao stem , as the input of the convolutional neural network;

[0062] The convolutional neural network extracts patterns from the input motion trajectory data through multiple convolutional layers, pooling layers, and fully connected layers. Each frame of image data and the motion changes therein are processed by multiple convolution kernels to extract the motion features of the Juncao stems. The motion features include:

[0063] Changes in the contact frequency and position between the Juncao stems and the cutter set;

[0064] The track of the repeated movement of the stems of the Juncao;

[0065] The changing trend of the speed and direction of the movement of the Juncao stems;

[0066] The feature map after convolution and pooling is used to identify complex motion patterns, and finally outputs a high-dimensional feature vector F, which represents the motion pattern of the stems of Juncao.

[0067] In convolutional neural networks, motion trajectory data can be discretized into frame data input. Assume that the motion trajectory segment captured by the camera at time t is a set of continuous frame sequences. , where each frame of the image contains the location information of the Juncao;

[0068] The convolutional neural network extracts motion features from the input image sequence. The specific process is:

[0069] 1. Input layer: frame sequence As input, it is represented as ,in, and are the image height and width, is the time dimension, is the number of channels ( );

[0070] 2. Convolution layer: The formula for the convolution operation is:

[0071] ,in, is the output feature map elements, is the weight matrix of the convolution kernel, is the input image sequence, The bias,convolutional layer extracts low-level features of the movement of the grass stems, such as edges, directions, and displacements; is the input image sequence, is the input image located at The pixel value at position In the present invention, the input image comes from a continuous frame sequence of the movement of the grass stem captured by the camera, and the movement pattern of the stem, such as the collision and rebound behavior in front of the tool group, is extracted from these images through convolution operation.

[0072] 3. Pooling layer: The pooling layer is used to downsample the feature map. The formula is:

[0073] ,in, is the output of the pooling operation;

[0074] is the pooling window size, ,The pooling layer compresses the size of the feature map, improving computational efficiency while retaining important motion features, is the input pooling window, which represents the input image from arrive and from arrive The pixel values ​​in this area are collected and used for pooling operation. The pooling window covers the local area where the grass stems move, and is used to extract the most significant motion features in the area to help with subsequent rebound effect detection.

[0075] 4. Fully connected layer: The motion features extracted by the convolution and pooling layers are finally passed through the fully connected layer to represent the motion trajectory features as a high-level feature vector , the output of the fully connected layer is expressed as: ,in, is the weight matrix, is the bias and ReLU is the activation function.

[0076] Judging whether a rebound effect occurs in the stems of the mushroom grass includes extracting motion features based on a convolutional neural network and using a double standard to judge whether a rebound effect occurs.

[0077] The double standards include that the stems of the mushroom grass contact the tool set multiple times without being completely cut, and that the stems of the mushroom grass move repeatedly within a fixed range in front of the tool set.

[0078] The criteria for the Juncao stems to not be completely cut by the knife set after multiple contacts include:

[0079] The motion characteristics analyzed by the convolutional neural network are used to detect whether the stems of the mushroom grass have frequent short-term contact behavior in front of the tool group. If the motion characteristics show that the mushroom grass contacts a certain area of ​​the tool group multiple times in a short period of time but is not cut by the tool group, it is regarded as a rebound effect. The number of contacts, that is, the number of rebounds N, is calculated: N>Nthreshold. If the number of rebounds exceeds the threshold Nthreshold, it is judged that the mushroom grass has not been cut and there is a rebound.

[0080] The criteria for the repeated movement of the stems of the grass within a fixed range in front of the cutter group include:

[0081] The motion trajectory extracted by the convolutional neural network is used to detect whether the stems of the Juncao move repeatedly within a fixed spatial range. This behavior is usually related to the rebound effect. The rebound is judged by the periodic changes and displacement data of the motion trajectory. If the trajectory moves back and forth within a fixed range, it is a rebound, and the rebound frequency M exceeds the threshold Mthreshold:

[0082] M>Mthreshold and |R(t1,t2)|<Rthreshold, it is judged that the Juncao has a rebound effect.

[0083] R(t1, t2) represents the displacement (motion trajectory range) of the Juncao stem between time t1 and t2. In other words, it represents the range or distance of movement of the Juncao in front of the tool group during this period. |R(t1, t2)| is the absolute value of the Juncao movement range R(t1, t2), which represents the physical distance that the Juncao moves during this period. If the Juncao repeatedly collides in front of the tool group and does not enter the tool group, this value will be relatively small, indicating that its movement range is limited to a certain area. Rthreshold is the preset movement range threshold.

[0084] The rebound effect scoring module is based on the number of rebounds. and rebound frequency Conduct a rebound effect score, including:

[0085] By detecting the rebound contact times of the stems of the grass in front of the cutting tool group and the rebound frequency per unit time , calculate the rebound effect score , the scoring formula is: ,in, and are weight parameters used to adjust the impact of rebound times and rebound frequency on the score; score The higher the score, the more rebounding grass stems there are and the more severe the rebound effect is. The working depth of the tool set is adjusted according to the score to cope with different degrees of rebound effect.

[0086] The motion trajectory calculation is expressed as:

[0087] ,in, Indicates that the stems of Juncao are The two-dimensional spatial position at the time of the movement, describing the movement trajectory of the Juncao stem from the initial position to the current moment, and It's time The speed of movement at any moment, and Respectively represent the time The initial position coordinates at the time of the start of monitoring are the initial spatial position of the Juncao in the camera field of view. In the present invention, the camera starts recording the movement trajectory of the Juncao from a specific position before the Juncao enters the tool group. The initial position is the starting point for judging the rebound effect. and Respectively represent the time Always Direction and The velocity components in the direction, in the present invention, are calculated from the continuous image frames by the optical flow method, and are used to analyze the change in the movement speed of the grass in front of the tool group. If the speed of the grass changes repeatedly, it indicates the presence of a rebound effect; and Respectively represent the time Internal and The cumulative displacement in the direction, Indicates that Juncao Displacement in direction, Indicates that Juncao Displacement in direction; in the present invention, these accumulated displacements are used to depict the complete motion trajectory of the Juncao stem to detect whether it moves or contacts repeatedly in front of the tool group.

[0088] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0089] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. Juncao adaptive harvesting depth control system, characterized by: The following steps are involved: Camera monitoring module: The camera is installed on the side wall of the cutter group protective cover to capture the initial image data of the harvesting situation in front of the cutter group in real time and record the movement trajectory of the Juncao stems; Rebound effect recognition module: Based on the motion trajectory recorded in the initial image data, the convolutional neural network is used to identify the motion characteristics of the Juncao stems to determine whether the Juncao stems have a rebound effect; Rebound effect scoring module: Based on the results of convolutional neural network recognition, the number of rebounds and the frequency of rebounds are recorded, and the rebound effect is scored according to the number of rebounds and the frequency of rebounds. The higher the score, the more stems of the rebounding grass are, and the more serious the rebound effect is. Depth adaptive adjustment module: dynamically adjusts the harvesting depth of the tool group according to the rebound effect scoring result, specifically including: When the rebound effect score exceeds the preset score threshold, the cutter group performs bidirectional depth adjustment, that is, repeatedly adjusts between deep and shallow depths to ensure that the rebounded grass stems enter the cutter group smoothly; When the rebound effect score is lower than the preset score threshold, the tool group only performs depth adjustment in one direction until the rebound effect mushroom grass stem enters the tool group and returns to the original depth.

2. The adaptive harvesting depth control system for Juncao according to claim 1 is characterized in that: The camera adopts an industrial camera and is installed on the side wall of the tool group protective cover through a fixed bracket. The camera records the movement trajectory of the mushroom grass stem when it approaches the tool group by sampling images multiple times and combining with a motion tracking algorithm.

3. The adaptive harvesting depth control system of Juncao according to claim 2 is characterized in that: The motion tracking algorithm adopts the optical flow method, which captures continuous image frames through a camera to record the motion trajectory of the grass stems in front of the tool set. The pixel points of the grass stems in each frame of the image will be tracked in real time by calculating the pixel points between two frames. and The displacement between , and obtain the motion trajectory of the stems during the harvesting process.

4. The adaptive harvesting depth control system for Juncao according to claim 3 is characterized in that: The method of identifying the movement pattern of the stems of Juncao by using a convolutional neural network specifically includes: Displacement obtained based on optical flow method Get the movement trajectory of the Juncao stem , as the input of the convolutional neural network; The convolutional neural network extracts patterns from the input motion trajectory data through multiple convolutional layers, pooling layers, and fully connected layers. Each frame of image data and the motion changes therein are processed by multiple convolution kernels to extract the motion features of the Juncao stems. The motion features include: Changes in the contact frequency and position between the Juncao stems and the cutter set; The track of the repeated movement of the stems of Juncao; The changing trend of the speed and direction of the movement of the Juncao stems; The feature map after convolution and pooling is used to identify the motion pattern, and finally outputs a high-dimensional feature vector F, which represents the motion pattern of the Juncao stem.

5. The adaptive harvesting depth control system for Juncao according to claim 4 is characterized in that: The method of judging whether a rebound effect occurs in the stems of the mushroom grass includes judging whether a rebound effect occurs by using a double standard based on motion features extracted by a convolutional neural network.

6. The adaptive harvesting depth control system for Juncao according to claim 5 is characterized in that: The double standards include that the mushroom grass stems contact the cutter group multiple times without being completely cut, and the mushroom grass stems move repeatedly within a fixed range in front of the cutter group.

7. The adaptive harvesting depth control system for Juncao according to claim 6 is characterized in that: The criteria for the Juncao stems to be not completely cut by the cutting tool set after multiple contacts include: The motion characteristics analyzed by the convolutional neural network are used to detect whether the stems of the mushroom grass have frequent short-term contact behavior in front of the tool group. If the motion characteristics show that the mushroom grass contacts a certain area of ​​the tool group multiple times in a short period of time but is not cut by the tool group, it is regarded as a rebound effect. The number of contacts, that is, the number of rebounds N, is calculated: N>Nthreshold. If the number of rebounds exceeds the threshold Nthreshold, it is judged that the mushroom grass has not been cut and there is a rebound.

8. The adaptive harvesting depth control system for Juncao according to claim 7, characterized in that: The criteria for the repeated movement of the stems of the grass within a fixed range in front of the cutter group include: The motion trajectory extracted by the convolutional neural network is used to detect whether the stems of the Juncao move repeatedly within a fixed spatial range. The rebound is judged by the periodic changes and displacement data of the motion trajectory. If the trajectory moves back and forth within a fixed range, it is a rebound, and the rebound frequency M exceeds the threshold Mthreshold: M>Mthreshold and |R(t1,t2)|<Rthreshold, it is judged that the rebound effect of Juncao occurs. R(t1, t2) represents the displacement of the Juncao stem between time t1 and t2, and |R(t1, t2)| is the absolute value of the Juncao movement range R(t1, t2).

9. The adaptive harvesting depth control system for Juncao according to claim 8, characterized in that: The rebound effect scoring module is based on the number of rebounds. and rebound frequency Conduct a rebound effect score, including: By detecting the rebound contact times of the stems of the grass in front of the cutting tool group and the rebound frequency per unit time , calculate the rebound effect score , the scoring formula is: ,in, and are weight parameters used to adjust the impact of rebound times and rebound frequency on the score; score The higher the score, the more rebounding grass stems there are and the more severe the rebound effect is. The working depth of the tool set is adjusted according to the score to cope with different degrees of rebound effect.

10. The adaptive harvesting depth control system of Juncao according to claim 3, characterized in that: The motion trajectory calculation is expressed as: ,in, Indicates that the stems of Juncao are The two-dimensional spatial position at the time of the movement, describing the movement trajectory of the Juncao stem from the initial position to the current moment, and It's time The speed of movement at any moment, and Respectively represent the time The initial position coordinates at time ; and Respectively represent the time Always Direction and Directional velocity component; and Respectively represent the time Internal and The cumulative displacement in the direction, Indicates that Juncao Displacement in direction, Indicates that Juncao Displacement in direction.

Images