A metering system and method for precision planting of potatoes

Abstract

The application discloses a potato precision seeding metering system and method, and the metering system comprises a collecting mechanism arranged on the top of a seeding machine body and used for collecting images of a region to be detected, a control device arranged on a seeding mechanism and connected with the collecting mechanism, the control device is used for calculating and counting a clean seed qualified rate and a reseeding qualified rate, and a reseeding mechanism arranged behind the seeding mechanism and used for receiving a reseeding command of the control device and completing reseeding.

Description

Technical Field

[0001] This invention relates to the field of agricultural sowing, and in particular to a metering system and method for precision potato sowing. Background Technology

[0002] Potatoes are one of the world's most important food crops and are of great significance to the sustainable development of agriculture. Mechanized planting is an indispensable technical link in mechanized production and an important component of potato cultivation. Mechanized potato planting technology is an effective way to increase potato yield. However, in the process of mechanized potato planting, to clearly understand the effect of potato planting, manual statistics are still required, which is very tedious, time-consuming, and inefficient, and almost impossible to handle large-scale fields.

[0003] Although potato planters are available on the market and can greatly improve planting efficiency, they also face some technical challenges.

[0004] When there are missed plantings, it is difficult to count the missed planting rate; after replanting, it is difficult to count whether the replanting is qualified; and when planting, because the seed scoop may pick up multiple potatoes, there is a possibility of replanting, resulting in the same area being planted multiple times.

[0005] Currently, these problems can only be solved through manual statistics, but this approach is completely inapplicable when dealing with large-scale data.

[0006] To address the shortcomings of existing technologies, this invention proposes a metering system and method for precision potato planting. Summary of the Invention

[0007] The purpose of this invention is to overcome the shortcomings of the prior art and provide a metering system and method for precision potato planting.

[0008] First aspect

[0009] This invention proposes a metering method for precision potato planting, the steps of which include:

[0010] S1. Obtain the image of the region to be detected;

[0011] S2. Obtain the first or second target in the image through target detection;

[0012] S3. Obtain the first tracking box of the first target or the second tracking box of the second target through the target tracking algorithm, and determine whether the first target or the second target has crossed the preset entry line through the tracking box.

[0013] If none of them are crossed, return to step S1;

[0014] If the tracking frame of the first target crosses the preset entry line, add the ID of the first target to the storage list and proceed to step S4;

[0015] If the tracking frame of the second target crosses the preset entry line, add the second target ID to the storage list and proceed to step S4;

[0016] S4. Obtain the third tracking box of the first target or the fourth tracking box of the second target through the target tracking algorithm;

[0017] The detection tracking box determines whether it has crossed the counting line. If the third tracking box of the first target crosses the counting line, the first target is counted to obtain a temporary spoon count, and the first target ID is deleted from the storage list. If the fourth tracking box of the second target crosses the counting line, the second target is counted to obtain a temporary potato count.

[0018] If none of the tracking frames cross the counting line, repeat step S4;

[0019] S5. Obtain the first empty spoon count based on the temporary spoon count and the temporary potato count;

[0020] If the number of temporary potatoes is greater than 1, then the number of replanted potatoes = temporary potatoes - 1;

[0021] S6. Obtain the number of the second spoonful, the number of the second empty spoonful, and the number of replacement spoonfuls based on the sensor;

[0022] S7. Based on the number of empty spoons, the number of empty spoons, the number of replantings, the number of qualified clearings, and the number of replantings, calculate the missed sowing rate, the accuracy rate of empty spoon detection, the qualification rate of clearings, the qualification rate of replantings, and the replanting rate.

[0023] Furthermore, the measurement method further includes:

[0024] The missed sowing rate, seed clearing qualification rate, empty spoon detection accuracy rate, replanting qualification rate, and reseeding rate are input into the display screen for display, so that staff can adjust the replanting number during the operation.

[0025] Furthermore, obtaining the tracking bounding box of the first target or the tracking bounding box of the second target through the target tracking algorithm includes:

[0026] A target tracking network is constructed based on the YOLO and DeepSort target tracking algorithms.

[0027] The tracking boxes of the first and second targets are obtained through a target tracking network.

[0028] Furthermore, the target tracking network includes a backbone network, a feature extraction network, and a prediction network;

[0029] The backbone network is used to extract feature descriptions of the first or second target;

[0030] The feature extraction network performs feature scale transformation on the first or second target based on the feature description to obtain feature maps of different scales;

[0031] The prediction network predicts and obtains the tracking boxes of the first or second target based on feature maps at different scales.

[0032] Furthermore, the existing C3 model is replaced with the C3-Faster model in the backbone network and feature extraction network; the RepConv module is introduced into the feature extraction network.

[0033] Furthermore, a LAMP pruning score is introduced into the object detection network, and the calculation formula for the LAMP pruning score is as follows:

[0034] Calculate the pruning score of W[u]:

[0035]

[0036] In the formula:

[0037] u represents the index of weight W;

[0038] v represents the index of weight W, and u <v;

[0039] (W[u]) 2 Represents the square of the weight magnitude of the target connection;

[0040] Σ v≥u (W[v]) 2 This represents the sum of the squared weight magnitudes of indices v ≥ u within the same layer.

[0041] Second aspect

[0042] This invention discloses a metering system for precision potato planting, installed on a potato planter. The potato planter includes a planter body, a seed box and a planting mechanism on the planter body, and a seed-collecting scoop on the planting mechanism. The metering system comprises:

[0043] The acquisition mechanism, located on the top of the seeder body, is used to acquire images of the area to be inspected;

[0044] A control device is installed on the seeding mechanism and is connected to the collection mechanism;

[0045] The control device is used to calculate and statistically analyze the missed sowing rate, empty spoon rate and replanting qualification rate based on the images acquired by the acquisition mechanism, and to issue a replanting command to the replanting mechanism.

[0046] A replanting mechanism is located behind the sowing mechanism. The replanting mechanism receives a replanting command from the control device and completes the replanting.

[0047] Furthermore, the metering system also includes:

[0048] A display screen is installed on the seeder body and is connected to the control device. The display screen is used to display the detection results of the control device.

[0049] Furthermore, the replanting mechanism includes:

[0050] A sensor is used to connect to the control device to detect the number of the second empty spoon and the number of the second spoon, and to transmit the detection results to the control device.

[0051] The push rod motor is used to connect to the control device to complete the replanting action and transmit the replanting number to the control device;

[0052] The sensor includes a first sensor and a second sensor;

[0053] The first sensor is used to detect whether there is a seed-taking spoon and transmits the detection result to the control device.

[0054] The second sensor is used to detect whether there are potatoes and transmits the detection result to the control device.

[0055] When the first sensor detects the seed-taking spoon, the number of spoons is recorded once; when the second sensor detects a potato, the number of potatoes is recorded once.

[0056] The number of empty spoons in the second round = the number of spoons in the second round - the number of potatoes in the second round.

[0057] Furthermore, the control device includes a master controller and a slave controller;

[0058] The main controller is used to connect to the acquisition mechanism and detect the images acquired by the acquisition mechanism;

[0059] The slave controller is used to connect to the replanting mechanism, and there is a communication connection between the slave controller and the master controller.

[0060] The master controller and slave controller communicate using Modbus.

[0061] The advantages of this invention are:

[0062] (1) This invention achieves comprehensive detection of image data and signal data by setting up a collection mechanism and sensor on the seeder, thereby ensuring the accuracy of calculating the replanting qualification rate and the clearing qualification rate.

[0063] (2) In this invention, the C3-Faster module is selected in the backbone network to ensure effective feature extraction of the image while reducing the computational load of the target tracking network model and accelerating the deriving of effective features. The RepConv module is introduced in the feature extraction network to further reduce the computational burden and memory consumption of the target tracking network model. Finally, the pruning score algorithm is introduced in the target tracking network model to remove redundant channels and reduce the number of model parameters. Thus, by introducing the C3-Faster module, RepConv module and LAMP pruning score into the YOLO algorithm, a lightweight target tracking network model of YOLOv5n is constructed. Compared with the traditional YOLOv5n model, the parameters are reduced by 56.8%, GFLOPs are reduced by 56.1%, the model size is reduced by 51.4%, and mAP@50 reaches 98%, which can more accurately and quickly identify various parameters in the seeding process and complete real-time recording and display of data. Attached Figure Description

[0064] Figure 1 This is a diagram illustrating the method steps of the present invention;

[0065] Figure 2 This is a schematic diagram of the detection process of the present invention;

[0066] Figure 3 This is a schematic diagram of the target tracking network architecture of the present invention;

[0067] Figure 4 A schematic diagram showing the expansion of the weight W in the LAMP pruning score;

[0068] Figure 5 This is a schematic diagram of the preset entry line (inlet line) and counting line;

[0069] Figure 6 This is a schematic diagram of the system structure of the present invention. Detailed Implementation

[0070] The present invention will be further described below with reference to the accompanying drawings, but the scope of protection of the present invention is not limited to the following description.

[0071] It should be noted that the orientation or positional relationship indicated by terms such as "left" and "right" is based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship in which the product of the invention is usually placed during use, or the orientation or positional relationship in which those skilled in the art would conventionally understand it. Such terms are only for the convenience of describing the invention and simplifying the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention.

[0072] It should be noted that, unless otherwise specified, the embodiments and features and technical solutions in the present invention can be combined with each other.

[0073] First aspect

[0074] like Figures 1-5 As shown, the first aspect of the present invention discloses a metering method for precision potato planting.

[0075] S1. Obtain the image of the region to be detected;

[0076] S2. Obtain the first or second target in the image through target detection;

[0077] S3. Obtain the tracking box of the first target or the tracking box of the second target through the target tracking algorithm, and determine whether the first target or the second target has crossed the preset entry line (entry line) through the tracking box;

[0078] If none of them are crossed, return to step S1;

[0079] If the tracking frame of the first target crosses the preset entry line, add the ID of the first target to the storage list and proceed to step S4;

[0080] If the tracking frame of the second target crosses the preset entry line, add the second target ID to the storage list and proceed to step S4;

[0081] S4. Obtain the third tracking box of the first target or the fourth tracking box of the second target through the target tracking algorithm;

[0082] The detection tracking box determines whether it has crossed the counting line. If the third tracking box of the first target crosses the counting line, the count of the first target is obtained as a temporary spoon count; if the fourth tracking box of the second target crosses the counting line, the count of the second target is obtained as a temporary potato count.

[0083] If none of the tracking frames cross the counting line, repeat step S4;

[0084] S5. Based on the number of temporary spoons and the number of temporary potatoes, obtain the number of empty spoons, the number of replantings, and the number of replantings;

[0085] S6. Obtain the number of the second spoonful, the number of the second empty spoonful, and the number of replacement spoonfuls based on the sensor;

[0086] S7. Calculate the missed sowing rate, empty sowing rate, and qualified replanting rate based on the number of empty scoops in the first, second, and third rounds of empty scoops and the number of replantings.

[0087] Preferably, the measurement method further includes:

[0088] Input the missed sowing rate, qualified seed cleaning rate, empty spoon detection accuracy rate, qualified reseeding rate and re-sowing rate into the display screen for display, so that the staff can adjust the reseeding number during the operation process; the empty spoon detection accuracy rate = the second number of spoons / the first number of empty spoons.

[0089] Preferably, obtaining the tracking frame of the first target or the tracking frame of the second target through the target tracking algorithm includes:

[0090] Construct a target tracking network based on the YOLO algorithm and the DeepSort target tracking algorithm,

[0091] Obtain the tracking frames of the first target and the second target through the target tracking network.

[0092] The target tracking network model includes a backbone network, a feature extraction network and a prediction network;

[0093] The backbone network is configured to extract the feature description of the first target or the second target;

[0094] The feature extraction network performs feature scale transformation on the first target or the second target according to the feature description to obtain feature maps of different scales;

[0095] The prediction network predicts and obtains the detection frame of the first target or the second target according to the feature maps of different scales;

[0096] The tracking network layer realizes the tracking of the first target and the second target by the detection frame according to the DeepSort target tracking algorithm.

[0097] Preferably, the backbone network adopts the C3-Faster model, and the C3-Faster model is combined with the convolutional PConv to form the architecture of the backbone network; the RepConv module is introduced into the feature extraction network.

[0098] Preferably, a pruning score is introduced into the target tracking network, and the calculation formula of the pruning score is as follows:

[0099] Calculate the pruning score of W[u]:

[0100]

[0101] In the formula:

[0102] u represents the index of the W weight;

[0103] v represents the index of the W weight, and u<v, as specifically shown in Figure 3 as shown;

[0104] (W[u]) 2 represents the square of the weight magnitude of the target connection;

[0105] ∑v≥u (W[v]) 2 This represents the sum of the squared weight magnitudes of indices v ≥ u within the same layer.

[0106] like Figure 5 As shown, during operation, the acquisition mechanism acquires video of the area to be detected, thereby obtaining an image of the area to be detected; a preset entry line (inlet line) and a counting line are set on the image of the area to be detected;

[0107] The first and second targets on the image to be detected are obtained through target detection; the first target is the seed-taking spoon, and the second target is the potato.

[0108] Obtain the tracking bounding boxes of the first target and the second target through target tracking;

[0109] Detect whether the center point coordinates of the tracking frame of the first target have crossed the preset entry line (inlet line); if the center point coordinates of the tracking frame of the first target have entered the preset entry line (inlet line), then add the ID of the first target to the storage list.

[0110] Detect whether the center point coordinates of the tracking frame of the second target cross the preset entry line (inlet line); if the center point coordinates of the tracking frame of the second target enter the preset entry line (inlet line), add the second target ID to the storage list;

[0111] Check if the center coordinates of the tracking box of the first target have crossed the counting line and if the target ID is in the storage list; if the center coordinates of the tracking box of the first target have crossed the counting line and its ID is in the storage list, then increment the temporary spoon count by 1 and delete its ID from the storage list.

[0112] Check if the center coordinates of the tracking box of the second target have crossed the counting line and if the target ID is in the storage list; if the center coordinates of the tracking box of the second target have crossed the counting line and its ID is in the storage list, then increment the temporary potato count by 1 and delete its ID from the storage list.

[0113] Determine if the number of temporary spoons is 1 and the number of temporary potatoes is 0. If the number of temporary spoons is 1 and the number of temporary potatoes is 0, then increment the number of empty spoons by 1 and set both the number of temporary spoons and the number of temporary potatoes to 0.

[0114] Determine if the number of temporary spoons is 1 and the number of temporary potatoes is 1. If both the number of temporary spoons and the number of temporary potatoes are 1, then increment the number of qualified seeders by 1 and set both the number of temporary spoons and the number of temporary potatoes to 0.

[0115] Determine if the number of temporary spoons is 1 and the number of temporary potatoes is greater than 1. If the number of temporary spoons is 1 and the number of temporary potatoes is greater than 1, then increment the replanting count by 1, and set both the number of temporary spoons and the number of temporary potatoes to 0.

[0116] This invention selects the C3-Faster module in the backbone network to ensure effective feature extraction from images while reducing the computational load of the target tracking network model and accelerating the deriving of effective features. The RepConv module is introduced into the feature extraction network to further reduce the computational burden and memory consumption of the target tracking network model. Finally, a pruning score algorithm is introduced into the target tracking network model to remove redundant channels and reduce the number of model parameters. Thus, by introducing the C3-Faster module, RepConv module, and LAMP (pruning score) into the YOLO algorithm, a lightweight YOLOv5n target tracking network model is constructed. Compared to the traditional YOLOv5n model, parameters are reduced by 56.8%, GFLOPs by 56.1%, Model Size by 51.4%, and mAP@50 reaches 98%, enabling more accurate and rapid identification of various parameters during the seeding process and real-time data recording and display.

[0117] Second aspect

[0118] like Figure 6 As shown, a metering system for precision potato planting is installed on a potato planter. The potato planter includes a planter body, a seed box and a planting mechanism on the planter body, and a seed-collecting scoop on the planting mechanism; see reference. Figure 3 The metering system includes:

[0119] The acquisition mechanism, located on the top of the seeder body, is used to acquire images of the area to be inspected;

[0120] A control device is installed on the seeding mechanism and is connected to the collection mechanism;

[0121] The control device is used to calculate and statistically analyze the missed sowing rate, empty spoon rate and replanting qualification rate based on the images acquired by the acquisition mechanism, and to issue a replanting command to the replanting mechanism.

[0122] A replanting mechanism is located behind the sowing mechanism. The replanting mechanism receives a replanting command from the control device and completes the replanting.

[0123] A display screen is installed on the seeder body and is connected to the control device. The display screen is used to display the detection results of the control device.

[0124] The acquisition unit faces the area after seeding and acquires video of the area to be detected; the video is transmitted to the control device, which processes the video to acquire images of the area to be detected, and obtains the first number of empty spoons and the number of temporary spoons through target detection, target tracking and counting methods;

[0125] The replanting mechanism collects the number of potatoes and the number of empty spoons using sensors, and transmits this data to the control device. The control device calculates the number of times replanting is needed based on the number of potatoes and the number of empty spoons, and then commands the push rod motor to perform the replanting. The control device records the number of times the push rod motor operates as the number of replanting operations.

[0126] The control device calculates the number of qualified seed clearings, the qualified replanting rate, the empty seed rate, and the replanting number by combining the first empty spoon count and the temporary spoon count obtained from the image with the second empty spoon count, the replanting number, and the second spoon count obtained from the sensor and the push rod motor, according to the rules set by the expert. The calculation results are then displayed on the screen.

[0127] Preferably, the control device includes a master controller and a slave controller;

[0128] The main controller is used to connect to the acquisition mechanism and detect the images acquired by the acquisition mechanism;

[0129] The slave controller is used to connect to the replanting mechanism and to communicate with the master controller.

[0130] The master controller and slave controller communicate using Modbus.

[0131] The master controller and slave controller communicate via Modbus to transmit the second empty spoon count, the second spoon count, and the number of replacement spoons.

[0132] The replanting mechanism includes: a sensor connected to a control device for detecting the number of empty spoons and the number of spoons, and transmitting the detection results to the control device; the sensor includes a first sensor and a second sensor;

[0133] In some specific embodiments, the first sensor is positioned at a low position on the planter. When the seed-collecting spoon is lowered, it blocks the laser of the first sensor, causing the first sensor to generate a low-level signal. The controller receives a low-level signal from the first sensor, indicating that the seed-collecting spoon has been detected, and increments the count of the second spoon by one. The second sensor is positioned higher than the first sensor to detect whether there is a potato on the seed-collecting spoon. When the seed-collecting spoon rises back up, if there is no potato in it, the higher sensor will not be blocked by the potato and will generate a high-level signal. The controller receives a high-level signal from the second sensor, indicating that the seed-collecting spoon is empty at this point, and increments the count of the second empty spoon by one.

[0134] The number of replants is calculated from the level signal returned by the sensor in the controller. The number of replants = the number of empty spoons in the second round - the number of spoons in the second round. Based on the number of replants, the controller sends a replanting command to the push rod motor. Each time the push rod motor performs a replanting action, it means that the replanting was successful once. The controller records the number of replants.

[0135] The main controller obtains the number of empty spoons and spoons in the images by image processing of the first and second images; and calculates the clearing qualification rate, replanting rate, missed planting rate, empty spoon detection accuracy rate and replanting qualification rate according to the number of empty spoons, spoons and replanting number transmitted from the controller and the number of empty spoons and spoons obtained from the images through preset rules.

[0136] The above embodiments only illustrate preferred implementation methods, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of this invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this invention, and these all fall within the protection scope of this invention.

Claims

1. A metering method for precision potato planting, characterized in that, include: S1. Obtain the image of the region to be detected; S2. Obtain the first or second target in the image through target detection; S3. Obtain the first tracking box of the first target or the second tracking box of the second target through the target tracking algorithm, and determine whether the first target or the second target has crossed the preset entry line through the tracking box. If none of them are crossed, return to step S1; If the tracking frame of the first target crosses the preset entry line, add the ID of the first target to the storage list and proceed to step S4; If the tracking frame of the second target crosses the preset entry line, add the second target ID to the storage list and proceed to step S4; S4. Obtain the third tracking box of the first target or the fourth tracking box of the second target through the target tracking algorithm; The detection tracking box determines whether it has crossed the counting line. If the third tracking box of the first target crosses the counting line, the count of the first target is obtained as a temporary spoon count, and the first target ID is deleted from the storage list. If the fourth tracking frame of the second target crosses the counting line, count the second target to obtain a temporary potato count; If none of the tracking frames cross the counting line, repeat step S4; S5. Based on the number of temporary scoops and the number of temporary potatoes, obtain the number of empty scoops, the number of qualified seed clearings, and the number of replantings; S6. Obtain the number of the second spoonful, the number of the second empty spoonful, and the number of replacement spoonfuls based on the sensor; S7. Based on the number of empty spoons, the number of empty spoons, the number of replanting, the number of qualified clearing, and the number of replanting, calculate the missed sowing rate, the empty spoon detection accuracy rate, the clearing qualification rate, the replanting qualification rate, and the replanting rate. Obtaining the tracking bounding box of the first target or the tracking bounding box of the second target through a target tracking algorithm includes: A target tracking network is constructed based on the YOLO and DeepSort target tracking algorithms. The tracking boxes of the first and second targets are obtained through a target tracking network; The target tracking network includes a backbone network, a feature extraction network, and a prediction network; The backbone network is used to extract feature descriptions of the first or second target; The feature extraction network performs feature scale transformation on the first or second target based on the feature description to obtain feature maps of different scales; The prediction network predicts and obtains the detection boxes of the first or second target based on feature maps at different scales; The existing C3 model was replaced with the C3-Faster model in the backbone network and feature extraction network; the RepConv module was introduced into the feature extraction network to obtain the following results; Introduce LAMP pruning scores into the object detection network.

2. The metering method for precision potato planting according to claim 1, characterized in that: The measurement method further includes: The missed sowing rate, seed clearing qualification rate, empty spoon detection accuracy rate, replanting qualification rate, and reseeding rate are input into the display screen for display, so that staff can adjust the replanting number during the operation.

3. The metering method for precision potato planting according to claim 1, characterized in that: The formula for calculating the LAMP pruning score is as follows: calculate LAMP pruning score: In the formula: u and v both represent the indices of the weights of W, and u <v; Represents the square of the weight magnitude of the target connection; Indicating the same level, index The sum of the squares of the weighted magnitudes.

4. A metering system for precision potato planting, installed on a potato planter, the potato planter comprising a planter body, the planter body having a seed box and a planting mechanism, the planting mechanism having a seed-collecting scoop, characterized in that, For implementing the method as described in any one of claims 1 to 3, the metering system comprises: The acquisition mechanism, located on the top of the seeder body, is used to acquire images of the area to be inspected; A control device is installed on the seeding mechanism and is connected to the collection mechanism; The control device is used to calculate and statistically analyze the missed sowing rate, the accuracy rate of empty spoon detection, and the replanting qualification rate based on the images acquired by the acquisition mechanism, and to issue a replanting command to the replanting mechanism. A replanting mechanism is located behind the sowing mechanism. The replanting mechanism receives a replanting command from the control device and completes the replanting.

5. The metering system for precision potato planting according to claim 4, characterized in that: The metering system also includes: A display screen is installed on the seeder body and is connected to the control device. The display screen is used to display the calculation results of the control device.

6. The metering system for precision potato planting according to claim 5, characterized in that: The re-implantation institutions include: A sensor is used to connect to the control device to detect the number of the second empty spoon and the number of the second spoon, and to transmit the detection result to the control device; The push rod motor is used to connect to the control device to complete the replanting action and transmit the replanting number to the control device; The sensor includes a first sensor and a second sensor; The first sensor is used to detect whether there is a seed-taking spoon and transmits the detection result to the control device. The second sensor is used to detect whether there are potatoes and transmits the detection result to the control device. When the first sensor detects the seed-taking spoon, the number of spoons is recorded once; when the second sensor detects a potato, the number of potatoes is recorded once. The number of empty spoons in the second round = the number of spoons in the second round - the number of potatoes in the second round.

7. The metering system for precision potato planting according to claim 6, characterized in that: The control device includes a master controller and a slave controller; The main controller is used to connect to the acquisition mechanism and detect the images acquired by the acquisition mechanism; The slave controller is used to connect to the replanting mechanism, and the slave controller is communicatively connected to the master controller; The master controller and slave controller communicate using Modbus.

Images