CN-116125483-B - Precision verification mechanism and method for tea canopy distance detection algorithm

CN116125483BCN 116125483 BCN116125483 BCN 116125483BCN-116125483-B

Abstract

The invention discloses an accuracy verification mechanism and method of a tea surface distance detection algorithm, which are characterized in that distance values of end points on two sides of a cutter and a simulated tea surface are obtained through a two-dimensional laser radar, then the distance values of the end points of the cutter and the simulated tea surface, which are measured by the two-dimensional laser radar, are used as distance initial values, the distance values of the end points of the cutter and the simulated tea surface are optimally estimated by adopting the tea surface distance detection algorithm fused with a plurality of sensors, finally the optimal estimated distance between the end points of the cutter and the simulated tea surface and the true value distance error directly measured by the sensor with high accuracy are calculated, the accuracy of the tea surface distance detection algorithm is evaluated from the angles of an error mean value and an error variance, and the parameter of the noise variance is measured by finely adjusting the acceleration in the vertical direction in the attitude sensor in the tea surface distance detection algorithm, so that the tea surface distance detection algorithm meeting the accuracy requirement is finally obtained.

Inventors

WU MIN
ZHAO RUNMAO
CHEN JIANNENG
HUAN XIAOLONG
JIA JIANGMING
Shao Baikai

Assignees

浙江理工大学

Dates

Publication Date: 20260505
Application Date: 20221229

Claims (2)

1. The accuracy verification mechanism of the tea awning surface distance detection algorithm is characterized by comprising a cutting knife and an analog tea awning surface underframe positioned below the cutting knife, wherein connecting rods are hinged to two sides of the cutting knife, the connecting rods at two sides are respectively hinged to two upper sliding blocks, the two upper sliding blocks and two screw rods respectively form screw pairs, the two screw rods and two sides of a frame respectively form sliding pairs along the vertical direction, the two screw rods are respectively driven by two motors, a pull rope sensor is fixed on the side face of the cutting knife, a pulley frame is connected with a pull rope of the pull rope sensor, pulleys are hinged to the pulley frame, a groove pin pair is formed by a simulated tea awning surface sliding groove formed by the analog tea awning surface underframe and the pulleys through pin shafts, connecting rods are hinged to two ends of the simulated tea awning surface underframe, the connecting rods at two ends of the simulated tea awning surface sliding groove are respectively hinged to two lower sliding blocks, the two lower sliding blocks and two sides of the frame respectively form sliding pairs along the vertical direction, the two sliding pairs are respectively fixed with the two sides of the frame through bolts, and the two-dimensional laser radar is arranged on the cutting knife and the simulated tea awning surface underframe.
2. The method for verifying the accuracy of the tea surface distance detection algorithm by the accuracy verification mechanism of the tea surface distance detection algorithm according to claim 1 is characterized by comprising the following specific steps: The method comprises the steps of firstly, placing a simulated tea awning surface on a simulated tea awning surface underframe, adjusting the horizontal inclination angle of the simulated tea awning surface underframe to a set value, enabling the relative positions of a cutting knife and the simulated tea awning surface to change in real time by asynchronous and non-periodic rotation of two motors, and obtaining the acceleration of the cutting knife in the vertical direction by a controller according to the detection signal of an attitude sensor on the cutting knife Horizontal inclination angle of cutter Real-time measurement values are obtained, and real-time measurement values of the distance between the cutting knife and the simulated tea awning surface underframe are obtained according to detection signals of the stay cord sensor; Scanning the simulated tea canopy surface by a two-dimensional laser radar to obtain the point cloud coordinates of the ROI area near the point C and the point D on the simulated tea canopy surface, and calculating the distance measurement value of the side end point A, B of the cutter and the simulated tea canopy surface in the vertical direction; Step three, optimally estimating the distance between the two sides of the cutter and the simulated tea canopy surface in the vertical direction by using a tea canopy surface distance detection algorithm with the accuracy to be verified; Setting the distance between a cutting knife obtained according to the detection signal of the stay cord sensor and the simulated tea awning surface chassis as l MN , and setting the distance between the top of the simulated tea awning surface and the simulated tea awning surface chassis as The distance between a connection point M of the stay rope sensor and the cutter and a projection point F of the point M on the simulated tea canopy surface along the vertical direction is obtained by the geometric relationship: Wherein beta is the horizontal inclination angle of the simulated tea awning surface underframe obtained according to the detection signal of the gesture sensor on the simulated tea awning surface underframe; if the stay cord sensor is arranged at the end point A of the cutter, the stay cord sensor is calculated If the stay rope sensor is arranged at the end point B of the cutter, the distance l AC between the end point A and the point C is calculated Distance l BD between end point B and point D; Step five, the step four is carried out Storing the real-time calculated value and the real-time optimal estimated value obtained in the third step, and performing precision evaluation on a tea canopy distance detection algorithm with precision to be verified after the set test time, wherein the specific steps of the precision evaluation are as follows: ① Extraction of Group of Value and the optimal estimated value data obtained in the third step, each group Personal (S) The optimal estimated value data obtained in the third step is selected according to the position of the end point A or the end point B of the cutter, which is provided with the stay cord sensor; ② For each set of test data, each The value is recorded as The optimal estimated values of the corresponding sides of the cutting knife are recorded as Calculating each of the test data Error mean value of optimal estimated value of value corresponding to each cutting knife Error variance ; ③ According to the first The mean value of the errors obtained from the set of test data is noted as According to Mean of error mean obtained from group test data Mean of error variance ; Step six, by And As the accuracy evaluation standard of the tea canopy distance detection algorithm of the accuracy to be verified, if And One of them does not meet the threshold requirement, fine tuning Repeating steps 3.4, fourth and fifth until And After all meet the threshold requirement, the parameters are The lower tea canopy distance detection algorithm is regarded as an optional algorithm; the second specific steps are as follows: 2.1, establishing a rectangular coordinate system xoy by taking a two-dimensional laser radar center point as an origin, wherein a y-axis coincides with a polar angle 0-degree direction of the two-dimensional laser radar polar coordinate system and points downwards, a cutter is perpendicular to the y-axis, and an x-axis is parallel to the cutter and rotates 90 degrees anticlockwise on the y-axis The horizontal axis in the direction that the axis rotates 90 degrees anticlockwise is Is vertically downward Establishing a static coordinate system by an axis, and establishing a coordinate system xoy and a coordinate system Horizontal inclination angle of phase difference cutter Setting the width AB=l of the cutter, the projections of the end points A, B on the two sides of the cutter on the simulated tea canopy surface along the vertical axis are respectively a point C and a point D, the y axis intersects with the cutter at a point E, oE =h is set, the two-dimensional laser radar scans the simulated tea canopy surface to obtain the point cloud of the ROI area near the point C and the point D on the simulated tea canopy surface, and the polar coordinates of the points in the point cloud of the ROI area near the point C and the point D under the polar coordinate system Conversion to a coordinate system In (a), namely: = * (1) 2.2 in coordinate System xoy, point A coordinates are The coordinates of the point B are ; In a coordinate system The point a coordinates are: = * (2) At the position of In the coordinate system, the coordinates of the point B are as follows: = * (3) 2.3 at In the coordinate system, point A and point C The axis coordinates being identical, i.e Recording the points in the ROI area point cloud near the point C In a coordinate system Axis coordinates and point C The absolute value of the difference of the axis coordinates is: (4) Traversing the ROI area point cloud data around point C will Taking the point corresponding to the smallest time as a point C, and similarly taking the point in the ROI area point cloud near the point D Axis coordinates and point D The point corresponding to the point with the minimum absolute value of the difference of the axis coordinates is taken as a point D, wherein the point D The axis coordinates being equal to point B Axis coordinates, i.e. ; 2.4 Taking the polar angle range around the midpoint C of the polar coordinate System as Is converted into a coordinate system Midpoint coordinates , , ,..., Obtaining the point cloud coordinates of the ROI area in the neighborhood of the point C, and then obtaining the points of the point cloud of the ROI area in the neighborhood of the point C The axis coordinate mean value is: (5) n is the number of points of the ROI area point cloud in the neighborhood of the point C; Similarly, the polar angle range near the midpoint D of the polar coordinate system is taken as Is converted into a coordinate system The midpoint coordinates obtain the point cloud coordinates of the ROI area in the neighborhood of the point D, and further obtain the points of the point cloud of the ROI area in the neighborhood of the point D Mean value of axis coordinates ; 2.5 From formulas (2) and (3), points A and B are obtained in the coordinate system In (a) and (b) The axis coordinates are respectively: (6) Taking the calculation result of the formula (7) as the measured value of the distance between the end point A and the point C and the distance between the end point B and the point D of the cutter: (7) The third specific steps are as follows: 3.1 construction of the kinematic equation as follows: (8) Wherein, the Sampling time intervals for the attitude sensor; After iteration of The distance between the endpoint A or the endpoint B of the cutter and the simulated tea canopy surface in the vertical direction at the moment; Is that Is a derivative of (2); Is that The acceleration of the cutter in the vertical direction is calculated according to the detection signal of the attitude sensor at the moment, Initial value of (2) Taking the distance between the end point A or the end point B of the cutter calculated by the method (7) at the moment 0 and the simulated tea canopy surface in the vertical direction, taking Initial value of (2) =0, Taking 0 as an initial value of (2); 3.2, establishing a state space model of a kinematic equation, and adopting a discrete state space to represent as follows: (9) wherein the state space model is at the first The system state variable at each sampling instant is State space model at The system state variable for +1 sampling instants is State space model at the first The process noise matrix iterated at each sampling moment is as follows Wherein the distance between the endpoint A or the endpoint B of the cutter and the simulated tea canopy surface in the vertical direction is iterative process noise in the state space model Iterative process noise in state space model of distance change rate between endpoint A or endpoint B of cutter and simulated tea canopy surface in vertical direction , The method is characterized in that acceleration measurement noise in the vertical direction in a detection signal of the attitude sensor is randomly valued in the accuracy range of the acceleration detection of the attitude sensor, and a normal distribution rule is met; In the first place for the state space model The output state variables for the individual sampling instants, ; In the first place for the state space model The control amount of the time of the sampling, = , Is that The moment is calculated according to the detection signal of the attitude sensor to obtain the true value of the cutter after the acceleration in the vertical direction is removed from the measurement noise, and the system matrix Input matrix Output matrix ; In the formula (9) of the present invention, = The formula (9) is simplified as follows: 3.3 to Is shown in the first Measuring the distance between the end point A or the end point B of the cutter and the simulated tea canopy surface in the vertical direction at the sampling time (10) Wherein, the Is a measurement matrix; in the second dimension for the two-dimensional laser radar Measurement noise at each sampling instant, in mm, Randomly taking values within the range of the ranging accuracy of the two-dimensional laser radar, wherein the values meet normal distribution; 3.4, fusing two-dimensional laser radar measurement data with acceleration measurement data of an attitude sensor on a cutter in the vertical direction through a tea awning distance detection algorithm with accuracy to be verified, wherein the specific process is as follows: ① Set the first The prior estimation matrix of the system state variable at each sampling moment is as follows: (11) Wherein, the , Is the first At each sampling time Is used to determine the value of the a priori estimate of (c), Is the first At each sampling time Is used to determine the value of the a priori estimate of (c), Is the first A priori estimates of the system state variables at the sampling instants, , Is the first At each sampling time Is used to determine the posterior estimate of (1), Is the first At each sampling time Is determined, when k=1, ; ② Calculate the first Priori estimation matrix of each sampling moment Covariance matrix of two variables in (a): (12) Wherein, the Is the first Covariance matrix of two variables in posterior estimation matrix at each sampling time, and when k=1, covariance matrix of two variables in posterior estimation matrix ; Is a system matrix Transpose of (a) process noise covariance matrix , The prior estimation process noise variance of the distance between the endpoint A or the endpoint B of the cutter and the simulated tea canopy surface in the vertical direction, The prior estimation of the process noise variance of the distance change rate of the endpoint A or the endpoint B of the cutter and the simulated tea canopy surface in the vertical direction, , Parameters (parameters) The initial value of the (1) is taken as the acceleration measurement noise variance in the vertical direction in the detection signal when the attitude sensor is static; ③ Calculate the first Gain at each sampling instant: (13) Wherein, the Is a measurement matrix Is a transpose of (2); the variance of the measurement noise of the two-dimensional laser radar is equal to the variance of the normal distribution which is consistent with the measurement noise of the two-dimensional laser radar; ④ Calculate the first Posterior estimation matrix of system state variables at each sampling instant: (14) Wherein, the ; ⑤ Updating covariance matrixes of two variables in a posterior estimation matrix at a kth sampling moment: (15) Wherein the method comprises the steps of Is a unit vector.

Description

Precision verification mechanism and method for tea canopy distance detection algorithm Technical Field The invention belongs to the field of digital signal processing of profile modeling harvesting of bulk tea, and particularly relates to an accuracy verification mechanism and method of a tea canopy distance detection algorithm. Background Tea is one of the well-known healthy plant beverages in the 21 st century, the development of the Chinese tea industry has promoted the increase of the world tea yield, the 2021 nationwide tea yield is 318 ten thousand tons, and the annual increase is nearly 1 time than the 2020 increase of 24.82 ten thousand tons. The domestic tea market sales amount breaks through 3000 hundred million yuan, and the tea consumer group is nearly 5 hundred million people. The fresh leaf harvesting is used as a key link of large-scale tea production, under the background of shortage of labor force, the mechanized tea picking machine is widely applied to standard tea gardens, however, the domestic tea gardens are mostly in hilly areas and have different management standards, the furrow of the tea garden is uneven, the height of the tea canopy varies in time and space, and the commercial mechanized tea picking machine cannot adapt to the influence of the height variation of the tea canopy and the fluctuation of the furrow, so that the mechanized tea picking machine is difficult to popularize and apply in China. For this reason, some automatic profiling tea-leaf picker are studied, the sensor senses the height change of the tea-leaf surface and uneven ridges and furrows of the tea garden, for example, the patent of the invention with publication number CN113039936A proposes to measure the distance between the cutter and the tea-leaf surface by ultrasonic, however, the ultrasonic sensor is used as a point-to-point distance measuring sensor, is easily influenced by the clearance of blades of the tea-leaf surface, and causes the distance measuring stability to be too low. The outdoor laser radar is adopted to measure the distance of the tea surface, the distances of a plurality of sample points of the tea surface are obtained, the distance measurement accuracy of the outdoor laser radar is insufficient due to the characteristic influence of the distance measurement principle of the outdoor laser radar, and the laser radar distance measurement accuracy is improved by adopting a specific algorithm. In an actual tea garden environment, the effectiveness of the improvement of the accuracy of the detection algorithm of the tea canopy distance is difficult to verify by using a higher accuracy sensor. Therefore, there is a need for an accuracy verification mechanism and related evaluation method for a tea canopy distance detection algorithm. Disclosure of Invention The invention provides a precision verification mechanism and method for a tea canopy distance detection algorithm, aiming at the problem that the precision of the tea canopy distance detection algorithm cannot be verified in the automatic profiling harvesting process of fresh tea leaves. The invention discloses an accuracy verification mechanism of a tea awning surface distance detection algorithm, which comprises a cutting knife and a simulated tea awning surface underframe positioned below the cutting knife, wherein connecting rods are hinged to two sides of the cutting knife, the connecting rods at two sides are respectively hinged to two upper sliding blocks, the two upper sliding blocks and two screw rods respectively form screw pairs, sliding pairs along the vertical direction are respectively formed on two sides of a frame, the two screw rods are respectively driven by two motors, a stay rope sensor is fixed on the side face of the cutting knife, a pulley frame is connected with a stay rope of the stay rope sensor, pulleys are hinged to the pulley frame, a simulated tea awning surface sliding groove formed by the simulated tea awning surface underframe and the pulleys form a groove pin pair through a pin shaft, connecting rods are hinged to the positions of two ends of the simulated tea awning surface sliding groove, the connecting rods at two ends are respectively hinged to two lower sliding blocks, the two lower sliding blocks and the two sides of the frame respectively form sliding pairs along the vertical direction, and are respectively fixed with the two sides of the frame through bolts. Posture sensors are arranged on the cutter and the simulated tea awning surface underframe; the cutter is provided with a two-dimensional laser radar. The verification method of the precision verification mechanism of the tea awning distance detection algorithm comprises the following specific steps: The method comprises the steps of firstly, placing a simulated tea awning surface on a simulated tea awning surface underframe, adjusting the horizontal inclination angle of the simulated tea awning surface underframe to a set value, enabling the tw