A carrot pulling device, system and method based on torque self-adaption of lever mechanism

Abstract

The application discloses a carrot pulling device, system and method based on lever mechanism torque self-adaption, and belongs to the technical field of vegetable harvesting machinery and artificial intelligence, which comprises a moving device, a collecting device installed above the moving device, a pulling device installed at the front end of the moving device, and a conveying device for connecting the pulling device and the collecting device; an environment sensing device and a controller are further installed on the moving device, the input end of the controller is connected with the environment sensing device, and the output end is connected with the moving device and the pulling device. The environment sensing device and the controller are fully utilized for information fusion, the data processing capacity of the device is strengthened, the environment sensing capacity is improved, manual participation is not needed, the pulling device realizes real-time monitoring and adjustment of torque output through self-adaptive torque, the needs in different situations are met, the carrots are prevented from being damaged, the quality is ensured, process parameters are collected and analyzed, torque output is automatically adjusted, and accurate positioning and accurate pulling of the carrots are realized.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of vegetable harvesting machinery and artificial intelligence, and particularly relates to a lever mechanism torque adaptive carrot pulling device, system and method. BACKGROUND

[0002] In the existing planting area, the harvesting method of white carrots is mostly manual, which has high labor intensity and low harvesting efficiency, limiting the development of the white carrot industry. A more advanced harvesting method is to use a combined harvester, which includes a tracked chassis, a hydraulic system, a digging device, a supporting and guiding device, an oblique clamping and conveying device, a soil removing device, a cutting device, and a collecting device. The combined harvester realizes the simultaneous digging, conveying, root and leaf separation, soil removing, and collecting of two rows of carrots, improving the white carrot harvesting efficiency. However, the driver is prone to break the carrots when pulling them out, resulting in a high rate of residual fruits, and the process is not intelligent.

[0003] With the continuous development of science and technology, the rise of large-scale plantations and the increase in labor costs, the field of intelligent fruit and vegetable picking has also become popular. For white carrot harvesters, there are already some large white carrot harvesters. However, most of the current white carrot harvesters use a single sensor to perceive the external environment or their own state, which has poor machine versatility, large size, and is affected by environmental conditions. The machine is mechanically automated. For example, the Kubota white carrot harvester, although it has a fast harvesting speed, has a large machine size, high cost, and is limited by environmental conditions, requiring regular maintenance and high fees. The machine is large and prone to missed pulling or damage to the carrots. Manual operation of the machine cannot accurately adjust and control the machine, and it is not torque adaptive. Most current white carrot harvesters are mechanically automated and complete through collection, transmission, and storage. Manual operation of the machine can greatly improve efficiency and economic benefits.

[0004] There are three major problems with current white carrot harvesters: 1. High power consumption, large machine size, inaccurate torque control, prone to missed pulling or damage to the carrots, affecting quality; 2. Complex mechanism, combining digging, clamping, conveying, cutting, and collecting, requiring automatic monitoring and control of performance and state, and realizing the coordinated optimization of each working component; 3. Large machines require manual assistance, as the carrots need to be cut, there is a risk of knife injury, and these machines can only handle simple environments, such as rain and snow, which greatly affect the sensing effect of the sensors. SUMMARY

[0005] In order to overcome the above-mentioned prior art defects, the purpose of the present application is to provide a lever mechanism torque adaptive carrot pulling device, system and method to solve the technical problems of large power consumption, easy to miss pulling, easy to scratch and damage the carrot, complex device structure, the need for manual assistance, poor perception of complex environment.

[0006] In order to achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0007] The present application discloses a lever mechanism torque adaptive carrot pulling device, comprising: a motion device, a collection device installed above the motion device, a pulling device installed at the front end of the motion device, and a conveying device for connecting the pulling device and the collection device;

[0008] The motion device is also provided with an environment perception device and a controller, the input end of the controller is connected with the environment perception device, and the output end is connected with the motion device and the pulling device.

[0009] Preferably, the environment perception device comprises a camera installed at the front end of the pulling device, a soil humidity sensor installed at the front end of the conveying device, a pressure sensor installed below the collection device, and a motion sensor, an IMU unit, a Beidou satellite navigation system, a light-sensitive sensor and an ultrasonic sensor installed on one side of the motion device.

[0010] Preferably, the motion device comprises a vehicle frame, a motion motor installed on the vehicle frame, a tire connected with the motion motor through a transmission shaft, and an output shaft of the motion motor connected with the transmission shaft through a coupling;

[0011] The motion device further comprises a servo motor installed on the vehicle frame and a solar panel installed above the collection device;

[0012] The pulling device comprises an end effector, a shearing device installed on the end effector, and a transmission device connected with the shearing device; the conveying device is connected with the end effector; the end effector comprises a first telescopic support and a claw body connected with the first telescopic support; the shearing device comprises a second telescopic support and a blade connected with the second telescopic support; the transmission device comprises a lever mechanism, a rotary motor and a vibration motor connected with the lever mechanism;

[0013] The collection device comprises a pressure sensor and a collection box, and the collection device is installed at the end of the vehicle frame;

[0014] The conveying device is also provided with a cleaning machine.

[0015] The present application also discloses a lever mechanism torque adaptive carrot pulling method, which adopts the above-mentioned lever mechanism torque adaptive carrot pulling device, comprising the following steps:

[0016] S1, positioning and recognizing the radish by an environment sensing device;

[0017] S2, obtaining a lever mechanism torque initial threshold range, BDS information, IMU information, and current position information of the radish pulling device based on the lever mechanism torque self-adaption by the environment sensing device;

[0018] S3, judging the information obtained in steps S1 and S2 by a controller to obtain corresponding scene information, and inputting the scene information into a decision system to make a decision and determine whether it is in a boundary environment;

[0019] S4, outputting the decision by the decision system, obtaining an optimal path for pulling the radish based on an ant colony algorithm, operating a motion device, adjusting the operation of the radish pulling device based on the lever mechanism torque self-adaption by a torque self-adaption system to the side of the recognized radish, and pulling the radish out of the soil by a pulling device;

[0020] S5, transporting the radish to a conveying device by the pulling device;

[0021] S6, transporting the radish to a collecting device by the conveying device;

[0022] S7, when the collecting device is full of radish, transporting the radish to a designated place for storage by the radish pulling device based on the lever mechanism torque self-adaption, and resetting to a stop place to continue the work.

[0023] Preferably, in step S1, the environment light intensity is obtained by a photosensitive sensor, and the environment sensing and positioning and recognizing the radish are realized by fusing a camera and an ultrasonic sensor;

[0024] In step S2, the soil dry hardness is determined by a soil humidity sensor in combination with the land type to determine the lever mechanism torque initial threshold range, and the high-precision position information of the radish pulling device based on the lever mechanism torque self-adaption is obtained by a fusion positioning algorithm based on Kalman filtering in combination with the BDS information obtained by a Beidou satellite navigation system and the IMU information obtained by a motion sensor;

[0025] In step S4, the environment sensing is realized by fusing the camera and the ultrasonic sensor, the radish is detected and recognized and positioned, the decision system outputs a decision based on the corresponding scene information, the radish pulling device based on the lever mechanism torque self-adaption performs the work by obtaining an optimal path for pulling the radish based on the ant colony algorithm in combination with the output decision, the radish pulling device based on the lever mechanism torque self-adaption operates by the motion device, and the torque self-adaption system adjusts the entire device to run to the side of the recognized radish;

[0026] In step S5, the radish is washed by a washing machine during the conveying process;

[0027] In step S7, when the pressure sensor reaches a specified threshold range, it is determined that the collection device is full of radishes.

[0028] Preferably, step S1 specifically comprises:

[0029] S11, obtaining the light intensity Lx through the light-sensitive sensor;

[0030] S12, comparing Lx with the set light intensity threshold w;

[0031] S13, if Lx≥w, using the camera to detect and identify the target and comparing with the constructed semantic map;

[0032] S14, if Lx<w, the camera cannot effectively obtain the environmental information, light compensation is performed, and the camera and the ultrasonic sensor are used for synchronous joint calibration to realize fusion target detection. According to the size of Lx, different weights are assigned to the identification results of the two to realize decision-level fusion, so as to realize the identification of obstacles;

[0033] S15, outputting the fused result to realize target detection and identification positioning;

[0034] The specific fusion result M is represented as follows:

[0035] M=f1(Lx)x1+f2(Lx)x2 (1)

[0036] Wherein, x1 represents the target detection result of the camera; f1(Lx) represents the weight value of the camera target detection result with respect to the change of the light intensity Lx; x2 represents the target detection result of the ultrasonic wave; f2(Lx) represents the weight value of the ultrasonic target detection result with respect to the change of the light intensity Lx.

[0037] Preferably, step S2 specifically comprises: determining the initial threshold range of the torque of the lever mechanism by combining the soil humidity sensor with the soil type to determine the soil hardness; based on the lever mechanism torque self-adaptive radish pulling device, the BDS data of the Beidou satellite navigation system and the IMU data of the IMU unit are preprocessed, when it is not at the initial moment, the error of the IMU is compensated according to the feedback of Kalman update; the data is obtained and processed;

[0038] Wherein, the state X' of the Kalman filter is described as:

[0039] X'=AX+W (2)

[0040] Z=HX+v (3)

[0041] Wherein, X represents the state parameter vector of the three of BDS, IMU and odometer; A represents the corresponding state transition matrix, which is a guess model for target state conversion; W represents the corresponding system error; Z represents the measurement observation vector; H represents the conversion matrix of state parameters to observation parameters; v represents the observation error;

[0042] Then, Kalman filtering step is performed again, including prediction and update, which is specifically as follows:

[0043] Pk|k-1=A[Pk-1|k-1]A T +Qk(4)

[0044] Pk|k=[I-KkHk]Pk|k-1(5)

[0045] Wherein, Pk|k-1is the covariance prediction at k-1time; Kkis the gain at k time; Pk|kcorresponds to the covariance update at k time; Qkis the process excitation noise covariance, which represents the error between the state conversion matrix and the actual process, and Hkis the measurement covariance, which is the known condition of the filter.

[0046] Preferably, step S4 specifically comprises: in the initial case, the pheromone concentration in the field is the same, and the probability of the i-th radish being pulled at t time is set as P k i，j (t), k is a mechanical device, the pheromone concentration is T i,j (t), the distance between the two radishes is d i,j ; the lever mechanism torque adaptive radish pulling device can only pull the radish without being cut off the leaves, and the heuristic value is increased to prevent the device from being out of order and lead to the algorithm degenerating into random search, and the heuristic value is denoted as V i，j 1 / d i,j In the ant colony algorithm, it is defined as:

[0047] P k i，j (t)=[T i,j (t)] α [V i，j (t)] β / ∑ s∈Ak [T i,s (t)] α [V i，s (t)] β , if j∈A k Otherwise P k i,j (t)=0, (6)

[0048] Wherein, α represents the importance of pheromone concentration, β represents the importance of heuristic value, A kA collection of carrots is picked by a lever mechanism torque adaptive carrot picking device;

[0049] P k i,j After (t), pheromone updating process is carried out, in the gth iteration, the updating formula of pheromone is as follows:

[0050] T i,j (g+1)=(1-р)*T i,j (g)+∑ n n=1 △T i,j (t) (7)

[0051] Wherein, (1-р)*T i,j (g) represents the decay of pheromone on the path over time, p∈(0,1); and △T i,j (t) represents the increasing pheromone on the path from carrot i to carrot j at time t; △T i,j (t)=∑ m k=1 △T i,j k (t), wherein △T i,j k (t) represents the pheromone left by mechanical device k on the path of picking carrots i, j at time t, if the lever mechanism torque adaptive carrot picking device does not pass through this path, the pheromone increase caused by it is recorded as 0; otherwise, the pheromone increase caused by it is inversely proportional to the path it passes through, and the logic is described by the following formula:

[0052] If k moves from i to j at time t, △T i,j k (t)=Q / L k , otherwise, △T i,j k (t)=0, (8)

[0053] Wherein, Q represents the pheromone constant; L k represents the path length passed in iteration;

[0054] Let d k (s) be the path length selected by the lever mechanism torque adaptive carrot picking device at time s, if k moves from i to j at time t,

[0055] △T i,j k (t)=Q / ∑ n s=0 d k (s), otherwise, △T i,j k (t)=0, (9)

[0056] wherein Q / ∑ n s=0 d k (s) represents the optimal path of the lever mechanism torque adaptive carrot pulling device to complete the entire operation;

[0057] Step S4 further comprises: further adjusting the lever mechanism torque size by identifying the shape and thickness of the carrot, the transmission device drives the end effector to move and the extension support to open the claw body, the end effector approaches the ground to embrace the carrot in the end effector, the end effector is lifted up to make the leaves fold, the extension support of the shearing device shears the carrot leaves, the end effector resets and grabs the carrot, and the force of the claw body grabbing the carrot is adjusted, the vibration motor operates, the vibration loosens the carrot, and then the transmission device operates; in combination with the soil dry hardness and the shape and thickness of the carrot, the lever mechanism gives the carrot a lever mechanism torque to pull the carrot out of the soil.

[0058] Preferably, step S3 specifically comprises: dividing the environment into three levels of scene sets L1, L2 and L3 according to the information of the environment brightness, map boundary and environmental obstacles obtained in steps S1 and S2; L1 includes low brightness scenes and regular brightness scenes; L2 includes boundary scenes and non-boundary scenes; L3 includes no-obstacle scenes, simple obstacle scenes and multi-dynamic obstacle scenes;

[0059] The priority and influence range of the three levels of scene sets are: L1>L2>L3, that is, the L1 scene is judged first, and the L1 scene will affect the L2 scene set, and the L2 scene will affect the L3 scene set;

[0060] For the map boundary including the outer boundary and the internal area boundary of the non-human movable obstacle boundary, that is, the boundary when the map is constructed, the recognition process includes using the camera and the prior semantic map to judge, and using the information of the fusion positioning and the constructed map to judge the boundary;

[0061] For the division of the L3 scene, according to the target information obtained in step S2, the number of static targets is set as Ns, and the number of dynamic targets is set as Nd, Ns and Nd are compared with the corresponding threshold set, and the division of the L3 obstacle scene is realized;

[0062] Step S5 specifically comprises: the pulling device operates, the transmission device, the rotary motor rotates to drive the lever mechanism to rotate, so that the end effector transports the pulled carrots above the conveying device, the first extension support of the end effector operates to loosen the claws of the carrot, and the carrot falls into the conveying device; the conveying device is installed between the pulling device and the collecting device, the washing machine operates to spray water on the carrot during the conveying process, and the surface soil of the carrot is washed;

[0063] Step S7 specifically comprises: when the pressure sensor at the bottom of the collecting device detects that the weight of the radish in the collecting box reaches a specified threshold range, i.e., the radish has been collected, the pulling device stops pulling the radish;

[0064] The pressure sensor calculates the pressure value P of the collecting box and converts the pressure value into an electrical signal output; the electrical signal is processed by the controller to obtain an accurate pressure value, and then it is judged whether the pressure value reaches the threshold range Q; if P < Q, the pulling device continues to pull the radish; if P >= Q, the pulling device stops pulling the radish; the radish collected by the radish pulling device based on the lever mechanism torque self-adaption is transported to a specified location and stored, and then reset to the stopped location to continue working.

[0065] The application also discloses an environment sensing system of the radish pulling device based on the lever mechanism torque self-adaption, which comprises an environment sensing module, a position positioning module, a data analysis module, a data recording module and a data transmission module.

[0066] The environment sensing module is used for acquiring environment data around the radish pulling device based on the lever mechanism torque self-adaption, and comprises illumination intensity data, semantic information and obstacle information.

[0067] The position positioning module is used for acquiring high-precision positioning information of the radish pulling device based on the lever mechanism torque self-adaption, and comprises IMU information and BDS information.

[0068] The data analysis module is used for identifying the scene of a complex environment around the radish pulling device based on the lever mechanism torque self-adaption, obtaining current scene judgment information, and establishing rule-based decision output information.

[0069] The data recording module is used for recording the environment data of the environment sensing module, the positioning information of the position positioning module and the scene judgment information of the data analysis module.

[0070] The data transmission module is used for transmitting the data recorded by the data recording module to a main controller, and comprises illumination intensity data, semantic information, obstacle information, IMU information, BDS information and scene judgment information.

[0071] The noise data processing module is further included and is used for denoising the sensor data with noise caused by the environment.

[0072] Compared with the prior art, the radish pulling device based on the lever mechanism torque self-adaption has the following beneficial effects:

[0073] The application discloses a carrot pulling device based on lever mechanism torque self-adaption, which comprises a motion device for realizing the functions of driving, braking, steering and the like of the whole device, a collecting device for collecting carrots installed above the motion device, a pulling device for pulling carrots installed at the front end of the motion device, and a conveying device for connecting the pulling device and the collecting device; the conveying device conveys the carrots pulled by the pulling device to the collecting device; the motion device is further provided with an environment sensing device and a controller, the input end of the controller is connected with the environment sensing device, and the output end of the controller is connected with the motion device and the pulling device. The environment sensing device has the characteristics of user friendliness, comprises various modern technologies to achieve high intelligence, only needs user input instructions, realizes the whole process without human supervision, guarantees that no dangerous situation occurs under high efficiency, and can help realize accurate positioning and accurate pulling of carrots. The traditional carrot harvester does not have the fusion sensing function of complex environment, cannot more accurately identify and pull carrots, the carrot pulling device based on lever mechanism torque self-adaption is fully utilized to carry various advanced scientific technologies such as the environment sensing device and the controller for information fusion, the processing capacity of the device for data is strengthened, the sensing capacity of the device for environment is improved, the size of the device is reduced through the fusion sensing and mutual cooperation of multiple environment sensing devices, the cutter design is optimized, the mechanical device is monitored and controlled by using intelligent technology, manual participation is not needed, the whole device realizes real-time monitoring and adjustment of torque output through self-adaptive torque, so as to meet the needs under different conditions, process parameters are collected and analyzed, torque output is automatically adjusted, the system can realize the best performance and efficiency, and the whole device realizes real-time monitoring and adjustment of torque output through self-adaptive torque, so as to meet the needs under different conditions, process parameters are collected and analyzed, torque output is automatically adjusted, the system can realize the best performance and efficiency.

[0074] Further, the environment perception device comprises a camera installed at the front end of the pulling device, a soil humidity sensor installed at the front end of the conveying device, a pressure sensor installed below the collecting device, and a motion sensor, an IMU unit, a Beidou satellite navigation system, a light sensor and an ultrasonic sensor installed on one side of the moving device; the pressure sensor is used to convert the information of the pressure of the collecting box filled with radishes into an electrical signal, and when the pressure reaches a specified threshold range, the pulling device of the radish pulling device based on the torque self-adaption of the lever mechanism stops working; the motion sensor is used to detect the motion state of the whole device during operation, including gravity, linear acceleration, rotation vector and vibration frequency, which is used to obtain IMU data to improve the working accuracy of the whole device; the Beidou satellite navigation system BDS is used to position and navigate the running direction of the radish pulling device based on the torque self-adaption of the lever mechanism; the light sensor is used to determine whether the light supplement is needed to determine whether the camera can see clearly; the ultrasonic sensor is used to assist in detecting and identifying and positioning the radish, and identifying and avoiding the surrounding obstacles; the soil humidity sensor is used to monitor the humidity of the soil, and the hardware control circuit thereof is inserted into the soil moisture sensor at the root of the crop to monitor the moisture of the root soil, and the soil hardness is judged according to the land type to determine the initial threshold range of the torque of the lever mechanism.

[0075] Further, the radish pulling device based on the torque self-adaption of the lever mechanism can automatically position and pull the radish by using the camera and the end effector (composed of the telescopic support and the claw body), and the radish can be accurately positioned and pulled by torque self-adaption, so that the radish is not damaged and missed, and manual participation is not needed. The end effector is used to grab the radish, the cutting device is used to cut the leaves, and the transmission device is used to vibrate the soil and pull out the radish after the radish is clamped by the end effector; the radish pulling device based on the torque self-adaption of the lever mechanism also has good energy saving and emission reduction, does not pollute the environment, uses energy, and is provided with a solar panel for storing energy by itself, which is used as a power device to provide power for the driving mechanical device, so that the environment is protected and the cost is saved.

[0076] The radish pulling device based on the torque self-adaption of the lever mechanism also has good energy saving and emission reduction, does not pollute the environment, uses energy, and is provided with a solar panel for storing energy by itself, which is used as a power device to provide power for the driving mechanical device, so that the environment is protected and the cost is saved.

[0077] The radish pulling device based on the torque self-adaption of the lever mechanism also has good energy saving and emission reduction, does not pollute the environment, uses energy, and is provided with a solar panel for storing energy by itself, which is used as a power device to provide power for the driving mechanical device, so that the environment is protected and the cost is saved. BRIEF DESCRIPTION OF DRAWINGS

[0078] Figure 1The specific flow chart of pulling out the radish based on the torque self-adaption of the lever mechanism for the decision system output result of the application;

[0079] Figure 2 The connection structure diagram of the environment perception device and the controller of the radish pulling device based on the torque self-adaption of the lever mechanism of the application;

[0080] Figure 3 The complete algorithm flow chart of the ant colony algorithm of the application;

[0081] Figure 4 The end effector structure diagram of the radish pulling device based on the torque self-adaption of the lever mechanism of the application;

[0082] Figure 5 The front end structure diagram of the pulling device of the radish pulling device based on the torque self-adaption of the lever mechanism of the application;

[0083] Figure 6 The front view of the radish pulling device based on the torque self-adaption of the lever mechanism of the application;

[0084] Figure 7 The front view of the radish pulling device based on the torque self-adaption of the lever mechanism of the application;

[0085] Figure 8 The top view of the radish pulling device based on the torque self-adaption of the lever mechanism of the application.

[0086] Wherein: 1-camera; 2-solar panel; 3-servo motor; 4-motion device; 4-1-carriage; 4-2-motion motor; 4-3-tire; 5-pulling device; 5-1-end effector; 5-1-1-first telescopic support; 5-1-2-claw body; 5-2-cutting device; 5-2-1-second telescopic support; 5-2-2-knife; 5-3-transmission device; 5-3-1-lever mechanism; 5-3-2-rotary motor; 5-3-3-vibration motor; 6-collecting device; 6-1-pressure sensor; 6-2-collecting box; 7-motion sensor; 8-Beidou satellite navigation system; 9-photosensitive sensor; 10-ultrasonic sensor; 11-soil humidity sensor; 12-conveying device; 12-1-cleaning machine. DETAILED DESCRIPTION

[0087] In the following, the technical solutions in the embodiments of the present application will be described clearly and completely in combination with the accompanying drawings in the embodiments of the present application, so that those skilled in the art can better understand the technical solutions of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work should belong to the protection scope of the present application.

[0088] It should be noted that the terms "first", "second", and the like in the specification and claims of the present application and the above-described drawings are used to distinguish similar objects, and do not necessarily indicate a specific order or a chronological sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but can include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0089] The present application will be described in further detail below in combination with the accompanying drawings:

[0090] Reference is made to Figure 1A specific flowchart for the decision system output result of the application to pull out the carrots based on the torque self-adaption of the lever mechanism; from the figure, it can be seen that it specifically comprises the following steps: S1, the carrot pulling out device based on the torque self-adaption of the lever mechanism acquires the ambient light intensity through the photosensitive sensor 9, judges whether the light supplement lamp needs to supplement light to determine whether the camera 1 can see clearly, so that the camera 1 and the ultrasonic sensor 10 are fused to realize the environment perception and the positioning identification of the carrots; S2, the carrot pulling out device based on the torque self-adaption of the lever mechanism judges the soil dry hardness by the soil humidity sensor 11 combined with the land type to determine the initial threshold range of the torque of the lever mechanism 5-3-1, and acquires the current high-precision position information of itself through the fusion positioning algorithm based on Kalman filtering combined with the BDS information and the IMU information acquired by the Beidou satellite navigation system 8 and the motion sensor 7; S3, the information acquired through the steps S1 and S2 is used for scene judgment, and the corresponding scene information is input into the decision system for decision to judge whether it is in the boundary environment; S4, the decision system outputs the decision, and the carrot pulling out device based on the torque self-adaption of the lever mechanism combines the output decision to acquire an optimal path for pulling out the carrots through the ant colony algorithm to work, the carrot pulling out device based on the torque self-adaption of the lever mechanism operates through the motion device 4, the torque self-adaption system adjusts the whole device, runs to the side of the recognized carrot, further adjusts the torque size of the lever mechanism 5-3-1 through identifying the thickness of the shape of the carrot, the transmission device 5-3 drives the pulling out device 5 to move and makes the claw body 5-1-2 open through the telescopic support 5-1-1, the pulling out device 5 approaches the ground to embrace the carrot in the pulling out device 5, at this time, the carrot is not grabbed, the pulling out device 5 is slowly lifted up by a certain distance to make the leaves fold, at this time, the telescopic support 5-2-1 of the shearing device 5-2 operates to shear the leaves of the carrot, the pulling out device 5 resets and grabs the carrot, adjusts the force of the claw body 5-1-2 to grab the carrot, the vibration motor 5-3-3 operates to make the carrot loosen through vibration, then the transmission device 5-3 operates the lever mechanism 5-3-1 to give the carrot a reasonable torque of the lever mechanism which is determined by the soil dry hardness and the thickness of the shape of the carrot to pull out the carrot from the soil; S5, the pulling out device 5 runs to the conveying device 12, puts the pulled carrot, and the cleaning machine 12-1 cleans the carrot in the conveying process; S6, the cleaned carrot is transported to the collecting device 6; S7, when the pressure sensor 6-1 reaches the specified threshold range, it represents that the collecting device 6 has collected the carrots, the carrot pulling out device based on the torque self-adaption of the lever mechanism transports the carrots to the designated place to store the collected carrots, and then the carrot pulling out device based on the torque self-adaption of the lever mechanism resets to the stopped place to continue working.The pulling device 5 (the main body is derived from the lever mechanism 5-3-1) pulls out the radish through the end effector 5-1 (composed of the first telescopic support 5-1-1 and the claw body 5-1-2), cuts the radish leaves through the cutting device 5-2 (composed of the second telescopic support 5-2-1 and the blade 5-2-2), and places the radish on the conveyor belt by rotating a certain angle; the cleaning machine 12-1 is used for cleaning the soil adhered to the outer surface of the radish; the conveying device 12 is used for conveying the radish to the collection box 6-2; the collection device 6 is used for collecting the pulled radish; when the pressure borne by the collection box 6-2 reaches a specified threshold range, the pulling device 5 of the radish pulling device based on the lever mechanism torque self-adaption will stop working, and the radish is transported to a designated place for storage; then the radish pulling device based on the lever mechanism torque self-adaption is reset to work. The environment perception module is used for obtaining the environmental data around the radish pulling device based on the lever mechanism torque self-adaption, including rainfall, light intensity data, semantic information, obstacle information and the like, integrating and analyzing these data, so that the machine obtains accurate environmental information; the position positioning module obtains high-precision positioning information of the radish pulling device based on the lever mechanism torque self-adaption, including odometer information, IMU information and BDS information.

[0091] Referring to Figure 2The connection structure diagram of the environment sensing device and the controller of the lever mechanism torque adaptive carrot pulling device of the present application is shown in the figure. The input end of the controller is connected to the environment sensing device. The environment sensing device includes a camera 1 installed at the front end of the pulling device 5, a soil humidity sensor 11 installed at the front end of the conveying device 12, a pressure sensor 6-1 installed below the collecting device 6, and a motion sensor 7, an IMU unit, a Beidou satellite navigation system 8, a light-sensitive sensor 9, and an ultrasonic sensor 10 installed on one side of the moving device 4. The camera module performs fusion to realize environment sensing, detection, positioning, and identification of carrots, sense semantic information of the environment, and perform multi-target detection. The shape and thickness of the carrot are used to further fine-tune the torque of the lever mechanism 5-3-1. The ultrasonic sensor 10 is used for supplementary sensing during the time from normal sensing to no sensing of the camera 1 in low light. Target detection is performed through decision-level fusion. The light-sensitive sensor 9 is used to obtain the light intensity and determine the fusion sensing weight of the camera 1 and the ultrasonic sensor according to the light intensity. It is determined whether to supplement light to determine whether the camera 1 can see clearly. The BDS and IMU are used for boundary judgment through fusion positioning. The BDS data and IMU data are obtained and preprocessed. When not at the initial time, the IMU data is error compensated according to the feedback of Kalman update. The data is obtained and processed. The camera 1 is used to sense the land environment and identify the shape and thickness of the carrot and the position information, etc. The shape and thickness of the carrot are used to further fine-tune the torque of the lever mechanism 5-3-1. The pressure sensor 6-1 is used to convert the information of the pressure received by the collecting box 6-2 into an electrical signal. When the pressure reaches a specified threshold range, the lever mechanism torque adaptive carrot pulling device is stopped. The motion sensor 7 is used to detect the motion state of the entire device during operation, including gravity, linear acceleration, rotation vector, and vibration frequency, to obtain IMU data and improve the working accuracy of the entire device. The Beidou satellite navigation system 8 is used for positioning and navigation of the running direction of the mechanical device. The light-sensitive sensor 9 obtains the ambient light intensity to determine whether light needs to be supplemented to determine whether the camera 1 can see clearly. The ultrasonic sensor 10 is used to assist in detecting, identifying, and positioning the carrot, and identifying and avoiding surrounding obstacles. The soil humidity sensor 11 is used to monitor the soil humidity. The hardware control circuit of the soil moisture sensor is inserted into the soil moisture sensor at the root of the crop to monitor the moisture of the root soil. The soil hardness is determined according to the land type to determine the initial threshold range of the torque of the lever mechanism 5-3-1.

[0092] Referring to Figure 3The complete algorithm flowchart of the ant colony algorithm of the application is shown in the figure, wherein the lever mechanism torque self-adaptive carrot pulling device is first placed in the carrot field, the heuristic value of each path is calculated according to the path length, and the current pheromone concentration on the path is calculated; the efficiency of starting work from different carrots is calculated according to the heuristic value of each path and the pheromone concentration, the lever mechanism torque self-adaptive carrot pulling device selects the next carrot according to the efficiency, updates the pheromone concentration of each path, judges whether all carrots are pulled by the device and whether the maximum iteration number is reached, and finally outputs the optimal path.

[0093] The data analysis module is used for scene recognition of the complex environment around the lever mechanism torque self-adaptive carrot pulling device, and establishes the decision output information based on the ant colony algorithm. The optimal method for pulling carrots is obtained by recognizing and avoiding obstacles through the camera 1, and the results of the current scene judgment are obtained. The data recording module is used for recording the sensor data and the analysis result initialization information; the data transmission module is used for transmitting the analyzed data and the recorded data to the main controller, including rainfall condition, light intensity data, semantic information, obstacle information, odometer information, IMU information, BDS information and scene judgment information; on the basis of the above technical solutions, the noise data processing module is further included for denoising the sensor data with noise caused by the environment, realizing green and noise-free pollution.

[0094] Referring to Figure 4 The end effector structure schematic diagram of the lever mechanism torque self-adaptive carrot pulling device of the application is shown in the figure. Figure 5 The front end structure schematic diagram of the lever mechanism torque self-adaptive carrot pulling device of the application is shown in the figure. Figure 6 The front view of the lever mechanism torque self-adaptive carrot pulling device of the application is shown in the figure. Figure 7 The front view of the lever mechanism torque self-adaptive carrot pulling device of the application is shown in the figure. Figure 8The application discloses a carrot pulling device based on lever mechanism torque self-adaption, which comprises a motion device 4, a collecting device 6 installed above the motion device 4, a pulling device 5 installed at the front end of the motion device 4, and a conveying device 12 for connecting the pulling device 5 and the collecting device 6; the motion device 4 is further provided with an environment sensing device and a controller, the input end of the controller is connected with the environment sensing device, and the output end is connected with the motion device 4 and the pulling device 5; the environment sensing device comprises a camera 1 installed at the front end of the pulling device 5, a soil humidity sensor 11 installed at the front end of the conveying device 12, a pressure sensor 6-1 installed below the collecting device 6, a motion sensor 7, an IMU unit, a Beidou satellite navigation system 8, a light-sensitive sensor 9 and an ultrasonic sensor 10 installed at one side of the motion device 4; the motion device 4 comprises a vehicle frame 4-1, a motion motor 4-2 installed on the vehicle frame 4-1, a tire 4-3 connected with the motion motor 4-2 through a transmission shaft, and the output shaft of the motion motor 4-2 is connected with the transmission shaft through a coupling; the motion device 4 further comprises a servo motor 3 installed on the vehicle frame 4-1 and a solar panel 2 installed above the collecting device 6; the pulling device 5 comprises an end effector 5-1, a shearing device 5-2 installed on the end effector 5-1 and a transmission device 5-3 connected with the shearing device 5-2; the conveying device 12 is connected with the end effector 5-1; the end effector 5-1 comprises a first telescopic support 5-1-1 and a claw body 5-1-2 connected with the first telescopic support 5-1-1; the shearing device 5-2 comprises a second telescopic support 5-2-1 and a blade 5-2-2 connected with the second telescopic support 5-2-1; the transmission device 5-3 comprises a lever mechanism 5-3-1, a rotary motor 5-3-2 and a vibration motor 5-3-3 connected with the lever mechanism 5-3-1; the collecting device 6 comprises the pressure sensor 6-1 and a collecting box 6-2, and the collecting device 6 is installed at the end of the vehicle frame 4-1; the conveying device 12 is further provided with a cleaning machine 12-1.

[0095] The application discloses a carrot pulling device based on lever mechanism torque self-adaption, which comprises a camera 1, a solar panel 2, a servo motor 3, a motion device 4, a pulling device 5, an end effector 5-1, a collecting device 6, a pressure sensor 6-1, a motion sensor 7, a Beidou satellite navigation system 8, a light-sensitive sensor 9, an ultrasonic sensor 10, a soil humidity sensor 11 and a conveying device 12.

[0096] The camera 1 is used for perceiving the land environment and positioning and identifying the radish; the solar panel 2 is used for self-storing energy as a reserved power device; the servo motor 3 is used for controlling and adjusting the speed of the whole device; the motion device 4 is of the tire type and comprises a vehicle frame 4-1, a motion motor 4-2 and tires 4-3; the motion motor 4-2 is installed on the vehicle frame 4-1, the output shaft of the motion motor 4-2 is connected with a transmission shaft through a shaft coupling, the motion motor 4-2 can drive the transmission shaft to rotate, the tires 4-3 are installed on the transmission shaft, and the rotation of the transmission shaft can drive the tires 4-3 to move; the pulling device 5 is installed at the front end of the vehicle frame 4-1 and comprises an end effector 5-1, i.e. the pulling device 5, which is composed of a first telescopic support 5-1-1 and a claw body 5-1-2; a shearing device 5-2 is composed of a telescopic support 5-2-1 and a blade 5-2-2; a transmission device 5-3 is composed of a lever mechanism 5-3-1, a rotary motor 5-3-2 and a vibration motor 5-3-3, the shearing device 5-2 is installed on the pulling device 5, the transmission device 5-3 is installed at the front end of the vehicle frame 4-1 and connected with the pulling device 5, wherein the pulling device 5 is used for pulling the radish, the shearing device 5-2 is used for cutting the leaves, and the transmission device 5-3 is used for vibrating the soil and pulling out the radish after the pulling device 5 holds the radish; a pressure sensor 6-1 converts the information of the pressure borne by a collecting box 6-2 into an electric signal, and the pulling device 5 stops working when the pressure reaches a specified threshold range; the camera 1 identifies the shape and thickness of the radish to further adjust the torque size of the lever mechanism 5-3-1 and prevent missing pulling; a motion sensor 7 is used for detecting the motion state of the mechanical device during operation, including gravity, linear acceleration, rotation vector and vibration frequency, for obtaining IMU data and improving the working accuracy of the whole device; a soil humidity sensor 11 is used for monitoring the humidity of the soil, the hardware control circuit of the soil humidity sensor is inserted into the soil at the root of the crop to monitor the moisture of the soil at the root, the soil humidity sensor is combined with the land type to judge the soil hardness, and the torque initial threshold range of the lever mechanism 5-3-1 is determined; the collecting box 6-2 is used for collecting the pulled radish; and a conveying device 12 is used for transporting the pulled radish to the collecting box 6-2.

[0097] In order to overcome the difficulties and problems of the radish harvester in the field, the radish pulling device based on the torque self-adaption of the lever mechanism provided by the present application can realize the fusion perception and mutual cooperation of multiple sensors, reduce the size of the harvester, optimize the design of the cutter, monitor and control the mechanical device by using intelligent technology, and realize the best performance and efficiency of the system through automatic torque output adjustment.

[0098] The radish pulling method based on the torque self-adaption of the lever mechanism comprises the following steps.

[0099] S1, the carrot pulling device based on the torque self-adaption of lever mechanism acquires the ambient light intensity through the light-sensitive sensor 9, judges whether the light supplement is needed, determines whether the camera 1 can see clearly, makes the camera 1 and the ultrasonic module fuse to realize the environment perception, and detects, identifies and positions the carrots;

[0100] S2, the carrot pulling device based on the torque self-adaption of lever mechanism judges the soil dry hardness through the soil humidity sensor 11 combined with the land type, determines the initial threshold range of the torque of lever mechanism, acquires the current high-precision position information of itself through the fusion positioning algorithm based on Kalman filtering combined with the BDS information and IMU information acquired by the Beidou satellite navigation system 8 and the motion sensor 7;

[0101] S3, the information acquired through the steps S1 and S2 is used for scene judgment, and the corresponding scene information is input into the decision system for decision;

[0102] S4, the decision system outputs the decision, the carrot pulling device based on the torque self-adaption of lever mechanism performs the work through the optimal path of pulling the carrot acquired based on the ant colony algorithm, the carrot pulling device based on the torque self-adaption of lever mechanism operates through the motion device 4, the torque self-adaption system adjusts the whole device, runs to the side of the carrot which has been identified, further adjusts the torque size of the lever mechanism 5-3-1 through identifying the thickness of the carrot, the pulling device 5 operates, the transmission device 5-3 drives the end effector 5-1 to move and makes the claw body 5-1-2 open through the telescopic support, the end effector 5-1 approaches the ground and embraces the carrot in the end effector 5-1 (at this time, the carrot is not grabbed), the end effector 5-1 is slowly lifted up by a certain distance to make the leaves fold, at this time, the shearing device 5-2 operates the telescopic support to shear the carrot leaves, the end effector 5-1 resets and grabs the carrot, adjusts the force of grabbing the carrot by the claw body 5-1-2, the vibration motor 5-3-3 operates, the vibration makes the carrot loose, then the transmission device 5-3 operates the lever mechanism 5-3-1 to give the carrot a reasonable torque of lever mechanism (determined by the soil dry hardness and the thickness of the carrot) to pull the carrot out of the soil;

[0103] S5, the pulling device 5 rotates to transport the carrot to the conveyor belt, and the cleaning machine 12-1 cleans the carrot;

[0104] S6, the cleaned carrot is transported to the carrot collecting box 6-2;

[0105] S7, when the pressure sensor 6-1 in the carrot collecting box 6-2 reaches the specified threshold range, that is, the carrot collecting box 6-2 is full of carrots, the pulling device 5 of the carrot pulling device based on the torque self-adaption of lever mechanism stops pulling the carrot and transports the carrot to the designated place for storage, and then resets to the stopped place to continue the work.

[0106] As a preferred solution, step S1 specifically comprises:

[0107] S11: Obtain the data of the light-sensitive sensor 9 to obtain the illumination intensity Lx;

[0108] S12: Compare Lx with the set illumination intensity threshold w;

[0109] S13: If Lx is greater than or equal to w, target detection and recognition positioning will be performed using the camera 1 to realize the detection of multiple targets in the environment and the extraction of semantic information, and then compared with the previously constructed semantic map to make the next decision;

[0110] S14: If Lx is less than w, it indicates that the ambient light intensity is low, and the camera 1 cannot effectively obtain semantic and other environmental information. There is a transition period before the camera 1 completely fails. At this time, the light supplement lamp is used for light supplement, and the camera 1 and the ultrasonic sensor 10 are fused for target detection. The identification results of the two are assigned different weights according to the size of Lx for decision-level fusion to realize obstacle recognition;

[0111] S15: Output the fused result to realize target detection and recognition positioning;

[0112] The specific fusion result M can be represented by formula (1):

[0113] M = f1(Lx) x1 + f2(Lx) x2 (1)

[0114] Wherein, x1 represents the target detection result of the camera; f1(Lx) represents the weight value of the camera target detection result with respect to the change of the illumination intensity Lx; x2 represents the target detection result of the ultrasonic wave; f2(Lx) represents the weight value of the ultrasonic wave target detection result with respect to the change of the illumination intensity Lx.

[0115] When the camera 1 and the ultrasonic sensor 10 perform target detection, joint calibration is needed to realize the synchronization of the two.

[0116] As a preferred solution, step S2 comprises:

[0117] Determine the initial threshold range of the lever mechanism torque by judging the soil hardness of the soil humidity sensor 11 combined with the land type;

[0118] The obtained BDS data and IMU data are preprocessed, and when at a non-initial time, the IMU is error compensated according to the feedback of Kalman update. The data is obtained and processed;

[0119] Wherein, the state X' of the Kalman filter is described as:

[0120] X' = AX + W (2)

[0121] Z = HX + v (3)

[0122] Where X represents the state parameter vector of BDS, IMU and odometer; A represents the corresponding state transition matrix, which is a guess model for target state conversion; W is the corresponding system error; Z represents the measurement observation vector; H represents the conversion matrix of state parameters to observation parameters; v represents the observation error;

[0123] Then, Kalman filtering step is performed, including prediction and update; formula (4) represents one-step state prediction mean square error, and formula (5) represents state estimation covariance:

[0124] Pk|k-1 = A[Pk-1|k-1]A T +Qk (4)

[0125] Pk|k = [I-KkHk]Pk|k-1 (5)

[0126] In the above formula, formula (4) represents one-step state prediction mean square error; Pk|k-1 is the covariance prediction at k-1 time; Kk is the gain at k time; Pk|k corresponds to the covariance update at k time; Qk is the process excitation noise covariance, representing the error between the state conversion matrix and the actual process; Hk is the measurement covariance, which is the known condition of the filter.

[0127] As a preferred scheme, step S4 specifically comprises: in the initial case, the pheromone concentration in the field is the same, and the probability of the i-th radish being pulled at t time is set as P k i，j (t), k is a radish pulling device based on adaptive lever mechanism torque, the pheromone concentration is T i,j (t), the distance between the two radishes is d i,j , in addition, it is stipulated that the radish pulling device based on adaptive lever mechanism torque can only pull radishes without cut leaves. In order to prevent the device from being out of order and causing the algorithm to degenerate into random search, a concept of heuristic value is added, and the heuristic value is denoted as V i,j (t) 1 / d i,j In the ant colony algorithm, we define:

[0128] P k i，j (t) = [T i,j (t)] α [V i，j (t)] β / ∑ s∈Ak [T i,s (t)] α [Vi，s (t)] β if j∈A k

[0129] otherwise

[0130] P k i，j (t)=0 (6)

[0131] α and β are used to represent the relative importance of pheromone concentration and heuristic value, respectively. k This represents a collection of radish-pulling devices based on lever mechanism torque self-adaptation.

[0132] Determine P k i,j After (t), the pheromone update process is performed. In the g-th iteration, the pheromone update formula is as follows:

[0133] T i,j (g+1)=(1-р)*T i,j (g)+∑ n n=1 △T i,j (t) (7)

[0134] Among them, (1-р)*T i,j (g) describes the decay of pheromones along the path over time, p∈(0,1); while △T i,j (t) describes the continuous increase of pheromones from radish i to radish j at time t:

[0135] △T i,j (t)=∑ m k=1 △T i,j k (t)

[0136] △T i,j k (t) represents the pheromone left by the lever mechanism torque-adaptive radish-pulling device k on the path of pulling radish i and j at time t. If the lever mechanism torque-adaptive radish-pulling device does not traverse this path, the increase in pheromone it causes is recorded as 0; otherwise, the increase in pheromone it causes is inversely proportional to the path it traverses. We describe this logic using the following formula:

[0137] If k moves from i to j at time t, ΔT i,j k (t)=Q / L k

[0138] otherwise

[0139] △T i,jk (t) = 0 (8)

[0140] where Q is pheromone constant, and L k denotes the path length walked in the iteration;

[0141] Let d k (s) denote the path length selected by the picking mechanism at time s, then the above equation can be rewritten as:

[0142] If k moves from i to j at time t, AT i,j k (t) = Q / ∑ n s=0 d k (s), otherwise

[0143] AT i,j k (t) = 0 (9)

[0144] In the above equation, Q / ∑ n s=0 d k (s) denotes the optimal path walked by the carrot picking mechanism based on the lever mechanism torque adaptation to complete the entire operation.

[0145] As a preferred solution, step S3 specifically comprises: according to the information of step S1, step S2, the acquired environmental brightness, the map boundary, and the environmental obstacles, the environment is divided into three levels of scene sets L1, L2 and L3. L1 includes low brightness scenes and regular brightness scenes, L2 includes boundary scenes and non-boundary scenes, and L3 includes no obstacle scenes, simple obstacle scenes and multiple dynamic obstacle scenes.

[0146] As a preferred solution, the priority and influence range of the three levels of scene sets are L1>L2>L3, that is, the L1 scene is prioritized for judgment, and the L1 scene will affect the L2 scene set, and the L2 scene will affect the L3 scene set. In order to better understand, an example is used here to illustrate, assuming that L1 contains two scenes L11 and L12, L2 contains two scenes L21 and L22, and after perception, it is judged that the environment is L11 scene, then according to the decision corresponding to L11 scene, the L2 scene is perceived and judged. If it is perceived as L21 scene, then the decision of L21 is made under the decision of L11 scene.

[0147] As a preferred solution, for the map boundary, it includes the outer boundary of the map and the boundary of the internal area such as the obstacle boundary that cannot be moved artificially such as picking carrots, that is, the boundary when the map is constructed. Its recognition process includes using the camera 1 to judge with the prior semantic map and using the information of fusion positioning and the constructed map to judge the boundary.

[0148] As a preferred solution, for the differentiation of the L3 scene, according to the target information obtained in step S2, the number of static targets is set as Ns, the number of dynamic targets is set as Nd, and the comparison of Ns, Nd with the corresponding threshold set realizes the differentiation of the L3 obstacle scene.

[0149] As a preferred solution, step S5 specifically includes:

[0150] In the fusion of the camera 1 and the ultrasonic module to realize the environment perception and the identification and positioning of the radish, after the decision system outputs the decision, the lever mechanism torque adaptive radish pulling device combines the output decision to perform the work by obtaining an optimal path for pulling the radish based on the ant colony algorithm, the lever mechanism torque adaptive radish pulling device operates through the motion device 4, the torque adaptive system adjusts the whole device, and runs to the side of the radish that has been identified, and the lever mechanism torque size is further adjusted through the identification of the shape and thickness of the radish. The pulling device 5 operates, drives the end effector 5-1 to move and makes the claw body 5-1-2 open through the first telescopic support 5-1-1, the end effector 5-1 approaches the ground to embrace the radish in the end effector 5-1 (at this time, the radish is not grabbed), the end effector 5-1 is slowly lifted up by a certain distance to make the leaves fold. After the leaves are folded, the shearing device 5-2 (composed of the second telescopic support 5-2-1 and the blade 5-2-2) operates to shear the radish leaves. The end effector 5-1 resets and grabs the radish, the vibration motor 5-3-3 operates, and the vibration makes the radish loose. The transmission device 5-3 operates the lever mechanism 5-3-1 to give the radish a reasonable lever mechanism torque (determined by the soil dry hardness and the shape and thickness of the radish) to pull the radish out of the soil. The radish is pulled out, and the leaves fall to the ground.

[0151] As a preferred solution, steps S5 and S6 specifically include:

[0152] In the pulling device 5 (composed of the end effector 5-1, the shearing device 5-2 and the transmission device 5-3), the transmission device 5-3 operates, the rotary motor 5-3-2 rotates to drive the lever mechanism 5-3-1 to rotate by a certain angle, so that the end effector 5-1 (composed of the first telescopic support 5-1-1 and the claw body 5-1-2) transports the pulled radish to the above of the conveying device 12, the first telescopic support 5-1-1 of the end effector 5-1 operates to make the claw body 5-1-2 release the radish, and the radish falls into the conveying device 12. The conveying device 12 is installed between the pulling device 5 and the collecting device 6. The cleaning machine 12-1 operates to spray water to clean the radish surface soil during the conveying of the radish; the cleaned radish will be transported to the radish collecting box 6-2.

[0153] As a preferred solution, step S7 specifically includes:

[0154] The collecting device 6 is installed at the end of the vehicle frame 4-1, and comprises a collecting box 6-2 and a pressure sensor 6-1. The pressure sensor 6-1 is arranged at the bottom of the radish collecting box 6-2. When the weight of the radishes in the radish collecting box 6-2 reaches a specified threshold range, it represents that the radish collecting box 6-2 is full of radishes, and the pulling device 5 of the radish pulling device based on the torque self-adaption of the lever mechanism stops pulling radishes; the radish pulling device based on the torque self-adaption of the lever mechanism transports the radishes to a specified place for storage, and then resets the work of the radish pulling device based on the torque self-adaption of the lever mechanism.

[0155] An environment perception system of a radish pulling device based on torque self-adaption of a lever mechanism, comprising: an environment perception module, a position positioning module, a data analysis module, a data recording module and a data transmission module;

[0156] The environment perception module is used for acquiring environment data around the radish pulling device based on torque self-adaption of a lever mechanism, including rainfall condition, illumination intensity data, obstacle information and the like.

[0157] The position positioning module acquires high-precision positioning information of the radish pulling device based on torque self-adaption of a lever mechanism, including odometer information, IMU information and BDS information.

[0158] The data analysis module obtains the result of current scene judgment based on scene recognition of the complex environment around the radish pulling device based on torque self-adaption of a lever mechanism, and simultaneously establishes decision output information based on the complex environment.

[0159] The data recording module is used for recording sensor data and analysis result initialization information.

[0160] The data transmission module is used for transmitting the analyzed data and recorded data to a main controller, including rainfall condition, illumination intensity data, semantic information, obstacle information, odometer information, IMU information, BDS information and scene judgment information.

[0161] Further comprising a noise data processing module used for denoising the sensor data with noise caused by the environment.

[0162] Compared with the prior art, the radish pulling device based on torque self-adaption of a lever mechanism has the following advantages:

[0163] 1、The traditional radish harvester does not have the fusion perception function of complex environment, and cannot accurately identify and pull out the radish, the application makes full use of the camera 1, solar panel 2, servo motor 3, end effector 5-1, pressure sensor 6-1, motion sensor 7, soil moisture sensor 11 and other advanced scientific technologies carried by the radish pulling device based on the torque self-adaptive lever mechanism to carry out information fusion, strengthen the data processing ability, improve the environmental perception ability, and the whole device can realize real-time monitoring and adjustment of torque output through adaptive torque, to meet the needs of different situations, and collect and analyze process parameters, through automatic torque output adjustment, the system can realize the best performance and efficiency.

[0164] 2、The application uses the camera 1 and the end effector 5-1 (composed of the first telescopic support 5-1-1 and the claw body 5-1-2) to position and pull out the radish, and through torque self-adaption, the radish can be accurately positioned and pulled out, ensuring that the radish is not damaged and missed, and having strong adaptability without manual participation.

[0165] 3、The application has good energy saving and emission reduction, no pollution, uses energy, and is provided with the solar panel 2 for storing energy, as a reserve power device, to provide power for the driving mechanical device, which can protect the environment and save costs.

[0166] 4、The application has the user-friendly feature, uses the camera 1, solar panel 2, servo motor 3, motion device 4, pulling device 5, conveying device 12, collecting device 6, pressure sensor 6-1, motion sensor 7, Beidou satellite navigation system 8 (BDS), photosensitive sensor 9, ultrasonic sensor 10, soil moisture sensor 11 and other modern technologies to achieve high intelligence, only needs the user to input the instruction, the whole process realizes unmanned supervision, ensures that there is no dangerous situation in the case of high efficiency.

[0167] The above content only illustrates the technical idea of the application, and cannot limit the protection scope of the application, any modification made according to the technical idea of the application on the basis of the technical scheme falls within the protection scope of the claims of the application.

Claims

1. A radish-pulling device based on lever mechanism torque self-adaptation, characterized in that, include: The motion device (4), the collection device (6) installed above the motion device (4), the extraction device (5) installed at the front end of the motion device (4), and the transmission device (12) for connecting the extraction device (5) and the collection device (6). The motion device (4) is also equipped with an environmental sensing device and a controller. The input end of the controller is connected to the environmental sensing device, and the output end is connected to the motion device (4) and the pulling device (5). The environmental sensing device includes a camera (1) installed at the front end of the extraction device (5), a soil moisture sensor (11) installed at the front end of the transmission device (12), a pressure sensor (6-1) installed below the collection device (6), and a motion sensor (7), an IMU unit, a Beidou satellite navigation system (8), a photosensitive sensor (9), and an ultrasonic sensor (10) installed on one side of the motion device (4). The motion device (4) includes a frame (4-1), a motion motor (4-2) mounted on the frame (4-1), a tire (4-3) connected to the motion motor (4-2) via a drive shaft, and the output shaft of the motion motor (4-2) connected to the drive shaft via a coupling. The motion device (4) also includes a servo motor (3) mounted on the frame (4-1) and a solar panel (2) mounted above the collection device (6). The extraction device (5) includes an end effector (5-1), a shearing device (5-2) mounted on the end effector (5-1), and a transmission device (5-3) connected to the shearing device (5-2); the conveying device (12) is connected to the end effector (5-1); the end effector (5-1) includes a first telescopic bracket (5-1-1) and a claw body (5-1-2) connected to the first telescopic bracket (5-1-1); the shearing device (5-2) includes a second telescopic bracket (5-2-1) and a blade (5-2-2) connected to the second telescopic bracket (5-2-1); the transmission device (5-3) includes a lever mechanism (5-3-1), a rotary motor (5-3-2) and a vibration motor (5-3-3) connected to the lever mechanism (5-3-1); The collection device (6) includes a pressure sensor (6-1) and a collection box (6-2), and the collection device (6) is installed at the end of the frame (4-1); The conveying device (12) is also equipped with a cleaning machine (12-1).

2. A method for pulling radishes based on the adaptive torque of a lever mechanism, characterized in that, The radish-pulling device based on lever mechanism torque adaptation as described in claim 1 includes the following steps: S1. Locate and identify the radish using an environmental sensing device; S2. Obtain the initial threshold range of the lever mechanism torque, BDS information, IMU information, and the current position information of the radish-pulling device based on lever mechanism torque adaptation through the environmental sensing device; S3. The controller performs scene judgment on the information obtained in steps S1 and S2 to obtain the corresponding scene information, and inputs it into the decision system for decision-making to determine whether it is in a boundary environment. S4. The decision system outputs a decision and then obtains the optimal path for pulling the radish based on the ant colony algorithm. The motion device (4) operates, and the torque adaptive system adjusts the radish pulling device based on the torque adaptive lever mechanism to run to the identified radish. The radish is then pulled out of the soil by the pulling device (5). S5. The picking device (5) moves the radish onto the conveying device (12); S6. The conveying device (12) transports the radish to the collecting device (6). S7. When the collecting device (6) is full of radishes, the radish-pulling device based on the torque adaptation of the lever mechanism transports the radishes to a designated location for storage, and then returns to the stop location to continue working.

3. The radish-pulling method based on lever mechanism torque adaptation according to claim 2, characterized in that, In step S1, the ambient light intensity is obtained by the photosensitive sensor (9), and the camera (1) and the ultrasonic sensor (10) are fused to realize environmental perception and location identification of the radish; In step S2, the soil dryness hardness is determined by the soil moisture sensor (11) in combination with the land type to determine the initial threshold range of the lever mechanism torque. Then, the BDS information obtained by the Beidou satellite navigation system (8) and the IMU information obtained by the motion sensor (7) are combined to obtain the high-precision position information of the radish pulling device based on lever mechanism torque adaptation through the fusion positioning algorithm based on Kalman filtering. In step S4, the camera (1) and the ultrasonic sensor (10) are fused to realize environmental perception, and the radish is detected, identified and located. The decision system outputs a decision based on the corresponding scene information input. The radish-pulling device based on the lever mechanism torque adaptive combines the output decision and obtains an optimal path for pulling the radish based on the ant colony algorithm. The radish-pulling device based on the lever mechanism torque adaptive operates through the motion device (4). The torque adaptive system adjusts the entire device and runs to the side of the identified radish. In step S5, the radish is cleaned by the cleaning machine (12-1) during the conveying process; In step S7, when the pressure sensor (6-1) reaches the specified threshold range, it means that the collecting device (6) has been filled with radishes.

4. The radish-pulling method based on lever mechanism torque adaptation according to claim 2, characterized in that, Step S1 specifically includes: S11. Obtain the light intensity Lx through the photosensitive sensor (9); S12. Compare Lx with the set light intensity threshold w; S13. If Lx≥w, use camera (1) to perform target detection and recognition positioning, and compare it with the constructed semantic map; S14. If Lx < w, the camera (1) cannot effectively acquire environmental information and supplement light. The camera (1) and the ultrasonic sensor (10) are simultaneously calibrated to perform target detection. According to the size of Lx, different weights are assigned to the recognition results of the two to perform decision-level fusion in order to realize the recognition of obstacles. S15. Output the fused result to achieve target detection, recognition and localization; The specific fusion result M is represented as follows: M = f1(Lx)x1 + f2(Lx)x2 (1) Where x1 represents the target detection result of the camera; f1(Lx) represents the weight value of the target detection result of the camera with respect to the change in light intensity Lx; x2 represents the target detection result of the ultrasound; and f2(Lx) represents the weight value of the target detection result of the ultrasound with respect to the change in light intensity Lx.

5. The radish-pulling method based on lever mechanism torque adaptation according to claim 2, characterized in that, Step S2 specifically includes: determining the soil dryness hardness by combining the soil moisture sensor (11) with the land type to determine the initial threshold range of the lever mechanism torque; preprocessing the BDS data of the acquired Beidou satellite navigation system (8) and the IMU data of the IMU unit based on the lever mechanism torque adaptive radish pulling device; when it is not at the initial time, performing error compensation on the IMU according to the feedback of Kalman update; obtaining data and processing it. The state X′ of the Kalman filter is described as follows: X′=AX+W(2) Z = HX + v (3) Where X represents the state parameter vector of the BDS, IMU and odometer; A represents the corresponding state transition matrix, which is a conjectured model of the target state transition; W represents the corresponding system error; Z represents the measurement observation vector; H represents the transformation matrix from state parameters to observation parameters; and v represents the observation error. Then, a Kalman filtering step is performed, which includes two parts: prediction and update, as detailed below: Pk|k-1=A[Pk-1|k-1]A T +Qk(4) Pk|k=[I-KkHk]Pk|k-1(5) Where Pk|k-1 is the covariance prediction at time k-1; Kk is the gain at time k; Pk|k corresponds to the covariance update at time k; Qk is the process excitation noise covariance, representing the error between the state transition matrix and the actual process; and Hk is the measurement covariance, which is a known condition of the filter.

6. The radish-pulling method based on lever mechanism torque adaptation according to claim 2, characterized in that, Step S4 specifically includes: In the initial case, the pheromone concentration in the field is the same, and the probability of the i-th radish being pulled at time t is set as P. k i，j (t), k is the mechanical device, and the pheromone concentration is T. i,j (t), the distance between the two radishes is d. i,j The lever-mechanism torque-adaptive radish-pulling device is designed to only pull radishes whose leaves have not been cut off. A heuristic value is added to prevent the device from moving randomly and degenerating into a random search. This heuristic value is denoted as V. i，j (t)1 / d i,j In the ant colony algorithm, it is defined as: P k i，j (t)=[T i,j (t)] α [V i，j (t)] β / ∑ s∈Ak [T i,s (t)] α [V i，s (t)] β ,if j∈A k Otherwise P k i,j (t)=0, (6) Where α represents the importance of pheromone concentration, β represents the importance of heuristic value, and A k This represents a collection of radish-pulling devices that utilize lever mechanism torque adaptation to pull radishes. Determine P k i,j After (t), the pheromone update process is performed. In the g-th iteration, the pheromone update formula is as follows: T i,j (g+1)=(1-р)*T i,j (g)+∑ n n=1 △T i,j (t)(7) Among them, (1-р)*T i,j (g) represents the decay of pheromones along the path over time, p∈(0,1); while △T i,j (t) indicates that at time t, the pheromone level from radish i to radish j continuously increases; △T i,j (t)=∑ m k=1 △T i,j k (t), where △T i,j k (t) represents the pheromone left by the mechanical device k on the path of pulling up radish i and j at time t. If the radish-pulling device based on the torque adaptation of the lever mechanism does not pass through this path, the increase in pheromone it causes is recorded as 0; otherwise, the increase in pheromone it causes is inversely proportional to the path it has traversed. This logic is described by the following formula: If k moves from i to j at time t, ΔT i,j k (t)=Q / L k Otherwise, △T i,j k (t)=0, (8) Where Q represents the pheromone constant; L k This represents the length of the path traversed during the iteration; d k (s) is denoted as the path length selected by the lever mechanism torque adaptive radish-pulling device at time s. If k moves from i to j at time t... △T i,j k (t)=Q / ∑ n s=0 d k (s), otherwise, △T i,j k (t)=0, (9) Where, Q / ∑ n s=0 d k (s) represents the optimal path taken by the lever mechanism torque-adaptive radish-pulling device to complete the entire operation; Step S4 further includes: by identifying the shape and thickness of the radish, further fine-tuning the torque of the lever mechanism; the transmission device (5-3) drives the end effector (5-1) to move and the claw (5-1-2) opens through the telescopic bracket (5-1-1); the end effector (5-1) closes to the ground and hugs the radish inside the end effector (5-1); the end effector (5-1) lifts up to make the leaves close; at this time, the telescopic bracket (5-2-1) of the shearing device (5-2) cuts the radish leaves; the end effector (5-1) resets and grabs the radish, and adjusts the force of the claw (5-1-2) to grab the radish; the vibration motor (5-3-3) operates, and the vibration loosens the radish; then the transmission device (5-3) operates; combined with the dryness and hardness of the soil and the shape and thickness of the radish, the lever mechanism (5-3-1) gives the radish a lever mechanism torque to pull the radish out of the soil.

7. The radish-pulling method based on lever mechanism torque adaptation according to claim 2, characterized in that, Step S3 specifically includes: based on the information of ambient brightness, map boundaries, and environmental obstacles obtained in steps S1 and S2, dividing the environment into three levels of scene sets L1, L2, and L3; L1 includes low-brightness scenes and normal-brightness scenes; L2 includes boundary scenes and non-boundary scenes; L3 includes obstacle-free scenes, simple obstacle scenes, and multi-dynamic obstacle scenes. The priority and scope of influence of the three levels of scene sets are as follows: L1 > L2 > L3, that is, L1 scene is judged first, and L1 scene will affect L2 scene set, and L2 scene will affect L3 scene set; For the map boundary, which includes the outer boundary of the map and the inner boundary of the region, and is an obstacle boundary that cannot be moved by humans, i.e. the boundary when the map is constructed, the recognition process includes using a camera (1) to make a judgment with the prior semantic map, and using the information of fused positioning and the constructed map to make a boundary judgment. To distinguish L3 scenarios, based on the target information obtained in step S2, the number of static targets is set to Ns and the number of dynamic targets is set to Nd. Ns and Nd are compared with the corresponding set thresholds to achieve the distinction of L3 obstacle scenarios. Step S5 specifically includes: the pulling device (5) operates, the transmission device (5-3) and the rotary motor (5-3-2) rotate to drive the lever mechanism (5-3-1) to rotate, so that the end effector (5-1) transports the pulled radish to the top of the conveying device (12), the first telescopic bracket (5-1-1) of the end effector (5-1) operates to make the claw (5-1-2) release the radish, and the radish falls into the conveying device (12); the conveying device (12) is installed between the pulling device (5) and the collecting device (6), and the washing machine (12-1) operates during the radish conveying process to spray water on the radish and clean the mud on the surface of the radish; Step S7 specifically includes: when the pressure sensor (6-1) at the bottom of the collecting device (6) detects that the weight of the radish in the collecting box (6-2) has reached the specified threshold range, that is, the radish has been collected, the pulling device (5) stops pulling the radish; The pressure sensor (6-1) calculates the pressure value P on the collection box (6-2) and converts the pressure value into an electrical signal output. The controller processes the electrical signal to obtain an accurate pressure value and then judges whether it reaches the threshold range Q. If P < Q, the pulling device (5) continues to pull the radish. If P ≥ Q, the pulling device (5) stops pulling the radish. The radish pulling device based on the torque adaptation of the lever mechanism transports the collected radish to the designated location and stores it, and then resets it to the stopping location to continue working.

8. An environmental sensing system for a radish-pulling device based on lever mechanism torque adaptation, characterized in that, It includes an environmental perception module, a location positioning module, a data analysis module, a data recording module, a data transmission module, and a noise data processing module; The environmental sensing module is used to acquire environmental data around the radish-pulling device based on the torque adaptation of the lever mechanism; This includes light intensity data, semantic information, and obstacle information; The position positioning module is used to acquire high-precision positioning information of the radish-pulling device based on lever mechanism torque adaptation, including IMU information and BDS information; The data analysis module is used to identify the complex environment around the radish-pulling device based on the torque adaptation of the lever mechanism, obtain the current scene judgment information, and establish rule-based decision output information. The data recording module is used to record environmental data from the environmental perception module, positioning information from the location positioning module, and scene judgment information from the data analysis module. The data transmission module is used to transmit the data recorded by the data recording module to the main controller, including light intensity data, semantic information, obstacle information, IMU information, BDS information and scene judgment information; The noise data processing module is used to denoise sensor data that is noisy due to environmental factors.

Images