CN113458004A - Seed sorting device and method based on machine vision - Google Patents
Seed sorting device and method based on machine vision Download PDFInfo
- Publication number
- CN113458004A CN113458004A CN202110933281.6A CN202110933281A CN113458004A CN 113458004 A CN113458004 A CN 113458004A CN 202110933281 A CN202110933281 A CN 202110933281A CN 113458004 A CN113458004 A CN 113458004A
- Authority
- CN
- China
- Prior art keywords
- seed
- seeds
- suction nozzle
- suction
- control system
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B07—SEPARATING SOLIDS FROM SOLIDS; SORTING
- B07C—POSTAL SORTING; SORTING INDIVIDUAL ARTICLES, OR BULK MATERIAL FIT TO BE SORTED PIECE-MEAL, e.g. BY PICKING
- B07C5/00—Sorting according to a characteristic or feature of the articles or material being sorted, e.g. by control effected by devices which detect or measure such characteristic or feature; Sorting by manually actuated devices, e.g. switches
- B07C5/04—Sorting according to size
- B07C5/10—Sorting according to size measured by light-responsive means
-
- A—HUMAN NECESSITIES
- A01—AGRICULTURE; FORESTRY; ANIMAL HUSBANDRY; HUNTING; TRAPPING; FISHING
- A01C—PLANTING; SOWING; FERTILISING
- A01C1/00—Apparatus, or methods of use thereof, for testing or treating seed, roots, or the like, prior to sowing or planting
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B07—SEPARATING SOLIDS FROM SOLIDS; SORTING
- B07C—POSTAL SORTING; SORTING INDIVIDUAL ARTICLES, OR BULK MATERIAL FIT TO BE SORTED PIECE-MEAL, e.g. BY PICKING
- B07C5/00—Sorting according to a characteristic or feature of the articles or material being sorted, e.g. by control effected by devices which detect or measure such characteristic or feature; Sorting by manually actuated devices, e.g. switches
- B07C5/34—Sorting according to other particular properties
- B07C5/342—Sorting according to other particular properties according to optical properties, e.g. colour
- B07C5/3422—Sorting according to other particular properties according to optical properties, e.g. colour using video scanning devices, e.g. TV-cameras
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B07—SEPARATING SOLIDS FROM SOLIDS; SORTING
- B07C—POSTAL SORTING; SORTING INDIVIDUAL ARTICLES, OR BULK MATERIAL FIT TO BE SORTED PIECE-MEAL, e.g. BY PICKING
- B07C5/00—Sorting according to a characteristic or feature of the articles or material being sorted, e.g. by control effected by devices which detect or measure such characteristic or feature; Sorting by manually actuated devices, e.g. switches
- B07C5/34—Sorting according to other particular properties
- B07C5/342—Sorting according to other particular properties according to optical properties, e.g. colour
- B07C5/3425—Sorting according to other particular properties according to optical properties, e.g. colour of granular material, e.g. ore particles, grain
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B07—SEPARATING SOLIDS FROM SOLIDS; SORTING
- B07C—POSTAL SORTING; SORTING INDIVIDUAL ARTICLES, OR BULK MATERIAL FIT TO BE SORTED PIECE-MEAL, e.g. BY PICKING
- B07C5/00—Sorting according to a characteristic or feature of the articles or material being sorted, e.g. by control effected by devices which detect or measure such characteristic or feature; Sorting by manually actuated devices, e.g. switches
- B07C5/36—Sorting apparatus characterised by the means used for distribution
- B07C5/361—Processing or control devices therefor, e.g. escort memory
- B07C5/362—Separating or distributor mechanisms
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
- G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Geometry (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Soil Sciences (AREA)
- Environmental Sciences (AREA)
- Multimedia (AREA)
- Pretreatment Of Seeds And Plants (AREA)
Abstract
The invention provides a seed sorting device and method based on machine vision, wherein the device comprises: support frame, seed transmission band, negative pressure device, good kind storage box, image acquisition device, control system and seed suction means, set up on the support frame the seed transmission band, the top parallel arrangement of seed transmission band seed suction device and image acquisition device, seed suction device passes through the support frame setting and is in the center of seed transmission band, image acquisition device passes through the support frame setting in seed suction device's front side, negative pressure device is provided with the multiunit, negative pressure device connects seed suction device, good kind storage box sets up in the terminal bottom of seed transmission band. The seed sorting device and method based on machine vision provided by the invention have the advantages that the accuracy is high, the appearance shape of the seeds cannot be damaged, the nondestructive detection of the seeds can be realized, the bad seeds can be removed, and the good quality of the seeds is ensured.
Description
Technical Field
The invention relates to the technical field of modern agricultural equipment, in particular to a seed sorting device and method based on machine vision.
Background
With the rapid development of national economy in China, the requirements of people on the yield and quality of agriculture are continuously improved, wherein the quality of seeds directly influences the yield and quality of crops. Most of seed screening methods in the prior art are to distinguish the quality of seeds through human eyes, and because the individual seeds are not large, once the human eyes detect for a long time, visual fatigue is easy to generate, and mistaken picking is caused. Under the background, many manufacturers begin to develop mechanical seed sorting methods, and currently, the seed screening apparatuses mainly used in the methods include: the seed sorting device comprises a gravity screening machine, an air screening and selecting machine, a dielectric winnowing machine and the like, wherein the seed screened by the screening modes is not high in quality and cannot be used for screening with high precision, and the screening modes can influence the seed quality and cannot ensure nondestructive testing, so that the seed sorting device and the seed sorting method based on machine vision are very necessary to design.
Disclosure of Invention
The invention aims to provide a seed sorting device and method based on machine vision, which have high accuracy, do not damage the appearance shape of seeds, can realize nondestructive detection of the seeds, can remove bad seeds and ensure good quality of the seeds.
In order to achieve the purpose, the invention provides the following scheme:
a seed sorting device and method based on machine vision comprise: a support frame, a seed conveying belt, a negative pressure device, a good seed storage box, an image acquisition device, a control system and a seed suction device, wherein the seed conveying belt is arranged on the support frame, used for driving the seed to be transmitted, the top of the seed transmission belt is provided with the seed suction device and the image acquisition device in parallel, the seed suction device is arranged at the center of the seed conveying belt through the support frame, the image acquisition device is arranged at the front side of the seed suction device through the support frame, the negative pressure devices are provided with a plurality of groups and are connected with the seed suction device, used for providing suction force for the seed suction device, the good seed storage box is arranged at the bottom of the tail end of the seed conveying belt, the control system is used for receiving the seeds and is electrically connected with the negative pressure device, the image acquisition device and the seed suction device;
the seed suction device comprises a first connecting rod, a suction nozzle and a crank and rocker mechanism, the first connecting rod is fixedly arranged on the support frames on two sides of the top of the seed conveying belt, the suction nozzle and the crank and rocker mechanism are correspondingly provided with a plurality of suction nozzles, the crank and rocker mechanisms are fixedly arranged on the first connecting rod, the suction nozzle is fixedly arranged at the lower end of the crank and rocker mechanism and is connected with the negative pressure device through a suction nozzle pipeline for sucking the bad seeds, the crank and rocker mechanism is electrically connected with the control system, and the crank and rocker mechanism is used for driving the suction nozzle to move up and down;
the negative pressure device is internally detachably provided with a seed cleaning box for cleaning seeds, and the seed cleaning box is used for storing and cleaning seeds.
Optionally, the image acquisition device includes second connecting rod and industry camera, the second connecting rod is fixed to be set up on the support frame of seed transmission band top both sides, industry camera sets up the center of support frame, the fixed light source that is provided with on the industry camera.
Optionally, the one end that the seed transmission band was kept away from to the good kind storage box is provided with the shelf baffle for prevent that the seed is too fast, fall out the good kind storage box.
Optionally, the suction nozzle is connected with the suction nozzle pipeline through a suction nozzle pipeline sealing pipe, the suction nozzle pipeline is connected with the negative pressure device, the suction nozzle pipeline is a rigid pipeline, the suction nozzle pipeline sealing pipe is in clearance fit with the suction nozzle and the suction nozzle pipeline, and is sealed through an oil seal, so that the suction nozzle and the suction nozzle pipeline move up and down along with the crank rocker mechanism.
Optionally, the support frame is fixedly connected to the suction nozzle pipeline through a plurality of fixing rods, and is used for fixing the suction nozzle pipeline.
Optionally, the negative pressure device comprises a box body, a motor and a fan, a top cover is arranged at the top of the box body, the motor is fixedly arranged at the upper part inside the box body through a fixing piece, the fan is arranged on an output shaft of the motor, a filter plate is arranged at the bottom of the fan, the seed cleaning box is arranged at the bottom of the filter plate, a plurality of pipe holes are formed in the side wall of the box body between the filter plate and the seed cleaning box, the suction nozzle pipeline is communicated with the box body through the pipe holes, and the motor is electrically connected with the control system.
The invention also provides a seed sorting method based on machine vision, which is applied to the seed sorting device based on machine vision and comprises the following steps:
step 1: placing the seeds on a seed conveying belt, transmitting the seeds to the shooting range of an industrial camera, shooting the seeds by the industrial camera, and transmitting the shot images to a control system;
step 2: the control system processes the image, after the processing is finished, the shape characteristic component and the color characteristic component are extracted and compared with two components in a standard library preset by the control system, if the comparison difference value is in a qualified range, the seeds are judged to be good seeds, otherwise, the seeds are judged to be bad seeds;
and step 3: if the seeds are judged to be good seeds, the seeds are not subjected to other operations and fall into a good seed storage box through the conveying of a seed conveying belt, and if the seeds are judged to be bad seeds, the seeds are sucked into a bad seed processing box through a seed sucking device corresponding to an industrial camera for shooting images.
Optionally, in step 2, the control system processes the image, after the processing is completed, extracts the shape characteristic component and the color characteristic component, and compares the shape characteristic component and the color characteristic component with two components in a standard library preset by the control system, if a comparison difference value is within a qualified range, the seed is determined to be good, otherwise, the seed is determined to be bad, specifically:
the control system carries out smooth filtering processing on the image and is used for eliminating the influence of noise and uneven field light source irradiation on the image, and after the processing is finished, the control system extracts the shape characteristic analysis and the color characteristic component of the seeds;
the shape characteristic components comprise nine groups of data which are respectively perimeter, area, circularity, equivalent diameter, long axis, short axis, length-width ratio, compactness and matching value of the seeds, the shape characteristic components are verified by Minitab data processing software by applying a hypothesis test formula according to a standard library, the nine groups of data are judged to contain data which accord with normal distribution, a 3 delta standard interval is adopted for standard value range demarcation of the data which accord with the normal distribution, wherein the qualified range defined by the standard value is (mu-3 delta, mu +3 delta), and the qualified range is set by adopting a margin requirement of +/-20% on the data which do not accord with the normal distribution;
the color characteristic component is the average value of RGB three channels, the color characteristic component is verified through Minitab data processing software by applying a hypothesis test formula according to a standard library, whether the average value of the RGB three channels accords with normal distribution or not is judged, if the average value accords with the normal distribution, a 3 delta standard interval is adopted for standard value range demarcation, wherein the qualified range defined by the standard value is (mu-3 delta, mu +3 delta), and if the average value does not accord with the normal distribution, the qualified range is set by adopting a margin requirement of +/-20%.
Optionally, the method for establishing the standard library specifically includes:
selecting a plurality of standard seeds, extracting shape characteristic components and color characteristic components of the standard seeds, and establishing a standard database according to the shape characteristic components and the color characteristic components of the standard seeds.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects: the device adopts a suction nozzle, a suction nozzle pipeline and a negative pressure device to achieve the purpose of sorting bad seeds in a matched suction mode, has higher accuracy rate compared with a mechanical arm sorting mode and the like, is simple and convenient to operate, can prevent seeds from being damaged, and ensures the integrity of the seeds to the maximum extent; the device is provided with a crank rocker mechanism which can drive the suction nozzle to move up and down, so that the suction nozzle can suck the bad seeds conveniently; the device's suction nozzle pipeline and the junction of suction nozzle are provided with suction nozzle pipeline sealing tube, with suction nozzle and suction nozzle pipe connection, in order to prevent that the seed card from the pipeline, suction nozzle pipeline design is the rigid pipe, and the cooperation between suction nozzle pipeline sealing tube and suction nozzle and the suction nozzle pipeline is clearance fit, seals with the mode of oil blanket, makes suction nozzle and suction nozzle pipeline can reciprocate along with crank rocker mechanism.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
FIG. 1 is a schematic structural diagram of a seed sorting device based on machine vision according to an embodiment of the present invention;
FIG. 2 is a schematic view of a negative pressure device;
FIG. 3 is a schematic flow chart of a seed sorting method based on machine vision according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of the crank and rocker mechanism.
Reference numerals: 1. a support frame; 2. a seed conveying belt; 3. a negative pressure device; 4. cleaning seed boxes from bad seeds; 5. a suction nozzle conduit; 6. a suction nozzle; 7. an industrial camera; 8. a second connecting rod; 9. a first connecting rod; 10. a good seed storage box; 11. a shelf partition; 12. a crank and rocker mechanism; 13. fixing the rod; 14. a filter plate; 15. a fan; 16. a box body; 17. a top cover; 18. a motor; 19. and a fixing member.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention aims to provide a seed sorting device and method based on machine vision, which have high accuracy, do not damage the appearance shape of seeds, can realize nondestructive detection of the seeds, can remove bad seeds and ensure good quality of the seeds.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
As shown in fig. 1, a seed sorting apparatus based on machine vision according to an embodiment of the present invention includes: a support frame 1, a seed conveying belt 2, a negative pressure device 3, a good seed storage box 10, an image acquisition device, a control system and a seed suction device, wherein the seed conveying belt 2 is arranged on the support frame 1, used for driving the seed transmission, the top of the seed transmission belt 2 is provided with the seed suction device and the image acquisition device in parallel, the seed suction device is arranged at the center of the seed conveying belt 2 through the support frame 1, the image acquisition device is arranged at the front side of the seed suction device through a support frame 1, a plurality of groups of negative pressure devices 3 are arranged, the negative pressure devices 3 are connected with the seed suction device, for providing suction force to the seed suction device, the good seed storage box 10 is provided at the bottom of the end of the seed conveying belt, the control system is used for receiving the seeds and is electrically connected with the negative pressure device 3, the image acquisition device and the seed suction device;
the seed suction device comprises a first connecting rod 9, a suction nozzle 6 and a crank and rocker mechanism 12, wherein the first connecting rod 9 is fixedly arranged on the support frame 1 at two sides of the top of the seed conveying belt 2, the suction nozzle 6 and the crank and rocker mechanisms 12 are correspondingly provided with a plurality of suction nozzles 6, the crank and rocker mechanisms 12 are fixedly arranged on the first connecting rod 9, the suction nozzle 6 is fixedly arranged at the lower end of the crank and rocker mechanism 12 and is connected with the negative pressure device 3 through a suction nozzle pipeline 5 for sucking the bad seeds, the crank and rocker mechanism 12 is electrically connected with the control system, and the crank and rocker mechanism 12 is used for driving the suction nozzle 6 to move up and down;
the negative pressure device 3 is internally detachably provided with a seed cleaning box 4 for cleaning seeds, and the seed cleaning box 4 is used for storing and cleaning seeds.
The top of the seed sorting device should be provided with a suitable light source for the image acquisition device to better acquire the seed image.
The image acquisition device comprises a second connecting rod 8 and an industrial camera 7, wherein the second connecting rod 8 is fixedly arranged on the support frames 1 on the two sides of the top of the seed conveying belt 2, the industrial camera 7 is arranged at the center of the support frame 1, and a light source is fixedly arranged on the industrial camera 7 so as to facilitate the industrial camera to acquire seed images better.
The one end that the seed transmission band 2 was kept away from to good kind storage box 10 is provided with shelf baffle 11 for prevent that the seed is too fast, fall out good kind storage box 10.
The suction nozzle 6 is connected with the suction nozzle pipeline 5 through a suction nozzle pipeline sealing pipe, the suction nozzle pipeline 5 is connected with the negative pressure device 3, the suction nozzle pipeline 5 is a rigid pipeline, the suction nozzle pipeline sealing pipe is in clearance fit with the suction nozzle 6 and the suction nozzle pipeline 5 and is sealed through an oil seal, and the suction nozzle 6 and the suction nozzle pipeline 5 move up and down along with the crank rocker mechanism 12.
The support frame 1 is fixedly connected with the suction nozzle pipeline 5 through a plurality of fixing rods 13 and used for fixing the suction nozzle pipeline 5.
As shown in fig. 2, the negative pressure device 3 includes a box 16, a motor 18 and a fan 15, a top cover 17 is disposed on the top of the box 16, the motor 18 is fixedly disposed on the upper portion inside the box 16 through a fixing member 19, the fan 15 is disposed on an output shaft of the motor 18, a filter plate 14 is disposed on the bottom of the fan 15, the seed cleaning box 4 is disposed on the bottom of the filter plate 14, a plurality of pipe holes are disposed on the sidewall of the box 16 between the filter plate 14 and the seed cleaning box 4, the suction nozzle pipe communicates with the box 16 through the pipe holes, and the motor 18 is electrically connected to the control system.
The control system adopts a computer.
As shown in fig. 4, the crank and rocker mechanism may be an actuator that can move up and down as in the conventional art.
Another embodiment of the present invention is: the invention is provided with a second group of image acquisition devices and seed suction devices besides the image acquisition devices and the seed suction devices, and is used for carrying out secondary screening on the screened seeds, wherein the second image acquisition device is arranged at the rear side of the first seed suction device, the second seed suction device is arranged at the rear side of the second image acquisition device, and more image acquisition devices and seed suction devices can be added according to the quality of the seeds to ensure the screening of the seeds.
The number of industrial cameras of the image acquisition device can be selected according to the width of the seed conveying belt so as to acquire images of all seeds on the seed conveying belt.
The number of the crank rocker mechanisms and the number of the suction nozzles are also selected according to the width of the seed conveying belt.
The seed suction device can also adopt a small mechanical arm, the control system analyzes images shot by the industrial camera, coordinates of the bad seeds are obtained according to the speed of the seed conveying belt, the small mechanical arm is controlled to pick up and discard the bad seeds, and the number of the small mechanical arms is determined according to the width of the seed conveying belt and the number of the seeds.
The invention achieves the purpose of sorting out the seeds by building a conveying system platform with a conveying belt, conveying the seeds in a parallel crawler belt mode, selecting a proper light source and a camera to collect seed images, then carrying out image processing, extracting shape characteristic components and color characteristic components of the processed images, comparing the extracted shape characteristic components and color characteristic components with two components in a standard library and judging whether the extracted shape characteristic components and color characteristic components are in a qualified range.
This seed sorting device can regard the conveyer belt as a coordinate system when using, judges the coordinate of seed initial position through industry camera, according to the speed of conveyer belt and the coordinate of the initial position of seed, judges the time that the seed got into the suction nozzle absorption range, and then the time that the corresponding crank rocker mechanism of control moved down to the bad kind of absorption of being convenient for better.
The working principle of the invention is as follows: the seed is passed to the industrial camera by the conveyer belt earlier and shoots the within range, treat that the industrial camera discernment is shot the back, pass to the computer on the picture, carry out analysis processes to the picture by algorithm program, if there is bad seed, then the suction nozzle that its anterior segment is connected is driven to crank rocker mechanism moves down, the inside fan of negative pressure device lasts the rotation this moment, so continuously produce the negative pressure in the suction nozzle pipeline, the suction nozzle produces stronger suction, absorb bad seed and get into the savings box, the first half of savings box has the filter, prevent to get into the fan because of the too big seed of suction. The lower half part of the storage box is provided with a seed cleaning box for the seeds. After the seeds are fully stored, the bad seed cleaning box can be taken out, good seeds directly enter the good seed storage box along the conveying belt, and the rack partition plate is arranged at the tail end of the seed conveying belt to prevent the seeds on the conveying belt from falling out due to overlarge speed. The suction nozzle pipeline is connected with the suction nozzle pipeline through a suction nozzle pipeline sealing pipe, in order to prevent seeds from being clamped in the pipeline, the pipeline is designed to be a rigid pipe, and therefore the sealing pipe and the sealing pipe are in clearance fit with each other and are sealed in an oil seal mode, and the suction nozzle and the pipeline can move up and down along with an actuating mechanism.
The invention also provides a seed sorting method based on machine vision, which is applied to the seed sorting device based on machine vision, as shown in fig. 3, and comprises the following steps:
step 1: placing the seeds on a seed conveying belt, transmitting the seeds to the shooting range of an industrial camera, shooting the seeds by the industrial camera, and transmitting the shot images to a control system;
step 2: the control system processes the image, after the processing is finished, the shape characteristic component and the color characteristic component are extracted and compared with two components in a standard library preset by the control system, if the comparison difference value is in a qualified range, the seeds are judged to be good seeds, otherwise, the seeds are judged to be bad seeds;
and step 3: if the seeds are judged to be good seeds, the seeds are not subjected to other operations and fall into a good seed storage box through the conveying of a seed conveying belt, and if the seeds are judged to be bad seeds, the seeds are sucked into a bad seed processing box through a seed sucking device corresponding to an industrial camera for shooting images.
In step 2, the control system processes the image, after the processing is finished, the shape characteristic component and the color characteristic component are extracted and compared with two components in a standard library preset by the control system, if the comparison difference is within a qualified range, the seed is judged to be good, otherwise, the seed is judged to be bad, and the method specifically comprises the following steps:
the control system carries out smooth filtering processing on the image and is used for eliminating the influence of noise and uneven field light source irradiation on the image, and after the processing is finished, the control system extracts the shape characteristic analysis and the color characteristic component of the seeds;
the shape characteristic components comprise nine groups of data which are respectively perimeter, area, circularity, equivalent diameter, long axis, short axis, length-width ratio, compactness and matching value of the seeds, the shape characteristic components are verified by Minitab data processing software by applying a hypothesis test formula according to a standard library, the nine groups of data are judged to contain data which accord with normal distribution, a 3 delta standard interval is adopted for standard value range demarcation of the data which accord with the normal distribution, wherein the qualified range defined by the standard value is (mu-3 delta, mu +3 delta), and the qualified range is set by adopting a margin requirement of +/-20% on the data which do not accord with the normal distribution;
the color characteristic component is the average value of RGB three channels, the color characteristic component is verified through Minitab data processing software by applying a hypothesis test formula according to a standard library, whether the average value of the RGB three channels accords with normal distribution or not is judged, if the average value accords with the normal distribution, a 3 delta standard interval is adopted for standard value range demarcation, wherein the qualified range defined by the standard value is (mu-3 delta, mu +3 delta), and if the average value does not accord with the normal distribution, the qualified range is set by adopting a margin requirement of +/-20%.
The method for establishing the standard library specifically comprises the following steps:
selecting a plurality of standard seeds, extracting shape characteristic components and color characteristic components of the standard seeds, and establishing a standard database according to the shape characteristic components and the color characteristic components of the standard seeds.
In the method, the control system processes the image, extracts the shape characteristic component and the color characteristic component after the processing is finished, and further extracts each texture characteristic, specifically:
the method comprises the steps of using a function in an OpenCV library to segment sticky seeds, firstly converting an original image into a gray image, then using an OSTU algorithm to obtain a binary image of the approximate outline of the seeds, removing small black and white noise in the image by using a morphological opening operation, performing an expansion operation on the binary image after the black and white noise is removed, wherein the image of the real seeds is a subset of the expanded image, using a distanceTransform algorithm to obtain a central area of each seed, then subtracting the central area from the expanded image to obtain an uncertain seed edge area, using a connected components algorithm to create a mask image to enable the central area image of each seed to have the number of the image, using different colors to represent the number of each seed, finally using a watershed segmentation algorithm to obtain the exact position of the edge of each seed, and segmenting the single seed by the position coordinates of each seed, simultaneously removing adjacent seeds to obtain a candidate region;
after the candidate region is obtained, the control system extracts the shape feature analysis and color feature components of the seeds, specifically:
extracting shape characteristic components, performing binarization processing on each seed, extracting the ratio of the perimeter to the area of the seed image, the diameter of a circle with the same area of the region, the ellipse eccentricity, the ratio of the major axis to the minor axis, the ratio of the area of the region to the area of a boundary external frame, the ratio of the total number of pixel points filled between the region and the external frame to the total number of pixel points in the region, and the ratio of the perimeter of the region to the diameter of the circle with the same area of the region, and taking 6 characteristic parameters as characteristic vectors for classification and identification;
extracting HOG texture features, extracting features by using a HOG function in Sciki-image, graying an image, fixing the size of the image to 64 x 64, dividing the image into blocks with the size of 8 x 8, dividing each block into cells with the size of 4 x 4, wherein the cells are the most basic unit of extracting gradient information of the lowest layer, counting the gradient direction information of all pixels in the cells, dividing the gradient direction information into preset gradient direction ranges to form basic gradient histogram information, combining the gradient information of the cells at the bottom layer into block features, and using the gradient information of the blocks in the final features combined into the image for classification and identification;
extracting LBP texture features, firstly graying an image, setting a window of an LBP operator to be 4 x 4 in size, taking a window center pixel as a threshold value, comparing adjacent 8 pixel gray values, wherein the gray value is more than or equal to the center pixel point and is 1, otherwise, the gray value is 0, so that 8-bit binary numbers are obtained in each window, namely the LBP value of the window center pixel point, the value is used for representing the texture information of the area, and finally, a statistical histogram of the LBP characteristic values is used as a characteristic vector for classification and identification;
extracting texture features of GLCM (Gray-level Co-occurrrence Matrix), wherein the Gray-level Co-occurrrence Matrix is not directly used as features for distinguishing textures due to the fact that the Gray-level symbiosis Matrix is large in dimension, but some statistics constructed based on the Gray-level symbiosis Matrix are used as texture classification features, and 6 feature parameters including Angular Second Moment (ASM), contrast (contrast), correlation (correlation), entropy (energy), cooperativity (homogeneity) and dissimilarity (dissimilarity) are used as feature vectors for classification and identification;
extracting color characteristic components to graye the image, fixing the image into 64 × 64, and changing the reduced image into a one-dimensional characteristic vector for final classification recognition.
There are also texture features in the standard database that should contain standard seeds.
The embodiment provided by the invention adopts a convolutional neural network algorithm in a deep learning technology to identify and analyze the image, and specifically comprises the following steps:
establishing a simplified convolutional neural network model for a seed sorting task, firstly, verifying and comparing a seed data set by adopting networks such as AlexNet, VGG, ResNet and the like, and then visually adjusting the number of convolutional layers in the network and parameters such as the size and the number of convolutional kernels and the like on the characteristics of an intermediate layer of ResNet18 to construct a ResNet10 network, so that the parameter quantity and the calculated quantity are obviously reduced under the condition of basically keeping the identification accuracy of the original network;
aiming at seed image data with less learning categories of a convolutional neural network model, a plurality of redundant parameters still exist, a heteronuclear convolution and channel attention mechanism module is introduced to ResNet10 to construct LHCNet, firstly, the complexity of the seed image is far less than that of a natural image, therefore, 1 × 1 convolution with small size and less calculation amount is fused with 3 × 3 convolution, the calculation amount and parameter amount of the model can be effectively reduced while the capability of extracting the seed image feature by a network is not reduced, and the inference speed of image recognition is improved;
aiming at the problem of low feature utilization rate of the convolutional neural network, the structure is further improved, a double-branch dense connection convolutional neural network model is constructed and combined with the SVM, the extracted feature information is richer due to the double-branch structure, the shallow feature and the deep feature are fused by the dense connection mode, multiplexing of the deep convolutional network on the shallow feature is fully improved, and therefore the accuracy of the network for seed type identification is improved.
The device adopts a suction nozzle, a suction nozzle pipeline and a negative pressure device to achieve the purpose of sorting bad seeds in a matched suction mode, has higher accuracy rate compared with a mechanical arm sorting mode and the like, is simple and convenient to operate, can prevent seeds from being damaged, and ensures the integrity of the seeds to the maximum extent; the device is provided with a crank rocker mechanism which can drive the suction nozzle to move up and down, so that the suction nozzle can suck the bad seeds conveniently; the device's suction nozzle pipeline and the junction of suction nozzle are provided with suction nozzle pipeline sealing tube, with suction nozzle and suction nozzle pipe connection, in order to prevent that the seed card from the pipeline, suction nozzle pipeline design is the rigid pipe, and the cooperation between suction nozzle pipeline sealing tube and suction nozzle and the suction nozzle pipeline is clearance fit, seals with the mode of oil blanket, makes suction nozzle and suction nozzle pipeline can reciprocate along with crank rocker mechanism.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.
Claims (9)
1. A seed sorting device based on machine vision, comprising: a support frame, a seed conveying belt, a negative pressure device, a good seed storage box, an image acquisition device, a control system and a seed suction device, wherein the seed conveying belt is arranged on the support frame, used for driving the seed to be transmitted, the top of the seed transmission belt is provided with the seed suction device and the image acquisition device in parallel, the seed suction device is arranged at the center of the seed conveying belt through the support frame, the image acquisition device is arranged at the front side of the seed suction device through the support frame, the negative pressure devices are provided with a plurality of groups and are connected with the seed suction device, used for providing suction force for the seed suction device, the good seed storage box is arranged at the bottom of the tail end of the seed conveying belt, the control system is used for receiving the seeds and is electrically connected with the negative pressure device, the image acquisition device and the seed suction device;
the seed suction device comprises a first connecting rod, a suction nozzle and a crank and rocker mechanism, the first connecting rod is fixedly arranged on the support frames on two sides of the top of the seed conveying belt, the suction nozzle and the crank and rocker mechanism are correspondingly provided with a plurality of suction nozzles, the crank and rocker mechanisms are fixedly arranged on the first connecting rod, the suction nozzle is fixedly arranged at the lower end of the crank and rocker mechanism and is connected with the negative pressure device through a suction nozzle pipeline for sucking the bad seeds, the crank and rocker mechanism is electrically connected with the control system, and the crank and rocker mechanism is used for driving the suction nozzle to move up and down;
the negative pressure device is internally detachably provided with a seed cleaning box for cleaning seeds, and the seed cleaning box is used for storing and cleaning seeds.
2. The seed sorting device based on machine vision as claimed in claim 1, wherein the image capturing device comprises a second connecting rod and an industrial camera, the second connecting rod is fixedly arranged on the supporting frame at two sides of the top of the seed conveying belt, the industrial camera is arranged at the center of the supporting frame, and a light source is fixedly arranged on the industrial camera.
3. The machine vision based seed sorting device according to claim 1, wherein the end of the good seed storage box away from the seed conveyor is provided with a shelf partition for preventing the seeds from falling out of the good seed storage box due to too fast speed.
4. The machine vision based seed sorting device according to claim 2, wherein the suction nozzle is connected to the suction nozzle pipe through a suction nozzle pipe sealing pipe, the suction nozzle pipe is connected to the negative pressure device, the suction nozzle pipe is a rigid pipe, the suction nozzle pipe sealing pipe is in clearance fit with the suction nozzle and the suction nozzle pipe and is sealed by an oil seal, and the suction nozzle pipe move up and down along with the crank rocker mechanism.
5. The machine vision based seed sorting apparatus of claim 4, wherein the support frame is fixedly connected to the nozzle duct by a plurality of fixing rods for fixing the nozzle duct.
6. The seed sorting device based on machine vision as claimed in claim 5, wherein the negative pressure device comprises a box, a motor and a fan, the top of the box is provided with a top cover, the motor is fixedly arranged at the upper part inside the box through a fixing member, the fan is arranged on an output shaft of the motor, the bottom of the fan is provided with a filter plate, the bottom of the filter plate is provided with the seed cleaning box, a plurality of pipe holes are arranged on the side wall of the box between the filter plate and the seed cleaning box, the suction nozzle pipeline is communicated with the box through the pipe holes, and the motor is electrically connected with the control system.
7. A seed sorting method based on machine vision, which is applied to the seed sorting device based on machine vision of claims 1-6, and is characterized by comprising the following steps:
step 1: placing the seeds on a seed conveying belt, transmitting the seeds to the shooting range of an industrial camera, shooting the seeds by the industrial camera, and transmitting the shot images to a control system;
step 2: the control system processes the image, after the processing is finished, the shape characteristic component and the color characteristic component are extracted and compared with two components in a standard library preset by the control system, if the comparison difference value is in a qualified range, the seeds are judged to be good seeds, otherwise, the seeds are judged to be bad seeds;
and step 3: if the seeds are judged to be good seeds, the seeds are not subjected to other operations and fall into a good seed storage box through the conveying of a seed conveying belt, and if the seeds are judged to be bad seeds, the seeds are sucked into a bad seed processing box through a seed sucking device corresponding to an industrial camera for shooting images.
8. The method for sorting seeds based on machine vision according to claim 7, wherein in step 2, the control system processes the image, extracts the shape characteristic component and the color characteristic component after the processing, compares the extracted shape characteristic component and color characteristic component with two components in a standard library preset by the control system, if the comparison difference is within a qualified range, the seed is determined to be good, otherwise, the seed is determined to be bad, specifically:
the control system carries out smooth filtering processing on the image and is used for eliminating the influence of noise and uneven field light source irradiation on the image, and after the processing is finished, the control system extracts the shape characteristic analysis and the color characteristic component of the seeds;
the shape characteristic components comprise nine groups of data which are respectively perimeter, area, circularity, equivalent diameter, long axis, short axis, length-width ratio, compactness and matching value of the seeds, the shape characteristic components are verified by Minitab data processing software by applying a hypothesis test formula according to a standard library, the nine groups of data are judged to contain data which accord with normal distribution, a 3 delta standard interval is adopted for standard value range demarcation of the data which accord with the normal distribution, wherein the qualified range defined by the standard value is (mu-3 delta, mu +3 delta), and the qualified range is set by adopting a margin requirement of +/-20% on the data which do not accord with the normal distribution;
the color characteristic component is the average value of RGB three channels, the color characteristic component is verified through Minitab data processing software by applying a hypothesis test formula according to a standard library, whether the average value of the RGB three channels accords with normal distribution or not is judged, if the average value accords with the normal distribution, a 3 delta standard interval is adopted for standard value range demarcation, wherein the qualified range defined by the standard value is (mu-3 delta, mu +3 delta), and if the average value does not accord with the normal distribution, the qualified range is set by adopting a margin requirement of +/-20%.
9. The machine vision based seed sorting method according to claim 8, wherein the standard library is established by:
selecting a plurality of standard seeds, extracting shape characteristic components and color characteristic components of the standard seeds, and establishing a standard database according to the shape characteristic components and the color characteristic components of the standard seeds.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110933281.6A CN113458004A (en) | 2021-08-14 | 2021-08-14 | Seed sorting device and method based on machine vision |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110933281.6A CN113458004A (en) | 2021-08-14 | 2021-08-14 | Seed sorting device and method based on machine vision |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113458004A true CN113458004A (en) | 2021-10-01 |
Family
ID=77868113
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110933281.6A Pending CN113458004A (en) | 2021-08-14 | 2021-08-14 | Seed sorting device and method based on machine vision |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113458004A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114160429A (en) * | 2021-11-12 | 2022-03-11 | 湖南省水稻研究所 | Method for sorting seeds based on appearance detection result |
| CN115382776A (en) * | 2022-10-28 | 2022-11-25 | 海南联新科技有限公司 | Vision-based agricultural seed sorting device and method |
-
2021
- 2021-08-14 CN CN202110933281.6A patent/CN113458004A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114160429A (en) * | 2021-11-12 | 2022-03-11 | 湖南省水稻研究所 | Method for sorting seeds based on appearance detection result |
| CN115382776A (en) * | 2022-10-28 | 2022-11-25 | 海南联新科技有限公司 | Vision-based agricultural seed sorting device and method |
| CN115382776B (en) * | 2022-10-28 | 2022-12-27 | 海南联新科技有限公司 | Sorting method of agricultural seed sorting device based on vision |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113458004A (en) | Seed sorting device and method based on machine vision | |
| CN111275679B (en) | Image-based solar cell defect detection system and method | |
| CN107966454A (en) | A kind of end plug defect detecting device and detection method based on FPGA | |
| CN111687069A (en) | Intelligent pecan shell and kernel sorting machine based on convolutional neural network algorithm | |
| CN111783693A (en) | Intelligent identification method of fruit and vegetable picking robot | |
| CN111583193B (en) | Pistachio nut framework extraction device based on geometric contour template matching and algorithm thereof | |
| CN109886170B (en) | Intelligent detection, identification and statistics system for oncomelania | |
| CN108549910A (en) | A kind of corn seed producing fruit ear image preliminary classification method based on convolutional neural networks | |
| CN113222959B (en) | Fresh jujube wormhole detection method based on hyperspectral image convolutional neural network | |
| CN113077450B (en) | Cherry grading detection method and system based on deep convolutional neural network | |
| CN111359907B (en) | Real-time online detection system and method for wheat scab infection particle rate based on embedded deep learning | |
| CN112977974A (en) | Cigarette packet appearance quality detection device and method and cigarette packet packaging machine | |
| CN111340019A (en) | Grain bin pest detection method based on Faster R-CNN | |
| CN108664839A (en) | A kind of image processing method and equipment | |
| CN111738279A (en) | Non-contact type automatic acquisition device and method for fish morphological phenotype | |
| CN114029943A (en) | Target grabbing and positioning method and system based on image data processing | |
| CN215613293U (en) | Seed sorting device based on machine vision | |
| CN112907516B (en) | Sweet corn seed identification method and device for plug seedling | |
| CN101172274B (en) | Matrimony vine classifying and sorting device and methods thereof | |
| CN113793385A (en) | Method and device for positioning fish head and fish tail | |
| CN111611953B (en) | Target feature training-based oil pumping unit identification method and system | |
| CN113567466B (en) | Intelligent identification method for appearance defects of microchip | |
| CN111369497A (en) | Walking type tree fruit continuous counting method and device | |
| CN115995017A (en) | Fruit identification and positioning method, device and medium | |
| CN115187878A (en) | Unmanned aerial vehicle image analysis-based blade defect detection method for wind power generation device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |