Tip cutting and sugarcane tip prediction method and tip cutting device of sugarcane harvesting machine

Abstract

The invention discloses a tip cutting and sugarcane tip prediction method for a sugarcane harvester, and the method comprises the steps: (1), adjusting the initial position of a camera after the position of the camera is obtained, and adjusting the initial position of the camera to one side without sugarcane interference; (2) obtaining sugarcane tip position data through a camera, processing the data collected in a camera picture by using a sugarcane tip identification model, and preliminarily judging whether a cutter head is too high or too low by using the obtained sugarcane tip position information; (3) performing control coarse adjustment on a tip cutting device through a control algorithm of a single-chip microcomputer according to the position information of the sugarcane tips; and (4) acquiring the position of the sugarcane tip by using a camera, acquiring accurate sugarcane tip position information from the data through a sugarcane tip identification model, judging according to the sugarcane tip identification model and a prediction result, and finely adjusting the tip cutting device through a control algorithm of a single chip microcomputer. According to the method, the impurity rate and the loss rate of the sugarcane harvester during tip cutting can be reduced, and meanwhile a large amount of manual labor is saved.

Description

Technical Field

[0001] The invention relates to the technical field of agricultural machinery, and in particular to a tip cutting and sugarcane tip prediction method and a tip cutting device of a sugarcane harvester. Background Art

[0002] Mechanized sugarcane topping refers to the use of mechanical equipment to complete the removal of sugarcane tips, which is an important part of the sugarcane harvesting process. Mechanized topping can improve efficiency, reduce manual labor intensity, and facilitate subsequent harvesting and processing. In the mechanized harvesting process, it is crucial to accurately find the sugarcane tips for cutting. If the cutting point is higher than the growth point of the sugarcane, a large amount of sugarcane heads with low sugar content will be included, resulting in an increase in the overall impurity rate, affecting the sugar mill's acceptance of sugarcane. On the contrary, if the cutting point is too low, a large amount of valuable sugarcane parts will be wasted, reducing the income of growers.

[0003] Through research, it was found that the growth point of sugarcane is usually located about 35-40cm below the tip of the sugarcane. However, the distribution of sugarcane in the sugarcane field is dense and chaotic, which poses a huge challenge to the dynamic identification of sugarcane. The existing tip cutting device of the sugarcane harvester has problems such as inflexible adjustment of the tip information collection equipment, inaccurate data collection, or inaccurate cutting position when adjusting the cutter disc for tip cutting.

[0004] The information disclosed in this background technology section is only intended to enhance the understanding of the overall background of the invention and should not be regarded as an acknowledgment or any form of suggestion that the information constitutes the prior art already known to a person skilled in the art. Summary of the invention

[0005] The object of the present invention is to provide a sugarcane tip cutting and sugarcane tip prediction method for a sugarcane harvester, thereby overcoming the shortcomings of inaccurate adjustment control position of the sugarcane harvester cutter disc and inflexible position adjustment of the camera used for collecting sugarcane tip information.

[0006] Another object of the present invention is to provide a tip cutting device for a sugarcane harvester to solve the problems that the sugarcane harvesting operation fails to react in time, resulting in the inability of the harvester's blade to be reasonably adjusted to the sugarcane growth point, and the impurity rate and loss rate of the harvester's tip cutting are high.

[0007] To achieve the above object, the present invention provides a sugarcane harvester tip cutting and sugarcane tip prediction method, comprising:

[0008] (1) After obtaining the camera position, adjust the initial position of the camera to the side without sugarcane interference;

[0009] (2) Obtaining sugarcane tip position data through a camera, using a sugarcane tip recognition model to process the data collected in the camera image, and using the obtained sugarcane tip position information to preliminarily determine whether the cutter disc is too high or too low;

[0010] (3) According to the position information of the sugarcane tip, the tip cutting device is roughly controlled through the control algorithm of the single chip microcomputer;

[0011] (4) The sugarcane tip position obtained by the camera is then used to obtain accurate sugarcane tip position information through a sugarcane tip recognition model. The tip recognition model and the predicted results are used to make a judgment, and the tip cutting device is fine-tuned through the control algorithm of the single-chip microcomputer, and then the tip is cut.

[0012] Preferably, in the above technical solution, the image information collected by the camera in step (2) includes sugarcane tip images and video information, and the camera is used to record or shoot images under different lighting conditions in the sugarcane field to obtain sugarcane tip image samples and obtain sugarcane tip position coordinate information.

[0013] Preferably, in the above technical solution, the sugarcane tip recognition model includes optimizing the backbone network based on the YOLOV8 architecture, and adding an LSTM long short time sequence memory algorithm to the backbone network to predict the position of the sugarcane tip in front.

[0014] Preferably, in the above technical solution, the single-chip microcomputer control algorithm is to control the top cutting device and adjust the position of the top cutting device by burning the control algorithm written in the host computer into the single-chip microcomputer; the data obtained by the single-chip microcomputer through the sensor can also be sent back to the host computer through the serial port to monitor and feedback the working status of the sugarcane harvester.

[0015] Preferably, in the above technical solution, the training process of the sugarcane tip recognition model includes: photographing sugarcane in clusters in the field, and using the polygon annotation of Lambelme annotation software to mark the sugarcane tips and sugarcane leaves as two types of labels for instance segmentation, wherein the sugarcane tips are used as the main object of recognition function training, and the sugarcane leaves are used as the main object of prediction function training; then, the Lambelme tool is used to generate a labeling file, and the file content includes the labeling points, the connection information between the labeling points, and the category information; the two types of labels are used as training sets with different functions for training, and finally, the two types of labeled data sets are randomly allocated in proportion to generate training sets, test sets, and validation sets, and the recognition model with the highest accuracy is selected through cross-validation, and then the accuracy, recall rate, average precision, mean square error, and mean absolute error are used as evaluation indicators of the prediction model.

[0016] A control system for tip cutting and sugarcane tip prediction of a sugarcane harvester, using the above method for prediction, the system comprises:

[0017] An image recognition unit, including a camera and a sugarcane tip recognition model, acquires a sugarcane tip image and world coordinate information through the camera, and the sugarcane tip recognition model processes the data acquired by the camera;

[0018] a tip cutting device, which is installed on the sugarcane harvester and is used to cut sugarcane tips; and

[0019] A control unit is connected with the image recognition unit and the cane cutting device data. The control unit adjusts the position of the camera and controls the cane cutting device to cut the cane tips according to the data of the image recognition unit.

[0020] A control method for a control system for cutting and predicting sugarcane tips of a sugarcane harvester comprises the following steps:

[0021] (1) inputting the sugarcane tip recognition model of the image recognition unit into the control unit;

[0022] (2) extracting the characteristics of the sugarcane tip and its world coordinate information through the camera and transmitting the data to the control unit;

[0023] (3) The control unit obtains the position information of the cutting device, and adjusts the height of the cutting device by comparing the relative position of the cutting device with the world coordinates of the sugarcane tip, so that the height of the cutting device is adjusted to be 35-40 cm higher than the height of the identified sugarcane tip growth point, and controls the cutting device to perform the cutting work.

[0024] A tip cutting device for a sugarcane harvester uses the above method to perform tip cutting and sugarcane tip prediction, the tip cutting device comprising: a mounting frame, which is movably connected to the sugarcane harvester through at least one supporting column, the mounting frame is provided with a rotating member, the rotating member is connected to one end of a camera support arm, the other end of the camera support arm is provided with a pan-tilt platform, and the pan-tilt platform is provided with a camera; a first driving mechanism, which is provided on the sugarcane harvester and the other end of which is connected to the mounting frame to drive the mounting frame to move up and down; a lifting mechanism, which is provided at the front end of the mounting frame; a knife disc, which is provided on the lifting mechanism, and the upper and lower positions of the knife disc are adjusted by the lifting mechanism; and a sugarcane gathering frame, which is provided at the front end of the mounting frame and is located below the knife disc, the sugarcane gathering frame comprising two gathering rods arranged opposite to each other, and the distance between the front ends of the two gathering rods is greater than the distance between the rear ends.

[0025] Preferably, in the above technical solution, the rotating member is a rotating cylinder, one end of the camera support arm is connected to the rotating end of the rotating cylinder, and the camera support arm is driven to rotate by the rotating cylinder to adjust the angle.

[0026] Preferably, in the above technical solution, the lifting mechanism is a screw mechanism, and the screw mechanism includes: a servo motor, which is arranged on the mounting frame; a screw, which is connected to the servo motor and is vertically arranged on the mounting frame; and a slide, which is arranged on the screw, and a pull-wire displacement sensor is provided on the slide, and the slide is connected to the cutter disc, and the pull-wire displacement sensor is used to obtain the position information of the cutter disc; wherein, the servo motor drives the screw to rotate, drives the slide to move up and down in the vertical direction, and the movement of the slide drives the cutter disc to move up and down.

[0027] Preferably, in the above technical solution, the first driving mechanism is a hydraulic mechanism or a cylinder mechanism.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] (1) The sugarcane tip cutting and sugarcane tip prediction method of the sugarcane harvester of the present invention improves the traditional sugarcane harvester in terms of physical structure and control logic, uses a binocular camera to identify sugarcane tips, and also uses a rotating mechanism to adjust the camera position. The depth map obtained by the camera is used to determine whether the camera is on the side where the sugarcane has not been harvested, thereby avoiding the camera being in an unnecessary complex environment to obtain accurate sugarcane tip images.

[0030] (2) The method of the present invention is a method for predicting the position of the sugarcane tip of a sugarcane harvester based on the LSTM model. The method combines the growth characteristics of sugarcane with the LSTM algorithm, improves the original recognition model based on YOLOv8, takes into account the growth characteristics of the branching point of the sugarcane tip, and starts from the branching growth morphology of the sugarcane leaves at the branching point of the sugarcane tip and the growth trend of being wide in the middle and narrow at both ends. The distance, thickness, and growth trend of the sugarcane leaves obtained by the recognition model are used to predict the position of the sugarcane tip in real time, thereby further improving the recognition accuracy of the model.

[0031] (3) The sugarcane cutting tip identification system of the present invention adds the pan-tilt anti-shake technology, obtains the camera's posture information through the built-in gyroscope and inertial sensor of the pan-tilt, and calls the brushless motor to adjust the camera's posture in real time, thereby playing the role of image anti-shake. The camera image can stably identify the sugarcane tip and obtain its position more accurately, so that the control unit can adjust the cutter disc more accurately when controlling.

[0032] (4) The control system and control method for tip cutting and sugarcane tip prediction of the present invention have two levels of coordination in the control process, namely, coarse adjustment and fine adjustment, which ensure that the sugarcane tip can be identified in the camera image at all times, thereby ensuring the efficiency and accuracy of tip cutting. Meanwhile, the accuracy of the identification system and the adjustment speed of the control system are further improved through the prediction algorithm and the blade adjustment strategy. This is of great significance for reducing the impurity rate of the harvester when cutting the tip, saving a large amount of manual labor, further improving the level of intelligent agricultural machinery, and promoting the development of agricultural mechanization and intelligence. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 is a flow chart of a sugarcane tip cutting and sugarcane tip prediction method of a sugarcane harvester according to the present invention;

[0034] Figure 2 is a schematic structural diagram of a tip cutting device according to the present invention;

[0035] Figure 3 is a schematic diagram of the position of the camera when the topping device according to the present invention is performing harvesting work in the forward process;

[0036] Figure 4 is a schematic diagram of the structure of the camera and the gimbal installation according to the present invention;

[0037] Figure 5 The present invention is a schematic structural diagram of the connection between the mounting frame and the lifting mechanism in the top cutting device.

[0038] Figure 6 It is a schematic diagram of the structure of the cutter disc in the topping device according to the present invention. DETAILED DESCRIPTION

[0039] The specific implementation modes of the present invention are described in detail below in conjunction with the accompanying drawings, but it should be understood that the protection scope of the present invention is not limited by the specific implementation modes.

[0040] Unless explicitly stated otherwise, throughout the specification and claims, the term “comprise” or variations such as “include” or “comprising”, etc., will be understood to include the stated elements or components but not to exclude other elements or components.

[0041] like Figure 1 As shown, a method for cutting and predicting sugarcane tips of a sugarcane harvester according to a specific embodiment of the present invention comprises the following steps:

[0042] Step 1: Use a camera to collect image information of sugarcane tips in the sugarcane field

[0043] First, the camera detects the environment position before the operation, shoots the surrounding environment in front of the camera, and uses the depth map obtained from the shooting to determine whether it is currently on the side of the sugarcane field that has been harvested.

[0044] The image information collected by the camera includes sugarcane tip images and video information. The camera records or shoots images under different lighting conditions in the sugarcane field to obtain sugarcane tip image samples and obtain the sugarcane tip position coordinate information.

[0045] If the camera is on the side where the harvest has not been completed, the microcontroller controls the rotating cylinder to deflect the camera to the side of the harvested sugarcane field, and then performs environmental position detection again. If the camera position adjustment is completed, you can prepare to start harvesting. The rotating cylinder is controlled by the stm32 microcontroller, and the camera is 70-80 cm away from the sugarcane tip. The recognition effect is best when the rotating arm is vertical to the ground.

[0046] After determining the position of the camera, control the gimbal to lock the current position in front of the camera and play an anti-shake role during operation.

[0047] Step 2: Obtain the sugarcane tip position data through the camera, use the sugarcane tip recognition model to process the data collected in the camera image, and use the obtained sugarcane tip position information to preliminarily determine whether the cutter disc is too high or too low.

[0048] Step 3: According to the position information of the sugarcane tops, the top cutting device is roughly controlled through the intelligent control algorithm based on the STM32 microcontroller.

[0049] The camera is used to record or shoot images of sugarcane tips in the sugarcane field to obtain sample pictures and information of sugarcane tips. The collected images and video information of sugarcane tips will be used as its image information. When there are too few images of sugarcane tips in the camera screen, or the positions of most sugarcane tips are too far from the median line of the screen, the overall position of the tip cutting device is adjusted to indirectly adjust the camera screen to ensure that there are sugarcane tips that can be cut within the range of the cutter. The adjustment position of the tip cutting device depends on the statistical average height of the overall outline of the sugarcane tips in the screen.

[0050] Step 4: Reuse the sugarcane tip position obtained by the camera, and obtain accurate sugarcane tip position information through the sugarcane tip recognition model. Make a judgment based on the sugarcane tip recognition model and the prediction results, and fine-tune the tip cutting device through the intelligent control algorithm based on the STM32 microcontroller, and then cut the tip.

[0051] The STM32 microcontroller control algorithm is to control the top cutting device and adjust the position of the top cutting device by burning the control algorithm written in the host computer into the microcontroller; the data obtained by the microcontroller through the sensor can also be sent back to the host computer through the serial port to monitor and feedback the working status of the sugarcane harvester.

[0052] The sugarcane tip recognition model adopts the target detection and tracking model. The target detection and prediction function adopts the YOLOV8+LSTM prediction model, which has the characteristics of high recognition rate, fast recognition and prediction speed, and high prediction accuracy. The backbone network is optimized based on the YOLOV8 architecture, and the LSTM prediction model is added to the backbone network. The sugarcane tip position obtained by the camera is used to predict the sugarcane tip using the LSTM algorithm based on the direction and distance data of the sugarcane leaves.

[0053] LSTM (Long Short-Term Memory) is a special recurrent neural network (RNN) used to process and predict time series data. Unlike traditional RNN, LSTM solves the common gradient vanishing and gradient exploding problems in long sequence data by designing a set of complex gating mechanisms. Through the direct path of the cell state, information can be effectively propagated in multiple time steps, which can alleviate the gradient vanishing problem and capture the dependencies between long time steps. It is suitable for long sequence data. Its core is its cell state and three gates, namely forget gate, input gate and output gate. These gates control the flow of information in the cell state through different weight matrices and activation functions.

[0054] The forget gate determines which information in the cell state needs to be forgotten. It combines the hidden state of the previous time step with the current input through a Sigmoid function:

[0055]

[0056] Among them, f t It is the output of the forget gate, indicating the proportion of old information retained, and its value range is between 0 and 1. f is the weight matrix, b f is the bias, h t-1 is the hidden state of the previous time step, x t is the current input, σ is the Sigmoid function, which limits the output to between 0 and 1.

[0057] The input gate determines what new information needs to be added to the cell state. It consists of two parts: a Sigmoid layer that determines which values ​​need to be updated, and a Tanh layer that generates a new candidate value vector:

[0058] i t =σ(W i ·[h t-1 ,x t ]+b i ) (2)

[0059] The cell state is updated by combining the results of the forget gate and the input gate.

[0060]

[0061] Where * indicates element-by-element multiplication. tIt is the output of the input gate, which is used to decide which new information will be updated into the memory. is the new candidate memory, and the tanh function is used to limit its range to -1 to 1. t is the current state of the memory unit. t-1 is the state of the memory cell at the previous moment. Wi and bi are the weight matrix and bias term of the input i of the input gate, and Wc and bc are the weight matrix and bias term of the new candidate memory C.

[0062] The output gate determines the hidden state of the current time step:

[0063] o t =σ(W o ·[h t-1 ,x t ]+b o ) (5)

[0064] h t =o t *tanh(C t ) (6)

[0065] Among them, W o is the weight matrix, b o is the bias term.

[0066] o t It is the output of the output gate, and the sigmoid function is used to control the proportion of output from the memory unit.

[0067] h t is the hidden state at the current moment, which is passed to the next time step or the final output.

[0068] C t is the state of the memory cell at the current moment, which is scaled by the tanh function so that the output range is between -1 and 1.

[0069] σ is the sigmoid function, which limits the output to between 0 and 1.

[0070] The training process of the sugarcane tip recognition model includes: photographing sugarcane in clusters in the field, and using the polygon annotation of Lambelme annotation software to mark the sugarcane tips and leaves as two types of labels for instance segmentation. The sugarcane tips are used as the main object of recognition function training, while the sugarcane leaves are used as the main object of prediction function training. Then, the Lambelme tool is used to generate an annotation file, which contains the annotation points, the connection information between the annotation points, and the category information. The two types of annotations are used as training sets with different functions for training. Finally, the two types of annotated data sets are allocated in proportion to generate 60% training sets, 20% test sets, and 20% validation sets. The recognition model with the highest accuracy is selected through cross-validation, and the accuracy, recall, average precision, mean square error, and mean absolute error are used as evaluation indicators of the prediction model.

[0071] The labeling method is as follows: First, use Lambelme labeling software to perform instance segmentation, and use polygonal labeling boxes to label the sugarcane tips. All sugarcane tips are labeled with the same label. Then, use the Lambelme tool to generate a labeling file, and the file content contains the labeling points, the connection information between the labeling points, and the category information. Finally, the labeled data set is randomly distributed in proportion. The present invention uses accuracy (Precision), recall rate (Recal1), average precision (mAP), mean square error (MSE), and mean absolute error (MAE) as evaluation indicators. The accuracy of the network model is required to be ≥93%.

[0072] Precision represents the detection accuracy, which means the ratio of the number of correctly predicted samples to the actual number of correct samples. TP (True Positive) is the true positive example, which is the number of samples correctly predicted as positive by the model. FP (False Positive) is the false positive example, which is the number of samples incorrectly predicted as positive by the model.

[0073]

[0074] Recall represents the recall rate, which is the ratio of the number of correct samples predicted by the model to the total number of samples, where TP represents the number of samples predicted as positive. , FN represents the number of samples predicted as negative.

[0075]

[0076] mAP stands for mean average precision, which is the sum of the average precision of all categories divided by the average precision of all categories in the dataset. The formula is shown in (9). The value of AP is equal to the area of ​​the curve enclosed by Precision and Recal1.

[0077]

[0078] MSE stands for mean square error, which refers to the average of the squares of the differences between the predicted value and the actual value. It is often used in regression problems. Its formula is shown in (10). Where y_i is the actual value, is the predicted value, and n is the number of samples.

[0079]

[0080] MAE stands for mean absolute error, which refers to the average value of the absolute value of the difference between the predicted value and the actual value. Its formula is shown in (11). Where y_i is the actual value, is the predicted value, and n is the number of samples.

[0081]

[0082] A control system for tip cutting and sugarcane tip prediction of a sugarcane harvester, using the above method for prediction, the system comprises:

[0083] An image recognition unit, including a camera and a sugarcane tip recognition model, acquires a sugarcane tip image and world coordinate information through the camera, and the sugarcane tip recognition model processes the data acquired by the camera;

[0084] a tip cutting device, which is installed on the sugarcane harvester and is used to cut sugarcane tips; and

[0085] A control unit is connected with the image recognition unit and the cane cutting device data. The control unit adjusts the position of the camera and controls the cane cutting device to cut the cane tips according to the data of the image recognition unit.

[0086] A control method for a control system for cutting and predicting sugarcane tips of a sugarcane harvester comprises the following steps:

[0087] (1) inputting the sugarcane tip recognition model of the image recognition unit into the control unit;

[0088] (2) extracting the characteristics of the sugarcane tip and its world coordinate information through the camera and transmitting the data to the control unit;

[0089] (3) The control unit obtains the position information of the cutting device, and adjusts the height of the cutting device by comparing the relative position of the cutting device with the world coordinates of the sugarcane tip, so that the height of the cutting device is adjusted to be 35-40 cm higher than the height of the identified sugarcane tip growth point, and controls the cutting device to perform the cutting work.

[0090] like Figure 2-4 As shown, the structure of the cutting device of the sugarcane harvester of the present invention is as follows:

[0091] A tip cutting device for a sugarcane harvester uses the above method to perform tip cutting and sugarcane tip prediction. The tip cutting device includes a mounting frame 1, which is movably connected to the sugarcane harvester through two support columns 11. A camera assembly 2 is installed on the mounting frame 1. The camera assembly 2 includes a rotating member, which is a rotating oil cylinder 21, and the rotating oil cylinder 21 is fixed to the mounting frame 1 through screws and nuts. The rotating end of the rotating oil cylinder 21 is connected to one end of a camera support arm 22, and the camera support arm 22 is driven to rotate by the rotating oil cylinder 21 to adjust the angle. A pan-tilt 23 is provided at the other end of the camera support arm 22, and a camera 24 is installed on the pan-tilt 23. The pan-tilt anti-shake technology is added to the sugarcane tip cutting recognition system, and the camera's posture information is obtained through the built-in gyroscope and inertial sensor of the pan-tilt, and the brushless motor is called to adjust the camera's posture in real time, thereby playing the role of image anti-shake. The camera image can stably identify the sugarcane tip and obtain its position more accurately, so that the control unit can adjust the cutter disc more accurately when controlling.

[0092] The sugarcane harvester is provided with a first driving mechanism, preferably, the first driving mechanism is a hydraulic mechanism or a cylinder mechanism. In this embodiment, a hydraulic mechanism 3 is selected, and the front end of the piston rod of the hydraulic mechanism 3 is hinged with the support column 11, and the telescopic rod of the hydraulic mechanism 3 is telescopically driven to drive the mounting frame 1 to move up and down. A lifting mechanism 4 is provided at the front end of the mounting frame 1, and a knife disc 5 is provided on the lifting mechanism 4, and the up and down position of the knife disc is adjusted by the lifting mechanism 4. The hydraulic mechanism 3 drives the mounting frame 1 to move up and down to roughly adjust the knife disc 5, and the lifting mechanism 4 is used to fine-tune the knife disc 5. A sugarcane gathering frame 6 is provided at the front end of the mounting frame 1, and the sugarcane gathering frame 6 is located below the knife disc 5. The sugarcane gathering frame 6 includes two gathering rods 61 arranged opposite to each other, and the distance between the front ends of the two gathering rods 61 is greater than the distance at the rear ends. The sugarcane gathering frame 6 is arranged in a trumpet shape, and the sugarcane is placed between the sugarcane gathering frames 6 when cutting the tops to support the sugarcane.

[0093] Preferably, the lifting mechanism 4 is a screw mechanism, which includes a servo motor 41, which is arranged on the mounting frame 1. The output end of the servo motor 41 is connected to a screw 42, which is arranged on the mounting frame 1 in a vertical direction. A slide 43 is provided on the screw 42, and a wire displacement sensor 44 is provided on the slide 43. The slide 43 is welded to the triangular support plate of the cutter disc 5. The wire displacement sensor 44 provided on the slide 43 is used to obtain the position information of the cutter disc.

[0094] The cutter disc is fine-tuned by the lifting mechanism 4 as follows: the servo motor 41 drives the screw rod 42 to rotate, driving the slide 43 to move up and down in the vertical direction, and the movement of the slide 43 drives the cutter disc 5 to move up and down. Fine-tune the position of the cutter disc 5. Use a servo motor to drive the screw rod to perform the knife adjustment operation. The speed of the servo motor is 50 revolutions per second, the stroke of the screw rod is 300 mm, and the slide displacement is ±10 mm per rotation, that is, the knife adjustment speed is 500 mm per second. When the sugarcane harvester is operating, the sugarcane mobile platform moves at a speed of 0.5 meters per second. By adjusting the rotation speed of the servo motor, the sugarcane mobile platform is driven to run at a speed of 0.5 meters per second. When the harvester runs at a speed of 0.5 meters per second, there is a cluster of sugarcane every approximately 0.5 meters. The knife adjustment system can perform one and a half back and forth knife adjustment operations for each cluster, thereby achieving continuous, accurate and fast knife adjustment.

[0095] The working process of the sugarcane harvester is as follows: before starting, the sugarcane harvester determines whether the current camera position is on one side of the sugarcane field 7 based on the depth map obtained by the camera. If so, the camera is rotated to the other side by rotating the oil cylinder. Then the sugarcane harvester starts to operate in the field at a speed of 0.5m / s. The camera always faces the sugarcane field through the pan-tilt to identify the sugarcane tip. When the sugarcane tip is detected, the screw mechanism will make the cutter head move at a uniform speed in a straight line to a position 30-40cm higher than the sugarcane tip for fine adjustment.

[0096] The control process and the method for predicting the position of sugarcane tips include the following steps:

[0097] (1) The depth map of the image recognition unit and the network model for recognizing the morphological features of the sugarcane tip are input into the control unit, which is a single-chip microcomputer.

[0098] (2) The camera extracts the depth information of the sugarcane field, the growth trend characteristics of the sugarcane tips and their world coordinate information, and transmits the data to the control unit.

[0099] The image information captured by the camera when the sugarcane harvester is working is used as the data source. The computer connects the camera and the microcontroller. The camera extracts the growth trend characteristics of the sugarcane tips and their world coordinate information. The computer sends the world coordinates to the microcontroller through the serial port. Before starting work, the harvester makes a judgment based on the depth map information obtained by the camera. If the depth map shows that it is close to the sugarcane field, the camera position is adjusted by adjusting the rotating cylinder to deflect to the side without sugarcane.

[0100] (3) During operation, the harvester operates at a speed of 0.5 m / s in the field, and drives the screw rod by adjusting the servo motor, and then drives the cutter disc to harvest sugarcane. The control unit obtains the position of the cutting unit through the pull-wire displacement sensor, and compares the relative position of the cutting unit with the world coordinates to adjust the height of the cutting unit so that the height of the cutting unit is 30-40 cm higher than the world coordinate position of the sugarcane tip, and controls the cutting unit to cut the tip.

[0101] The camera and the single-chip computer are connected by a computer. The camera extracts the growth trend characteristics of the sugarcane tip and its world coordinate information. After the adjustment result of the cutter disc is obtained through the sugarcane tip recognition model and the prediction model, the adjustment result is sent to the serial port through the computer to the single-chip computer. The single-chip computer measures the relative displacement of the cutter disc through the pull-wire displacement sensor and compares the height with the adjustment. If there is no sugarcane tip in the camera screen, the extracted growth trend characteristics of the sugarcane tip are used to judge the position of the sugarcane tip through the prediction model. If most of the screen is sugarcane nodes, the hydraulic mechanism is lifted upward; if most of the screen is sugarcane leaves, the hydraulic mechanism is adjusted downward. Two methods are used for coarse adjustment. If the position of the cutter disc is 30-40 cm from the position of the sugarcane tip, the servo motor does not move. If the pull-wire displacement sensor data minus 30-40 cm> adjustment result, the servo motor rotates to make the cutter disc drop to the height of the pull-wire distance sensor data at the adjustment result. If the pull-wire distance sensor data plus 30-40 cm < adjustment result, the servo motor rotates to make the cutter disc drop to the same height as the pull-wire distance sensor data and the adjustment result, so as to achieve fine control of the cutter adjustment.

[0102] The foregoing description of specific exemplary embodiments of the present invention is for the purpose of illustration and demonstration. These descriptions are not intended to limit the present invention to the precise form disclosed, and it is clear that many changes and variations can be made based on the above teachings. The purpose of selecting and describing the exemplary embodiments is to explain the specific principles of the present invention and its practical application, so that those skilled in the art can realize and utilize various different exemplary embodiments of the present invention and various different selections and changes. The scope of the present invention is intended to be limited by the claims and their equivalents.

Claims

1. A sugarcane tip cutting and sugarcane tip prediction method for a sugarcane harvester, characterized in that: include: (1) After obtaining the camera position, adjust the initial position of the camera to the side without sugarcane interference; (2) Obtaining sugarcane tip position data through a camera, using a sugarcane tip recognition model to process the data collected in the camera image, and using the obtained sugarcane tip position information to preliminarily determine whether the cutter disc is too high or too low; (3) According to the position information of the sugarcane tip, the tip cutting device is roughly controlled through the control algorithm of the single chip microcomputer; (4) The sugarcane tip position obtained by the camera is then used to obtain accurate sugarcane tip position information through a sugarcane tip recognition model. The tip recognition model and the predicted results are used to make a judgment, and the tip cutting device is fine-tuned through the control algorithm of the single-chip microcomputer, and then the tip is cut.

2. The sugarcane tip cutting and sugarcane tip prediction method of a sugarcane harvester according to claim 1, characterized in that: The image information collected by the camera in step (2) includes sugarcane tip images and video information. The camera records or shoots images under different lighting conditions in the sugarcane field to obtain sample pictures of the sugarcane tips and obtain the sugarcane tip position coordinate information.

3. The sugarcane tip cutting and sugarcane tip prediction method of a sugarcane harvester according to claim 1, characterized in that: The sugarcane tip recognition model includes optimizing the backbone network based on the YOLOV8 architecture, and adding the LSTM long short time sequence memory algorithm to the backbone network to predict the position of the sugarcane tip in front.

4. The sugarcane tip cutting and sugarcane tip prediction method of a sugarcane harvester according to claim 1, characterized in that: The single-chip microcomputer control algorithm is to control the cutting device and adjust the position of the cutting device by burning the control algorithm written in the host computer into the single-chip microcomputer; the data obtained by the single-chip microcomputer through the sensor can also be sent back to the host computer through the serial port to monitor and feedback the working status of the sugarcane harvester.

5. The sugarcane tip cutting and sugarcane tip prediction method of a sugarcane harvester according to claim 1, characterized in that: The training process of the sugarcane tip recognition model includes: photographing sugarcane in clusters in the field, using the photographed photos to mark sugarcane tips and sugarcane leaves as two types of labels for instance segmentation through polygon annotation of Lambelme annotation software, wherein the sugarcane tips are used as the main object of recognition function training, and the sugarcane leaves are used as the main object of prediction function training; then using Lambelme tool to generate annotation files, the file content includes annotation points, connection information between annotation points and category information; the two types of annotations are used as training sets with different functions for training, and finally the two types of data sets that have been marked are randomly allocated in proportion to generate training sets, test sets and validation sets, and the recognition model with the highest accuracy is selected through cross-validation, and then the accuracy, recall rate, average precision, mean square error and mean absolute error are used as evaluation indicators of the prediction model.

6. A control system for sugarcane harvester tip cutting and sugarcane tip prediction, characterized in that: The method according to claim 1 is used to make predictions, wherein the system comprises: An image recognition unit, including a camera and a sugarcane tip recognition model, acquires a sugarcane tip image and world coordinate information through the camera, and the sugarcane tip recognition model processes the data acquired by the camera; a tip cutting device, which is installed on the sugarcane harvester and is used to cut sugarcane tips; and A control unit is connected with the image recognition unit and the cane cutting device data. The control unit adjusts the position of the camera and controls the cane cutting device to cut the cane tips according to the data of the image recognition unit.

7. A control method of the control system according to claim 6, characterized in that: The following steps are involved: (1) inputting the sugarcane tip recognition model of the image recognition unit into the control unit; (2) extracting the characteristics of the sugarcane tip and its world coordinate information through the camera and transmitting the data to the control unit; (3) The control unit obtains the position information of the cutting device, and adjusts the height of the cutting device by comparing the relative position of the cutting device with the world coordinates of the sugarcane tip, so that the height of the cutting device is adjusted to be 35-40 cm higher than the height of the identified sugarcane tip growth point, and controls the cutting device to perform the cutting work.

8. A sugarcane harvester toe cutting device, characterized in that: The method according to claim 1 is used to perform tip cutting and sugarcane tip prediction, wherein the tip cutting device comprises: A mounting frame, which is movably connected to the sugarcane harvester through at least one supporting column, the mounting frame is provided with a rotating member, the rotating member is connected to one end of a camera supporting arm, the other end of the camera supporting arm is provided with a pan platform, and the pan platform is provided with a camera; A first driving mechanism is provided on the sugarcane harvester, and the other end of the first driving mechanism is connected to the mounting frame to drive the mounting frame to move up and down; A lifting mechanism, which is arranged at the front end of the mounting frame; A cutter disc is arranged on the lifting mechanism, and the upper and lower positions of the cutter disc are adjusted by the lifting mechanism; and The sugarcane gathering frame is arranged at the front end of the mounting frame and is located below the knife disc. The sugarcane gathering frame comprises two gathering rods arranged opposite to each other, and the distance between the front ends of the two gathering rods is greater than the distance between the rear ends.

9. The sugarcane harvester topping device according to claim 8, characterized in that: The rotating member is a rotating cylinder, one end of the camera support arm is connected to the rotating end of the rotating cylinder, and the camera support arm is driven to rotate by the rotating cylinder to adjust the angle.

10. The sugarcane harvester topping device according to claim 8, characterized in that: The lifting mechanism is a screw mechanism, and the screw mechanism includes: A servo motor, which is arranged on the mounting frame; A screw rod, which is connected to the servo motor and is vertically arranged on the mounting frame; and A slide, which is arranged on the screw rod, a wire displacement sensor is arranged on the slide, the slide is connected to the cutter disc, and the wire displacement sensor is used to obtain the position information of the cutter disc; The servo motor drives the screw to rotate, thereby driving the slide to move up and down in the vertical direction, and the movement of the slide drives the cutter disc to move up and down.

Images