A three-dimensional space visual positioning method, system and device based on non-complete mushroom images

Abstract

The present invention relates to the field of three-dimensional space visual positioning, and discloses a three-dimensional space visual positioning method, system and device based on an incomplete mushroom image, comprising an image acquisition module, a distance measurement module and a main control module; the image acquisition module is used to acquire image information of mushrooms to be detected; the distance measurement module is used to measure the vertical distance between the image acquisition module and the mushrooms to be detected; the main control module is used to pre-process the acquired image, and then identify and locate the mushrooms in the pre-processed image, and then perform integrity detection on the identified mushrooms. During identification and positioning, the center points of the mushrooms in the image are determined according to the image integrity, and finally the three-dimensional coordinates of the mushrooms are generated. Compared with the prior art, the method proposed by the present invention can realize the positioning of mushrooms with incomplete image acquisition and further improve the positioning accuracy of mushrooms.

Description

Technical Field

[0001] The present invention relates to the field of three-dimensional space visual positioning, and in particular to the application of positioning based on non-complete mushroom images, namely, a three-dimensional space visual positioning method, system and device based on non-complete mushroom images. Background Art

[0002] In recent years, the mushroom picking method has shifted from manual picking to robot picking. The most mature mushroom picking method is the robot method based on a monocular camera. However, the robot method based on a monocular camera still has a certain deviation in positioning accuracy when positioning mushrooms due to occlusion and other problems, and usually directly abandons positioning for mushrooms with incomplete image acquisition. In this way, the efficiency of the entire picking process is greatly reduced, and its accuracy is not high. There may be cases of missed picking or wrong picking during the picking process. For the incomplete images collected, how to perform three-dimensional spatial visual positioning of mushrooms and improve positioning accuracy is an urgent problem to be solved. Summary of the invention

[0003] Technical purpose: In order to solve the problems existing in the prior art, the present invention provides a three-dimensional space visual positioning method, system and device for incompletely collected mushroom images, which makes up for the shortcomings of the robot operation mode based on monocular camera in terms of positioning accuracy and positioning processing of incompletely collected mushroom images.

[0004] Technical solution: The present invention 1. A three-dimensional space visual positioning method based on a non-complete mushroom image, characterized in that it includes the following steps:

[0005] Step 1: Fix the shooting height H, perform camera calibration, and obtain calibration parameters. The camera includes a left camera and a right camera. The calibration parameters include an intrinsic parameter matrix N, an extrinsic parameter rotation matrix W, and a translation matrix B.

[0006] Step 2: Capture images of the mushrooms to be detected, and perform image preprocessing operations on the images captured by the left camera;

[0007] Step 3: Pass the preprocessed image into the neural network model to identify the mushrooms in the image and generate a bounding box to determine the location of the mushrooms in the image;

[0008] Step 4: According to the bounding box generated in step 3, determine the integrity r of each mushroom in the image, calculate the center point of the measured mushroom, thereby obtaining the pixel coordinates of the measured mushroom and calculating the world coordinates of the mushroom;

[0009] Step 5: Collect the distance from the left camera to the mushroom, calculate the height of the measured mushroom, and obtain the three-dimensional coordinates of the measured mushroom;

[0010] Step 6: Perform the same steps 2 to 5 on the image captured by the right camera, and make a one-to-one correspondence between the image captured by the right camera and the image captured by the left camera to obtain the coordinates of the center point of the mushroom in the image captured by the left camera. w The corresponding point O k , O k Used to fine tune O w The coordinates of the final three-dimensional coordinates O f :

[0011] 2. According to claim 1, the three-dimensional space visual positioning method based on incomplete mushroom images is characterized in that the preprocessing operation in step 2 is: scaling, rotation, and then image enhancement to obtain an image data set for network training.

[0012] 3. According to claim 1, the three-dimensional space visual positioning method based on the incomplete mushroom image is characterized in that the neural network model used in step 3 is a YOLOv3 network model, which recognizes and locates the preprocessed image and generates a bounding box for each mushroom in the image.

[0013] 4. The three-dimensional space visual positioning method based on the incomplete mushroom image according to claim 1 is characterized in that the specific steps of calculating the center point of the measured mushroom in step 4 are as follows:

[0014] Step 4.1: Based on the width of the bounding box w The height of the bounding box h , calculate the completeness of the mushrooms in the collected image

[0015] Step 4.2: The mushroom images with integrity r lower than 0.25 or higher than 4 in the image captured by the left camera are recorded as pseudo targets, and the pseudo targets are removed;

[0016] Step 4.3: The mushrooms and soil in the image are respectively regarded as the foreground object and background object of the image, and the pixels in the image are classified into two clusters using the clustering algorithm K-means, so as to separate the mushroom image from the soil image in the collected image;

[0017] Step 4.4: Perform pixel-level contour detection on the separated mushroom image, and use the prior circle to fit the contour of the mushroom image to obtain the contour of the complete mushroom image;

[0018] Step 4.5: When the integrity r of the mushrooms in the collected image is 1 or in the interval (0.5, 2), in the mushroom image region Q ij Search for the center point O l , o l= argmax{Q ij}; where i is the number of rows to search and j is the number of columns to search; otherwise, go to step 4.4;

[0019] Step 4.6: When the mushroom integrity r in the acquired image is in the interval (0.25, 0.5)U(2, 4), calculate the corresponding center point.

[0020] 5. The three-dimensional space visual positioning method based on the incomplete mushroom image according to claim 4 is characterized in that the step of determining the contour of the mushroom image according to the calculated contour edge in step 4.4 comprises the following steps:

[0021] Step 4.4.1: Operate the foreground area with the edge detection filter to extract the pixel-level contour edges of the mushrooms in the captured image;

[0022] Step 4.4.2: Take each pixel of the obtained contour edge as an anchor point, and assume that these anchor points are points in the prior circle of the mushroom, so that each anchor point corresponds to a prior circle;

[0023] Step 4.4.3: Calculate the number of pixels of the prior circles, and select the prior circle with the largest common mushroom image area among all the prior circles as the outline of the mushroom image.

[0024] 6. The three-dimensional space visual positioning method based on the incomplete mushroom image according to claim 5 is characterized in that the calculation method of the center point of the incomplete mushroom image with integrity in the interval (0.25, 0.5)U(2, 4) in step 4.6 comprises the following steps:

[0025] Step 4.6.1: If the integrity of the mushroom image is in the interval (0.25, 0.5)U(2, 4), fill the image captured by the left camera with a certain pixel value P in width and height with pure white, and the range of P is (25, 100). Establish a pixel coordinate system with the upper left corner of the filled image as the origin;

[0026] Step 4.6.2: Select four anchor point coordinates a(a x , a y )、b(b x , b y )、c(c x , c y )、d(d x , d y ) can be obtained respectively as a(a x , a y )、b(b x , b y ) and c(c x , c y)、d(d x , d y ) can be calculated by the perpendicular bisectors l1 and l2 to obtain the intersection point O of the two lines. L (O Lx , O Ly ), which is the pixel coordinate of the center point of the mushroom image, where:

[0027]

[0028]

[0029]

[0030]

[0031] Among them, x1 is the horizontal coordinate of the point that l1 passes through, y1 is the vertical coordinate of the point that l1 passes through, x2 is the horizontal coordinate of the point that l2 passes through, y2 is the vertical coordinate of the point that l2 passes through;

[0032] Step 4.6.3: Calculate the height O of the mushroom based on the vertical distance h between the camera and the mushroom. wz :

[0033] O wz =Hh

[0034] Where H is the fixed shooting height;

[0035] Step 4.6.4: The pixel coordinates of the obtained center point can be obtained by calculation L , converted into world three-dimensional coordinates O w :

[0036] O w =W -1 N -1 T z O L +W -1 B

[0037] Among them, T z is the distance from the optical center to the mushroom in the camera coordinate system: O wz is the height of the mushroom.

[0038] The present invention also discloses a three-dimensional space visual positioning system based on non-complete mushroom images, including an image acquisition module, a distance measurement module, and a main control module;

[0039] The image acquisition module is used to acquire image information of the mushrooms to be detected; it includes a left camera and a right camera, the left camera and the right camera jointly acquire image data of the same mushroom object, and perform fine-tuning one by one;

[0040] The distance measuring module is used to measure the distance between the image acquisition module and the mushroom to be detected;

[0041] The main control module includes an image preprocessing unit, a target detection unit, an integrity detection unit, and a coordinate generation unit:

[0042] The image preprocessing unit is used to preprocess the collected image;

[0043] The target detection unit is used to identify and locate mushrooms in the preprocessed image;

[0044] The integrity detection unit is used to detect the integrity of the mushrooms in the captured image;

[0045] The coordinate generating unit is used to generate the three-dimensional coordinates of the measured mushroom.

[0046] Preferably, the specific operations of the image preprocessing unit, the target detection unit, the integrity detection unit, and the coordinate generation unit of the main control module are:

[0047] The image preprocessing unit is used to scale and rotate the collected image, and then perform image enhancement to obtain an image data set for network training;

[0048] The object detection unit is used to identify and locate the mushroom object in the preprocessed image, pass the image into the neural network model, identify the mushroom in the image and generate a bounding box to determine the position of the mushroom in the image;

[0049] The integrity detection unit is used to detect and determine the integrity of the mushrooms in the captured image. When the mushroom image has an integrity r lower than 0.25 or higher than 4, it is removed as a pseudo target; for the mushroom image with an integrity r in the interval (0.5, 2) and an integrity of 1, the center point is searched within the mushroom outline; when the integrity r of the mushroom in the measured image is in the interval (0.25, 0.5)U(2, 4), the center point of the incomplete mushroom image is calculated; the coordinate generation unit is used to convert the pixel coordinates of the center point of the measured mushroom, and calculate the height of the mushroom according to the vertical distance from the image acquisition module to the mushroom measured by the ranging module, so as to obtain the three-dimensional coordinates of the measured mushroom.

[0050] Preferably, the distance measuring module comprises a laser distance measuring sensor for measuring the vertical distance from the image acquisition module to the mushroom so as to calculate the height of the measured mushroom.

[0051] The present invention also discloses a three-dimensional space visual positioning device based on a non-complete mushroom image, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the three-dimensional space visual positioning method based on a non-complete mushroom image as described in any one of claims 1 to 6 when executing the computer program.

[0052] Beneficial effects:

[0053] 1. The technical solution of the present invention solves the problem of locating mushrooms with incomplete image acquisition and the problem of the offset of the measured mushroom center point caused by occlusion. The present invention can locate mushrooms with incomplete image acquisition and further improve the positioning accuracy of mushrooms through a three-dimensional space visual positioning method, system and device based on incomplete mushroom images.

[0054] 2. Although the present invention cannot obtain the complete outline of the mushrooms due to incomplete images caused by shooting or occlusion between mushrooms or images of mushrooms covered with soil when using the clustering algorithm, the present invention uses a priori circles to fit the outlines of the mushroom images, so that the mushrooms can obtain complete outlines. Then we take points on the outline to calculate the position of the center point of the mushroom, and finally convert it into three-dimensional coordinates. Using the three-dimensional coordinates for positioning, the positioning accuracy is greatly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a schematic diagram of the process of the three-dimensional space visual positioning method based on the incomplete mushroom image of the present invention;

[0056] Figure 2 The structure block diagram of the three-dimensional space visual positioning system based on the incomplete mushroom image of the present invention;

[0057] Figure 3 A schematic diagram of an edge detection filter provided by an embodiment of the present invention;

[0058] Figure 4 A schematic diagram of the principle of calculating the center point of an incomplete mushroom image provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0059] The embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings, but it should be understood that the protection scope of the present invention is not limited by the specific implementation methods.

[0060] Reference below Figure 1 The present invention describes in detail a three-dimensional space visual positioning method based on a non-complete mushroom image. Taking Agaricus bisporus as an example, the present invention discloses a three-dimensional space visual positioning system based on a non-complete mushroom image. Figure 2 , the positioning system mainly includes:

[0061] Image acquisition module, ranging module, main control module;

[0062] The image acquisition module is used to acquire image information of the mushrooms to be detected; it includes a left camera 1-1 and a right camera 1-2, the left camera 1-1 and the right camera 1-2 jointly acquire image data of the same mushroom object, and perform fine-tuning one by one;

[0063] The distance measuring module is used to measure the distance between the image acquisition module and the mushroom to be detected; the distance measuring module includes a laser distance measuring sensor 2-1, which is used to measure the vertical distance between the image acquisition module and the mushroom to be detected so as to calculate the height of the mushroom to be detected.

[0064] The main control module includes an image preprocessing unit 3-1, a target detection unit 3-2, an integrity detection unit 3-3, and a coordinate generation unit 3-4:

[0065] The image preprocessing unit 3-1 is used to preprocess the collected images, including scaling and rotating the collected images, and then performing image enhancement to obtain an image data set for network training.

[0066] The target detection unit 3-2 is used to identify and locate the mushrooms in the preprocessed image; the image is passed into the neural network model, the mushrooms in the image are identified and a bounding box is generated to determine the position of the mushrooms in the image.

[0067] The integrity detection unit 3-3 is used to detect the integrity of the mushrooms in the captured image; when the integrity r is lower than 0.25 or higher than 4, the mushroom image is removed as a pseudo target; for the mushroom image with integrity r in the interval (0.5, 2) and the integrity of 1, the center point is searched within the mushroom outline; when the integrity r of the mushroom in the measured image is in the interval (0.25, 0.5)U(2, 4), the center point of the incomplete mushroom image is calculated.

[0068] The coordinate generating unit 3-4 is used to generate the three-dimensional coordinates of the measured mushroom, to convert the pixel coordinates of the center point of the measured mushroom, and to calculate the height of the mushroom according to the vertical distance from the image acquisition module to the mushroom measured by the distance measuring module, so as to obtain the three-dimensional coordinates of the measured mushroom.

[0069] like Figure 1 As shown, the method disclosed in this embodiment includes the following steps:

[0070] Step 1: First, fix the shooting height H of the image acquisition module, print a chessboard as a calibration object, calibrate the image acquisition module, and calculate the camera's intrinsic parameter matrix N, extrinsic parameter rotation matrix W, and translation matrix B.

[0071] Step 2: Use the image acquisition module to acquire images of the Agaricus bisporus to be detected, and the image acquired by the left camera in the image acquisition module is sent to the image processing unit of the main control module for scaling and rotation, and then image enhancement is performed to complete image preprocessing.

[0072] Step 3: The preprocessed image is passed into the neural network model. In this embodiment, YOL0v3 is used as an example to identify and locate the preprocessed image, so as to identify the Agaricus bisporus in the image and generate a bounding box to determine the position of the Agaricus bisporus in the image.

[0073] Step 4: According to the generated bounding box, the integrity r of each Agaricus bisporus in the image is determined, and the center point of the measured Agaricus bisporus is calculated, thereby obtaining the pixel coordinates of the measured Agaricus bisporus, and then the world coordinates of the Agaricus bisporus can be calculated.

[0074] Step 4.1: Based on the width of the bounding box w The height of the bounding box h , calculate the integrity r of Agaricus bisporus in the collected image:

[0075] Step 4.2: The Agaricus bisporus images with integrity r lower than 0.25 or higher than 4 in the images captured by the left camera are recorded as pseudo targets, and the pseudo targets are removed.

[0076] Step 4.3: The Agaricus bisporus and soil in the image are respectively regarded as the foreground object and background object of the image, and the pixels in the image are classified into two clusters using a clustering algorithm, such as K-means, so as to separate the Agaricus bisporus image from the soil image in the collected image.

[0077] Step 4.4: Perform pixel-level contour detection on the separated Agaricus bisporus image, and use the prior circle to fit the contour of the Agaricus bisporus image, so as to obtain the contour of the complete Agaricus bisporus image and reduce the impact of occlusion between Agaricus bisporus and occlusion of Agaricus bisporus by soil.

[0078] Step 4.4.1: Align the foreground region with Figure 3 The edge detection filter shown is operated to extract the pixel-level contour edge of the Agaricus bisporus in the captured image.

[0079] Step 4.4.2: Take each pixel of the obtained contour edge as an anchor point, and assume that these anchor points are points in the prior circle of Agaricus bisporus, so that each anchor point corresponds to a prior circle.

[0080] Step 4.4.3: Calculate the number of pixels of the priori circle, and select the priori circle with the largest Agaricus bisporus image area among all the priori circles as the contour of the Agaricus bisporus image.

[0081] Step 4.5: When the integrity of Agaricus bisporus in the collected image is 1 or in the interval (0.5, 2), in the Agaricus bisporus image area Q ij Search for the center point o l = argmax{Q ij}, where i is the number of rows to search and j is the number of columns to search. Otherwise, go to step 4.4.

[0082] Step 4.6: When the integrity of the Agaricus bisporus in the acquired image is in the interval of (0.25, 0.5)U(2, 4), calculate the corresponding center point.

[0083] The method for calculating the center point of the incomplete mushroom image with integrity in the interval (0.25, 0.5)U(2, 4) in step 4.6 includes the following steps:

[0084] Step 4.6.1: If the integrity of the Agaricus bisporus image is in the interval (0.25, 0.5)U(2, 4), then Figure 4 As shown, the image captured by the left camera is filled with a certain pixel value P in width and height with pure white, and the range of P is (25, 100). The filling pixel value of 25 is taken as an example for explanation, and a pixel coordinate system is established with the upper left corner of the filled image as the origin.

[0085] Step 4.6.2: If Figure 4 As shown in the figure, take any four anchor point coordinates a(a x , a y )、b(b x , b y )、c(c x , c y )、d(d x , d y ) can be obtained respectively as a(a x , a y )、b(b x , b y ) and c(c x , c y )、d(d x , d y ) can be calculated by the perpendicular bisectors l1 and l2 to obtain the intersection point O of the two lines. L (O Lx , O Ly ), which is the pixel coordinate of the center point of the Agaricus bisporus image.

[0086] in,

[0087]

[0088]

[0089]

[0090]

[0091] Among them, x1 is the horizontal coordinate of the point through which l1 passes, y1 is the vertical coordinate of the point through which l1 passes, x2 is the horizontal coordinate of the point through which l2 passes, and y2 is the vertical coordinate of the point through which l2 passes.

[0092] Step 4.6.3: The vertical distance h between the image acquisition module and the Agaricus bisporus can be measured by the distance measurement module, and then the height O of the Agaricus bisporus can be calculated. wz :

[0093] O wz =Hh

[0094] Where H is the fixed shooting height;

[0095] Step 4.6.4: The pixel coordinates of the obtained center point can be obtained by calculation L , converted into world three-dimensional coordinates O w :

[0096] O w =W -1 N -1 T z O L +W -1 B

[0097] Among them, T z is the distance from the optical center to the Agaricus bisporus in the camera coordinate system:

[0098]

[0099] Step 5: Measure the distance from the left camera in the image acquisition module to the Agaricus bisporus according to the laser distance measurement sensor of the distance measurement module, calculate the height of the Agaricus bisporus measured, and then obtain the three-dimensional coordinates of the Agaricus bisporus measured.

[0100] Step 6: Perform the same steps 2 to 5 on the image captured by the right camera in the image acquisition module, and make a one-to-one correspondence between the captured image and the image captured by the left camera, so as to obtain the coordinates of the center point of the Agaricus bisporus in the image captured by the left camera. w The corresponding point O k , O k Used to fine tune O w The coordinates of the final three-dimensional coordinates O f :

[0101] The above embodiments are only for illustrating the technical concept and features of the present invention, and their purpose is to enable people familiar with the technology to understand the content of the present invention and implement it accordingly, and they cannot be used to limit the protection scope of the present invention. Any equivalent transformation or modification made according to the spirit of the present invention should be included in the protection scope of the present invention.

Claims

1. A three-dimensional space visual positioning method based on non-complete mushroom images, characterized in that: The following steps are involved: Step 1: Fix the shooting height H, perform camera calibration, and obtain calibration parameters. The camera includes a left camera and a right camera. The calibration parameters include an intrinsic parameter matrix N, an extrinsic parameter rotation matrix W, and a translation matrix B. Step 2: Capture images of the mushrooms to be detected, and perform image preprocessing operations on the images captured by the left camera; Step 3: Pass the preprocessed image into the neural network model to identify the mushrooms in the image and generate a bounding box to determine the location of the mushrooms in the image; Step 4: According to the bounding box generated in step 3, determine the integrity r of each mushroom in the image, calculate the center point of the measured mushroom, thereby obtaining the pixel coordinates of the measured mushroom and calculating the world coordinates of the mushroom; The specific steps for calculating the center point of the measured mushroom are as follows: Step 4.1: Based on the width of the bounding box w The height of the bounding box h , calculate the completeness r of the mushrooms in the collected image: Step 4.2: The mushroom images with integrity r lower than 0.25 or higher than 4 in the image captured by the left camera are recorded as pseudo targets, and the pseudo targets are removed; Step 4.3: The mushrooms and soil in the image are respectively regarded as the foreground object and background object of the image, and the pixels in the image are classified into two clusters using the clustering algorithm K-means, so as to separate the mushroom image from the soil image in the collected image; Step 4.4: Perform pixel-level contour detection on the separated mushroom image, and use the prior circle to fit the contour of the mushroom image to obtain the contour of the complete mushroom image; Step 4.4.1: Operate the foreground area with the edge detection filter to extract the pixel-level contour edges of the mushrooms in the captured image; Step 4.4.2: Take each pixel of the obtained contour edge as an anchor point, and assume that these anchor points are points in the prior circle of the mushroom, so that each anchor point corresponds to a prior circle; Step 4.4.3: Calculate the number of pixels of the priori circles, and select the priori circle with the largest total mushroom image area among all the priori circles as the outline of the mushroom image; Step 4.5: When the integrity r of the mushrooms in the collected image is 1 or in the interval (0.5, 2), in the mushroom image region Q ij Search for the center point O l , O l = argmax{O ij }; where i is the number of rows to search and j is the number of columns to search; otherwise, go to step 4.4; Step 4.6: When the mushroom integrity r in the collected image is in the interval (0.25, 0.5) ∪ (2, 4), calculate the corresponding center point; Step 5: Collect the distance from the left camera to the mushroom, calculate the height of the measured mushroom, and obtain the three-dimensional coordinates of the measured mushroom; Step 6: Perform the same steps 2 to 5 on the image captured by the right camera, and make a one-to-one correspondence between the image captured by the right camera and the image captured by the left camera to obtain the coordinates of the center point of the mushroom in the image captured by the left camera. w The corresponding point O k , O k Used to fine tune O w The coordinates of the final three-dimensional coordinates O f :

2. The three-dimensional space visual positioning method based on the incomplete mushroom image according to claim 1 is characterized in that: The preprocessing operation in step 2 is: scaling, rotating, and then performing image enhancement to obtain an image data set for network training.

3. The three-dimensional space visual positioning method based on non-complete mushroom images according to claim 1 is characterized in that: The neural network model used in step 3 is a YOLOv3 network model, which recognizes and locates the preprocessed image and generates a bounding box for each mushroom in the image.

4. The three-dimensional space visual positioning method based on non-complete mushroom images according to claim 1 is characterized in that: The method for calculating the center point of the incomplete mushroom image in the interval (0.25, 0.5) ∪ (2, 4) in step 4.6 comprises the following steps: Step 4.6.1: If the integrity of the mushroom image is in the interval (0.25, 0.5) ∪ (2, 4), fill the image captured by the left camera with a certain pixel value P in width and height with pure white, and the range of P is (25, 100). Establish a pixel coordinate system with the upper left corner of the filled image as the origin; Step 4.6.2: Select four anchor point coordinates a(a x ,a y )、b(b x ,b y )、c(c x ,c y )、d(d x ,d y ) can be obtained respectively as a(a x ,a y )、b(b x ,b y ) and c(c x ,c y )、d(d x ,d y ) can be calculated by the perpendicular bisectors l1 and l2 to obtain the intersection point O of the two lines. L (O Lx ,O Ly ), which is the pixel coordinate of the center point of the mushroom image, where: Among them, x1 is the horizontal coordinate of the point that l1 passes through, y1 is the vertical coordinate of the point that l1 passes through, x2 is the horizontal coordinate of the point that l2 passes through, and y2 is the vertical coordinate of the point that l2 passes through; Step 4.6.3: Calculate the height O of the mushroom based on the vertical distance h between the camera and the mushroom. wz : O wz =H-h Where H is the fixed shooting height; Step 4.6.4: The pixel coordinates of the obtained center point can be obtained by calculation L , converted into world three-dimensional coordinates O w : O w =W -1 N -1 T z O L +W -1 B Among them, T z is the distance from the optical center to the mushroom in the camera coordinate system: O wz is the height of the mushroom.

5. A system based on the three-dimensional space visual positioning method based on the incomplete mushroom image according to any one of claims 1 to 4, characterized in that: Including image acquisition module, ranging module and main control module; The image acquisition module is used to acquire image information of the mushrooms to be detected; it includes a left camera and a right camera, the left camera and the right camera jointly acquire image data of the same mushroom object, and perform fine-tuning one by one; The distance measuring module is used to measure the distance between the image acquisition module and the mushroom to be detected; The main control module includes an image preprocessing unit, a target detection unit, an integrity detection unit, and a coordinate generation unit: The image preprocessing unit is used to preprocess the collected image; The target detection unit is used to identify and locate mushrooms in the preprocessed image; The integrity detection unit is used to detect the integrity of the mushrooms in the captured image; The coordinate generating unit is used to generate the three-dimensional coordinates of the measured mushroom.

6. The system according to claim 5, characterized in that The specific operations of the image preprocessing unit, target detection unit, integrity detection unit, and coordinate generation unit of the main control module are: The image preprocessing unit is used to scale and rotate the collected image, and then perform image enhancement to obtain an image data set for network training; The object detection unit is used to identify and locate the mushroom object in the preprocessed image, pass the image into the neural network model, identify the mushroom in the image and generate a bounding box to determine the position of the mushroom in the image; The integrity detection unit is used to detect and determine the integrity of the mushrooms in the captured image. When the integrity r is lower than 0.25 or higher than 4, the mushroom image is removed as a pseudo target; for the mushroom image with integrity r in the interval (0.5, 2) and the integrity of 1, the center point is searched within the mushroom outline; when the integrity r of the mushroom in the measured image is in the interval (0.25, 0.5) ∪ (2, 4), the center point of the incomplete mushroom image is calculated; the coordinate generation unit is used to convert the pixel coordinates of the center point of the measured mushroom, and calculate the height of the mushroom according to the vertical distance from the image acquisition module to the mushroom measured by the ranging module, so as to obtain the three-dimensional coordinates of the measured mushroom.

7. The system according to claim 6, characterized in that The distance measuring module includes a laser distance measuring sensor, which is used to measure the vertical distance from the image acquisition module to the mushroom so as to calculate the height of the measured mushroom.

8. A three-dimensional space visual positioning device based on an incomplete mushroom image, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the three-dimensional space visual positioning method based on the non-complete mushroom image are implemented as described in any one of claims 1 to 4.

Images