CN112102258A - Air-suction type seeder seeding detection method based on machine vision - Google Patents

Abstract

The invention relates to a sowing machine sowing inspection method, in particular to a machine vision-based air-suction type sowing machine sowing inspection method. The method comprises the steps of acquiring pictures simultaneously containing a suction hole and a process groove of a seeder by a camera; calibrating the position of the process tank in the picture; secondly, when the seeder works, determining whether new suction holes appear according to the gray value distribution of the sampling point set, and recording the new suction holes as a pulse when the new suction holes appear; then, cutting, histogram equalization and self-adaptive thresholding operations are carried out on the picture to obtain a binary image; then, judging whether the suction hole is a cavity or not according to the binary image; then, marking pictures with different colors according to whether the suction holes are cavities or not and displaying the pictures; and finally, calculating the miss-seeding rate. The detection method is not easily influenced by the environment, is not easy to generate misjudgment, and has higher correctness and better intuition of the detection result.

Description

Air-suction type seeder seeding detection method based on machine vision

Technical Field

The invention relates to a sowing machine sowing inspection method, in particular to a machine vision-based air-suction type sowing machine sowing inspection method.

Background

For the air-suction type seeder, the miss-seeding rate is an important parameter, and how to carry out seed sowing detection and count the miss-seeding rate is an important proposition. At present, in an existing air-suction type seeder system, sensors such as eddy current sensors, photoelectric sensors and infrared sensors are generally used for detecting seed leakage, the precision of the mode is influenced by the environment, misjudgment is easy to occur, the correctness of a detection result is low, and the intuitiveness is poor.

Disclosure of Invention

The invention aims to solve the technical problem of providing a machine vision-based air-aspiration seeder seeding detection method, which is not easily influenced by environment and is not easy to generate misjudgment, and has higher correctness of detection results and better intuition.

In order to solve the problems, the following technical scheme is provided:

the invention relates to a machine vision-based air-suction seeder seeding detection method, which is characterized by comprising the following steps:

step 1, a camera is installed on the seeder, and the camera can acquire pictures containing the suction hole and the process groove of the seeder.

Step 2, calibrating the position of the process tank in the picture to obtain a sampling point set { P }_i}, sample point P_iThe coordinates x, y of (a) represent the pixel coordinates in the picture.

Step 3, the seeder works, the suction holes and the process grooves rotate along with the seeding disc of the seeder, and the { P is set according to the sampling points_iDetermining whether a new suction hole appears according to the gray value distribution, and recording as a pulse when the new suction hole appears, wherein the method specifically comprises the following steps:

step 301, convert the picture into a grayscale I_grey。

Step 302, each sampling point P_i(x_i，y_i) Can correspond to a gray scale image I_greyGray value G of one point_iThen get the gray value set { G }_iIn which G_i(x_i，y_i)＝I_grey(x＝x_i，y＝y_i)。

Step 303, set of sampling points { P }_iDividing the process into a set of points (P) which are positioned in the process tank and have low gray value_inAnd a set of points outside the process tank with high gray scale values P_outI.e. that

if G_i≤T_tr P_i∈{P_in}

if G_i＞T_tr P_i∈{P_out}

Wherein, T_trIs a preset threshold value to distinguish between high and low gray values.

Step 304, initially setting a pulse state S_pulseFalse and a hole state S_emptyWhen { P ═ False_inThe number of points is greater than lambdan_PWhen it is, let S_pulseTrue, then { P }_outThe number of points is greater than gamma N_PWhen it is, let S_pulseFalse and the number of pulses N_puslePlus one, if the suction hole is judged to be a cavity at the moment, the number of the cavities is N_emptyAdd one, namely

St.S_pulse＝False，S_empty＝False，N_pulse＝0，N_empty＝0

if S_pulse＝False and N_in＞λN_P

{S_pulse＝True

if S_pulse＝True and N_out＞γN_P

Wherein N is_in、N_out、N_PAre respectively a set of points { P_in}、{P_out}、{P_iThe number of the middle points, λ and γ are two preset thresholds.

Step 4, performing cutting, histogram equalization and adaptive thresholding operations on the picture to obtain a binary image, specifically comprising the following steps:

when S is_pulseTrue and { P ═ P_inThe number of points is greater than lambdan_PTo gray scale image I_greyCutting to obtain an interested area I_ROIThen histogram equalization is carried out to obtain an image I with enhanced contrast_histFinally, self-adaptive thresholding operation is carried out to obtain a binary image I_bin，

I.e. I_ROI＝Rect(I_grey)，I_hist＝equalizeHist(I_ROI)，I_bin＝adaptiveThreshold(I_hist)。

When S is_pulseTrue and { P ═ P_inThe number of points is less than lambdan_PThen, the process returns to step 301.

Step 5, according to the binary image I_binAnd determining whether the suction hole is a hole, namely:

if sum(I_bin)＞T_e

S_empty＝True

wherein, sum (I)_bin) Is I_binPixel sum of (1), T_eThe threshold is preset to determine whether the hole is present.

Step 6, marking pictures with different colors according to whether the suction holes are cavities or not and displaying the outer disc

Step 7, according to the pulse number N_pulseAnd number of holes N_emptyCalculating the rate rho of missed seeding by the formula

And 8, repeating the steps 3 to 7 until the air-suction seeder stops operating, and finishing the seeding detection of the operation of the air-suction seeder.

In the step 2, the position of the process tank in the picture is calibrated manually.

By adopting the scheme, the method has the following advantages:

the invention discloses a machine vision-based air-suction seeder seeding detection method, which adopts a camera to acquire pictures simultaneously containing a suction hole and a process groove of a seeder; calibrating the position of the process tank in the picture; secondly, when the seeder works, determining whether new suction holes appear according to the gray value distribution of the sampling point set, and recording the new suction holes as a pulse when the new suction holes appear; then, cutting, histogram equalization and self-adaptive thresholding operations are carried out on the picture to obtain a binary image; then, judging whether the suction hole is a cavity or not according to the binary image; then, marking pictures with different colors according to whether the suction holes are cavities or not and displaying the pictures; and finally, calculating the miss-seeding rate. The detection method is based on machine vision, is less influenced by the environment, and is not easy to generate misjudgment, so that the correctness of the detection result is greatly improved. And whether the suction holes are holes or not is marked with pictures in different colors and displayed, so that the intuitiveness is better.

Drawings

FIG. 1 is a schematic structural diagram of a seeding tray in the sowing detection method of the air-suction seeder based on machine vision;

FIG. 2 is an enlarged schematic view of a camera capturing picture in the sowing detection method of the air-suction seeder based on machine vision;

fig. 3 is a flow chart of a software program of the air-suction type seeder seeding detection method based on machine vision.

Detailed Description

The present invention will be described in further detail with reference to the accompanying drawings.

As shown in fig. 1 to 3, the machine vision-based air-suction seeder seeding detection method of the present invention includes the following steps:

step 1, a camera is installed on the seeder, and the camera can acquire a picture 5 containing the suction hole 2 of the seeder and the process groove 1. The seeding disc 3 of seeder is located the outer dish 4 of seeder, and outer dish 4 is as the background of seeding disc 3, for the convenient colour of distinguishing picture 5, and the seeder outer dish 4 is black.

Step 2, manually calibrating the position of the process tank 1 in the picture 5 to obtain a sampling point set { P }_i}, sample point P_iThe coordinates x, y of (a) represent the pixel coordinates in the picture 5.

Step 3, the seeder works, the suction holes 2 and the process grooves 1 rotate along with the seeding tray 3 of the seeder, and the { P is set according to the sampling points_iDetermining whether a new suction hole 2 appears according to the gray value distribution, and recording the new suction hole 2 as a pulse when the new suction hole appears, wherein the method specifically comprises the following steps:

step 301, convert picture 5 into grayscale I_grey。

Step 302, each sampling point P_i(x_i，y_i) Can correspond to a gray scale image I_greyGray value G of one point_iThen get the gray value set { G }_iIn which G_i(x_i，y_i)＝I_grey(x＝x_i，y＝y_i)。

Step 303, set of sampling points { P }_iIs divided into a point set P which is positioned in the process tank 1 and has low gray value_inAnd a set of points (P) outside the process tank 1 with high gray scale values_outI.e. that

if G_i≤T_tr P_i∈{P_in}

if G_i＞T_tr P_i∈{P_out}

Wherein, T_trIs a preset threshold value to distinguish between high and low gray values.

Step 304, initially setting a pulse state S_pulseFalse and a hole state S_emptyWhen { P ═ False_inThe number of points is greater than lambdan_PWhen it is, let S_pulseTrue, then { P }_outThe number of points is greater than gamma N_PWhen it is, let S_pulseFalse and the number of pulses N_puslePlus one, if the suction hole 2 is judged to be a cavity at the moment, the number of the cavities N is_emptyAdd one, namely

St.S_pulse＝False，S_empty＝False，N_pulse＝0，N_empty＝0

if S_pulse＝False and Nⁱⁿ＞λN_P

{S_pulse＝True

if S_pulse＝True and N_out＞γN_P

Wherein N is_in、N_out、N_PAre respectively a set of points { P_in}、{P_out}、{P_iThe number of the middle points, λ and γ are two preset thresholds.

Step 4, performing cutting, histogram equalization and adaptive thresholding operations on the picture 5 to obtain a binary image, specifically:

when S is_pulseTrue and { P ═ P_inThe number of points is greater than lambdan_PTo gray scale image I_greyCutting to obtain an interested area I_ROIThen histogram equalization is carried out to obtain an image I with enhanced contrast_histFinally, self-adaptive thresholding operation is carried out to obtain a binary image I_bin，

I.e. I_ROI＝Rect(I_grey)，I_hist＝equalizeHist(I_ROI)，I_bin＝adaptiveThreshold(I_hist)；

When S is_pulseTrue and { P ═ P_inThe number of points is less than lambdan_pThen, the process returns to step 301.

Step 5, according to the binary image I_binAnd whether the suction hole 2 is a hole, namely:

if sum(I_bin)＞T_e

S_empty＝True

wherein, sum (I)_bin) Is I_binPixel sum of (1), T_eThe threshold is preset to determine whether the hole is present.

And 6, marking the picture 5 with different colors according to whether the suction hole 2 is a cavity or not and displaying the picture.

Step 7, according to the pulse number N_pulseAnd number of holes N_emptyCalculating the rate rho of missed seeding by the formula

And 8, repeating the steps 3 to 7 until the air-suction seeder stops operating, and finishing the seeding detection of the operation of the air-suction seeder.

The Rect, equalizehost and adaptiveThreshold are all common algorithms for image processing, and corresponding implementations can be found in a relevant library.

The air-suction seeder seeding detection method based on machine vision marks air holes with different colors according to whether the air holes are holes, and simultaneously directly displays pictures, so that the intuition is strong. Moreover, the influence of the environment is small, and misjudgment is not easy to generate, so that the correctness of the detection result is greatly improved.

Claims (2)

1. A machine vision-based sowing detection method for an air-suction type seeder is characterized by comprising the following steps:

step 1, mounting a camera on a seeder, wherein the camera can acquire pictures simultaneously including a suction hole and a process groove of the seeder;

step 2, calibrating the position of the process tank in the picture to obtain a sampling point set { P }_i}, sample point P_iThe coordinates x, y of (a) represent the pixel coordinates in the picture;

step 3, the seeder works, the suction holes and the process grooves rotate along with the seeding disc of the seeder, and the { P is set according to the sampling points_iDetermining whether a new suction hole appears according to the gray value distribution, and recording as a pulse when the new suction hole appears, wherein the method specifically comprises the following steps:

step 301, convert the picture into a grayscale I_grey；

Step 302, each sampling point P_i(x_i，y_i) Can be used forCorresponding gray scale map I_greyGray value G of one point_iThen get the gray value set { G }_iIn which G_i(x_i，y_i)＝I_grey(x＝x_i，y＝y_i)；

Step 303, set of sampling points { P }_iDividing the process into a set of points (P) which are positioned in the process tank and have low gray value_inAnd a set of points outside the process tank with high gray scale values P_outI.e. that

if G_i≤T_tr P_i∈{P_in}

if G_i＞T_tr P_i∈{P_out}

Wherein, T_trA preset threshold value is set so as to distinguish the gray value;

step 304, initially setting a pulse state S_pulseFalse and a hole state S_emptyWhen { P ═ False_inThe number of points is greater than lambdan_PWhen it is, let S_pulseTrue, then { P }_outThe number of points is greater than gamma N_PWhen it is, let S_pulseFalse and the number of pulses N_puslePlus one, if the suction hole is judged to be a cavity at the moment, the number of the cavities is N_emptyAdd one, namely

St.S_pulse＝False，S_empty＝False，N_pulse＝0，N_empty＝0

if S_pulse＝False and N_in＞λN_P

{S_pulse＝True

if S_pulse＝True and N_out＞γN_P

Wherein N is_in、N_out、N_PAre respectively a set of points { P_in}、{P_out}、{P_iThe number of the middle points, lambda and gamma are two preset thresholds;

step 4, performing cutting, histogram equalization and adaptive thresholding operations on the picture to obtain a binary image, specifically comprising the following steps:

when S is_pulseTrue and { P ═ P_inThe number of points is greater than lambdan_pTo gray scale image I_greyCutting to obtain an interested area I_ROIThen histogram equalization is carried out to obtain an image I with enhanced contrast_histFinally, self-adaptive thresholding operation is carried out to obtain a binary image I_bin，

I.e. I_ROI＝Rect(I_grey)，I_hist＝equalizeHist(I_ROI)，I_bin＝adaptiveThreshold(I_hist)；

When S is_pulseTrue and { P ═ P_inThe number of points is less than lambdan_pIf yes, go back to step 301;

step 5, according to the binary image I_binAnd determining whether the suction hole is a hole, namely:

if sum(I_bin)＞T_e

S_empty＝True

wherein, sum (I)_bin) Is I_binPixel sum of (1), T_eA preset threshold value for judging whether the hole exists or not;

step 6, marking pictures with different colors according to whether the suction holes are cavities or not and displaying the pictures;

step 7, according to the pulse number N_pulsAnd number of holes N_emptyCalculating the rate rho of missed seeding by the formula

And 8, repeating the steps 3 to 7 until the air-suction seeder stops operating, and finishing the seeding detection of the operation of the air-suction seeder.

2. The machine vision-based air-suction seeder seeding detection method according to claim 1, wherein in the step 2, the position of the process groove in the picture is manually calibrated.

Images