Trait collection device and collection method for small-particle seeds and poplar seeds

Abstract

The invention discloses a character acquisition device for small-particle seeds, relates to the technical field of plant seed detection, and aims to solve the problems of low efficiency, strong subjectivity and unstable image quality of traditional poplar seed detection. The device comprises a seed feeding system, an image acquisition system, a vibration system, a material clearing system, a weighing system, a supporting system and the like, and all the parts cooperate to achieve the whole-process operation of seed image acquisition. The seed feeding system is linked with a camera through a direct vibration stock bin to achieve quantitative feeding. The image acquisition system utilizes backlight and surface light to alternately illuminate, acquires backlight / surface light image pairs, and comprehensively extracts the shape and color features of the grains. A flexible vibration disc of the vibration system separates the seeds in different vibration modes. The material cleaning system exhausts air through negative pressure, so that the collecting container tightly butted with the cyclone separator falls back to the weighing platform after the fan stops. According to the device, high-throughput, high-precision and manual-intervention-free image acquisition of small-particle seeds is realized, and the detection efficiency and the data consistency are improved.

Description

Technical Field

[0001] This invention relates to the field of plant seed detection technology, and in particular to a device for collecting the traits of small seeds and its application in collecting the traits of poplar seeds. Background Technology

[0002] Poplar is an important fast-growing timber species in my country, and its seed quality directly affects afforestation success. Traditional methods for poplar seed quality testing mainly rely on manual observation and measurement, which are inefficient, highly subjective, and difficult to meet the needs of large-scale seed testing. In recent years, with the rapid development of image processing technology, machine vision-based seed quality testing methods have gradually become a research hotspot.

[0003] Existing technology mentions a "multi-type small-particle seed image acquisition system based on suction cleaning," with Chinese patent publication number CN120001547B. This technical solution relies on a suction cleaning device for rapid cleaning of small-particle seeds, replacing manual cleaning and vibration cleaning, improving cleaning efficiency and reducing seed damage. Furthermore, using poplar seed image acquisition as an example, this technical solution demonstrates its high operational efficiency and applicability to small-particle seeds.

[0004] Further research into this technical solution revealed the following problems:

[0005] 1. The trait data is limited to seed images. When seed weight is required, the seed container containing the seeds must be manually removed and weighed individually. Because small seeds are very light, the manual weighing process is easily affected by external factors, resulting in inaccurate weight data.

[0006] 2. Existing image recognition algorithms face difficulties in identifying the contour, color, and texture features of small seeds such as poplar seeds. They struggle to accurately extract the characteristics of the seeds from images, and even with extensive training and data cleaning, the results still fall short of engineering application requirements. The main objective reasons are that small seeds like poplar seeds have a much lower uniformity in shape compared to common seeds, and they may possess fuzzy or semi-transparent membrane-like structures. Images acquired using existing methods are unstable and cannot meet the demands of high-throughput, high-precision non-destructive testing of seeds in practical engineering. Summary of the Invention

[0007] The technical problem to be solved by the present invention is to provide a high-throughput automated image acquisition device for poplar seeds, so as to realize the rapid, automatic and high-quality acquisition of poplar seed images and weight information.

[0008] To address the aforementioned problems in the existing technology, this invention proposes a trait collection device for small-particle seeds, comprising: a feeding system, an image acquisition system, a vibration system, and a cleaning system; all systems are connected to the same frame.

[0009] The discharge port of the feeding system is above the vibrating plate of the vibration system; the camera lens of the image acquisition system is vertically facing the surface of the vibrating plate from top to bottom; the feed port of the cleaning system is connected to the discharge port of the vibrating plate (the material tray) through a pipe; the discharge port of the cleaning system is connected to the material bottle.

[0010] In the image acquisition system, there are two light sources corresponding to the camera: a front light source and a back light source. The brightness of the front light source and the back light source can be adjusted.

[0011] The backlight is a flat panel light source that is attached to the surface of the vibratory feeder;

[0012] The positive light source is a ring light source, which is located above the vibrating plate; the ring light source is coaxial with the camera lens, and the projection of the camera lens is within the ring of light of the ring light source.

[0013] To meet the imaging requirements of dual-mode illumination (positive light + backlight) and to balance the details and contour acquisition of phenotypic detection: the distance between the ring light source and the lens surface is 40mm, and the distance between the lens surface and the flat light source surface is 210mm; the outer diameter of the camera lens is 61mm, and the inner diameter of the ring light source is 200mm.

[0014] The ring light is closely attached to the lens, and the light shines on the sample vertically downward along the lens optical axis, forming coaxial positive light. The light can evenly cover the effective field of view of the lens, avoiding light scattering and uneven brightness on the sample surface caused by the light source being too far away. At the same time, it eliminates hard shadows in the seed texture / depression, ensuring clear imaging of the phenotypic features such as seed surface texture and color.

[0015] As a backlight source, the flat panel light source needs to maintain a sufficient distance from the lens to allow the emitted light to diffuse sufficiently and form a uniform "area light source," rather than a point / localized light. A 210mm spacing allows light to penetrate the sample area perpendicularly, creating a high-contrast black-and-white outline image when photographing poplar seeds, without localized overexposure or underexposure, ensuring accurate measurement of shape parameters such as seed area and roundness. This spacing combination is a customized design for phenotypic detection of small, lightweight poplar seeds: close-range ring light preserves details, while the flat panel light preserves the outline at a distance. Furthermore, the coaxial layout of the dual light sources perfectly matches the lens optical axis, resulting in images with both clear texture features and well-defined outline boundaries, significantly improving the accuracy of segmentation and parameter calculation.

[0016] The purpose of a ring light source is to provide uniform, shadow-free illumination for the target object (such as poplar seeds). If the inner diameter of the light source is too small, it will directly block the effective shooting angle of the lens, resulting in vignetting or missing fields of view in the image, making it impossible to fully capture the phenotypic information of the seed. An inner diameter much larger than the outer diameter of the lens ensures that the lens's field of view is entirely within the illumination range of the light source. A larger inner diameter allows the light from the ring light source to be projected onto the seed surface at a more reasonable angle (especially in frontal illumination mode), reducing glare and reflection interference caused by direct light hitting the lens, and ensuring clear imaging of the seed surface texture and color characteristics.

[0017] The cleaning system includes a cyclone separator and a blower; the air inlet of the cyclone separator is the feed inlet of the cleaning system; the exhaust port at the top of the cyclone separator is connected to the air intake port of the blower; the discharge port at the bottom of the cyclone separator is below the material bottle; and the weighing platform of the weighing system is below the material bottle.

[0018] The top opening of the bottle corresponds to the cross-sectional shape of the discharge port of the cyclone separator; when the fan is working, the bottle is drawn upward due to the negative pressure inside the cyclone separator, so that the top opening of the bottle is in close contact with the discharge port of the cyclone separator; when the fan is stopped, the bottle falls onto the weighing platform.

[0019] In the practice of collecting poplar seeds, the inner diameter of the bottle opening is 47mm; a plastic connector is fitted onto the outlet of the cyclone separator, with the outer diameter of the connector being approximately 0.5mm smaller than the inner diameter of the bottle opening. This fit between the connector and the bottle opening is a clearance fit combined with a sealed contact, leaving a tiny gap to allow air to be drawn away by the fan, creating negative pressure. The plastic, relying on its rigidity and slight elasticity, adheres tightly to the bottle opening under negative pressure, achieving a seal. This design ensures sufficient adsorption force while allowing the bottle to easily detach after the machine stops.

[0020] With a bottle weight of 6.155g, the weight of poplar seeds varies by approximately 0.5 to 1g depending on the collection process;

[0021]

[0022]

[0023]

[0024]

[0025] Indicates vacuum adsorption force; The total weight of the bottle (including seeds) is 10g; K represents the safety factor, which is 1.8.

[0026] Indicates the effective adsorption area; This indicates negative pressure, which is the difference between the external atmospheric pressure and the air pressure inside the cyclone separator cavity;

[0027] Taking the LWK7575-24V fan as an example, the working air pressure is 150Pa to 400Pa, which fully meets the requirements.

[0028] Furthermore, the feeding system includes a direct vibration hopper and a first camera; the direct vibration hopper stores the seeds to be tested; the lens of the first camera is vertically oriented towards the material tray of the direct vibration hopper, and the material tray outlet is above the vibrating plate of the vibration system;

[0029] The first camera captures images of the seeds on the feed tray, and the number of seeds is identified by machine vision. Once the number of seeds reaches a preset value, a quantitative amount of seeds is dropped into the flexible vibrating plate of the vibration system through the feed tray outlet.

[0030] The linear vibrating hopper (linear vibrating feeder) is an outsourced product. It is a device that uses the principle of linear vibration to achieve uniform and continuous material conveying. It employs linear vibration, with its vibration source (piezoelectric linear vibrator or linear vibrating motor) generating high-frequency linear reciprocating motion along the axis of the material trough. The linear vibrating hopper mainly consists of a material tray, a linear vibrator, and a support / suspension structure. The material in the tray moves forward along the feed channel under the action of vibration.

[0031] Furthermore, there is a limiting structure between the bottom of the cyclone separator and the weighing platform to prevent the material bottle from shifting up or down.

[0032] The acquisition method of the acquisition device includes the following steps: (i) job preparation and (ii) image acquisition;

[0033] In step (a): take a portion of the seeds from the same batch and spread them evenly on a vibrating plate;

[0034] S1. First, use the camera of the image acquisition system to acquire seed images; then use machine vision software to measure the roundness and area of ​​each seed; then calculate the threshold for the roundness and area of ​​the seed.

[0035] a. Measure the area of ​​the seed:

[0036] The formula for calculating the pixel area A of the seed in the image is:

[0037]

[0038] Where x represents the column coordinate (horizontal direction) of the pixel.

[0039] y represents the row coordinate (vertical direction) of the pixel.

[0040] R represents the seed region (the set of all pixels belonging to the seed).

[0041] 1 indicates: the count contributed by each pixel (each pixel is counted as 1);

[0042] The formula for calculating the roundness C of a seed is:

[0043] ,

[0044] in, Indicates the area of ​​the region. Represents the area of ​​the convex hull;

[0045] (Area of ​​region) = Two-dimensional projected area of ​​the seed;

[0046] The definition of convex hull: the smallest convex polygon that can completely enclose the seed region. It can be understood as "the smallest convex contour formed by stretching the outline of the seed outward and filling all the depressions".

[0047] It is the total area covered by this smallest convex polygon, not the area of ​​the seed's "convex part";

[0048] Convex hull area acquisition: Preprocess the captured seed image (grayscale conversion, binarization, denoising) to obtain a binary image retaining only the seed region. Then, extract the edge contour point set of the seed and call a convex hull algorithm (such as Graham scan) to calculate the "minimum convex hull" on the contour point set, obtaining the vertex coordinate set of the convex hull (pixel counting method): Generate a binary image containing only the convex hull region, and count the total number of pixels within that region, which is the convex hull area. .

[0049] b. Seed area threshold: = ×a, where a represents the seed area threshold coefficient;

[0050] Seed roundness threshold: = ×c, where c represents the seed's roundness threshold coefficient;

[0051] in , The average pixel area and average roundness of the seed are represented by the fraction;

[0052] The above parameters are obtained in order to achieve machine vision feedback control of the vibratory feeder vibration.

[0053] When determining seed adhesion, if traditional machine vision methods (such as background subtraction and determining seed overlap by roundness and area thresholds) are used, and seeds are still found to be adhered, the vibratory disk will continue to vibrate until no seeds are adhered, and then image acquisition will be performed.

[0054] S2. Method for obtaining the vibration frequency of the vibratory feeder:

[0055] Multiple vibration frequency points were tested to obtain the frequency f1 for dispersing seeds, the frequency f2 for gathering seeds, and the frequency f3 for seeds moving towards the edge of the vibrating plate (discharge port).

[0056] S2. Obtain the vibration parameters of the vibratory feeder using the following method:

[0057] The vibratory feeder is vibrated by multiple vibratory motors distributed at its bottom. Through experiments, different combinations of seed state and corresponding vibratory motor working modes are obtained: combination a corresponds to dispersed seeds, combination b corresponds to gathered seeds, and combination c corresponds to seeds moving towards the discharge port at the edge of the vibratory feeder.

[0058] Taking flexible vibrating discs as an example, the vibrating motor is a voice coil motor. Common small flexible vibrating discs use four voice coil motors. Because different types of seeds exhibit different resonance characteristics, different combinations are needed for different seeds. For example:

[0059] If only the middle voice coil motor is turned on and vibrates at frequency f1, the poplar seeds will spread evenly, and this combination will be recorded as combination a; similarly, if the poplar seeds spread randomly when vibrating at frequency f2, this combination will be discarded.

[0060] When only the voice coil motors on the left and front sides are turned on, vibrating at frequencies f3 and f4 respectively, the poplar seeds move towards the discharge port on the right. This combination is recorded as combination c.

[0061] The above is for illustrative purposes only. In actual operation, the product performance of the flexible vibratory feeder, the type of seed, and the degree of seed dryness will all affect the vibration state. Therefore, it is necessary to obtain different combinations of vibratory motor operating modes through experiments before the collection operation. Since the number of voice coil motors is small and the number of operating frequency points is limited, the number of experiments can be controlled.

[0062] In step (ii), one round of image acquisition is as follows:

[0063] First, the seeds are introduced into the direct vibration hopper, the feeding system works, and the required number of seeds are fed into the flexible vibrating plate;

[0064] Then, control the vibratory feeder to operate in combination mode a; based on the images captured by the second camera, if it is determined that the seeds are separated, stop the vibration, and then capture images again;

[0065] After image acquisition is completed, the vibratory feeder is controlled to operate in combination b; after determining that the seeds have gathered based on the images acquired by the second camera, the operation is switched to combination c; after determining that the seeds have moved to the discharge port of the vibratory feeder based on the images acquired by the second camera, the vibration is stopped; the cleaning system is started to remove the seeds from the vibratory feeder.

[0066] After cleaning, turn off the blower and weigh the collected seeds;

[0067] Repeat step (ii) until the entire batch of seeds has completed the image acquisition task.

[0068] Furthermore, in step S1, the area A of the seed in the image is calculated using the AreaCenter operator of the Halcon machine vision software.

[0069] Specifically, the seeds are poplar seeds;

[0070] In step S1, the area threshold coefficient of poplar seeds is 0.8, and the roundness threshold coefficient of poplar seeds is 0.9.

[0071] The selection criteria for the area threshold coefficient of 0.8 are: it can cover most of the separation seeds, the 20% tolerance can accommodate size differences and measurement errors, exclude excessively small areas (noise, etc.), and retain most of the normal-sized separation seeds.

[0072] The selection criteria for the roundness threshold coefficient of 0.9 are as follows: the roundness of separated seeds is usually concentrated between 0.85 and 1.0, while the roundness of overlapping seeds is usually < 0.65. The coefficient of 0.9 ensures that about 95% of separated seeds are correctly identified, forming a buffer of about 0.1 with the overlap judgment threshold of 0.65.

[0073] Furthermore, in step (ii), the image acquisition system acquires images under combined front and backlight illumination conditions, and the brightness of both the front and back lights can be adjusted; the combined illumination methods include:

[0074] a. Backlighting is applied using only the backlight source to achieve a silhouette effect, increasing the brightness contrast between the seed target outline and the background; and the backlight brightness is adjusted according to the light transmittance of the seed target size.

[0075] b. Illuminate the seed target surface with only a positive light source to capture the color and texture features;

[0076] c. Backlight and front light sources are used alternately for backlighting and front lighting. The color and texture information of the seed target are extracted from the front lighting image by using the reliable contour obtained by processing the backlight illumination image.

[0077] Furthermore, in step (ii), the cleaning system is started. When the blower is working, the inner cavity of the cyclone separator is under negative pressure. The bottle is drawn upward to the outlet of the cyclone separator and held there. At this time, the negative pressure of the cyclone separator on the outlet of the vibrating plate is stronger. The seeds are sucked into the cyclone separator and fall into the bottle after separation. After that, the blower stops working, and the bottle falls freely onto the weighing platform for weighing.

[0078] The beneficial effects of this invention include:

[0079] 1) Achieve automated and precise linkage between cleaning and weighing, which is efficient and non-destructive.

[0080] By using the negative pressure suction mechanism of the cyclone separator and the fan, the seeds in the flexible vibrating plate are 100% cleaned by airflow and centrifugal force, avoiding physical damage to the seeds by mechanical contact, preserving the integrity and vigor of the seeds to the greatest extent, and ensuring the subsequent germination rate and sowing quality.

[0081] During the cleaning process, the container (material bottle) is closely connected with the cyclone separator to collect seeds through negative pressure suction. After the fan stops, the container automatically falls back to the weighing platform. Combined with a high-precision electronic scale (accuracy 0.001g), the weighing is automated without manual intervention, which reduces human error and improves the efficiency and data reliability of batch testing.

[0082] 2) Multi-mode illumination adapts to different imaging needs, significantly improving image quality and phenotypic extraction accuracy.

[0083] A ring light source (front light) and a flexible vibrating plate are integrated with a rear light source to form a combined lighting system, supporting three modes: backlight, front light, and alternating lighting with adjustable brightness ratios.

[0084] Backlight mode can generate a silhouette effect, maximizing the brightness contrast between the seed outline and the background, laying the foundation for accurate outline extraction;

[0085] Front light mode can clearly capture the color and texture features of the seed surface;

[0086] Alternating illumination mode combines the advantages of both, using the reliable contours of backlighting to accurately extract comprehensive features from front-lit images, effectively avoiding the problems of blurred contours or missing textures under a single light source, and providing high-quality image support for the accurate extraction of subsequent seed phenotypic traits such as length, width, area, and perimeter.

[0087] 3) The data acquisition process of this device can be realized through automated control, which greatly reduces reliance on manual labor and operational errors.

[0088] With the help of automated control programs and machine vision technology, it is easy to achieve fully automated linkage of the entire process, including parameter acquisition, quantitative feeding, vibration and dispersion, dispersion detection, image acquisition, automatic cleaning, and weighing statistics. For example:

[0089] Quantitative feeding is achieved by counting the number of seeds using a first camera and machine vision technology;

[0090] The seed gathering and dispersing effects of a flexible vibratory feeder can be achieved by preset combinations of vibration positions and frequencies. In automated control, these combinations can be used to control the working mode / state of the flexible vibratory feeder via combined commands.

[0091] The second camera and machine vision technology provide real-time feedback on the dispersion status of seeds on the flexible vibrating disk. Once the seeds are dispersed, an image acquisition operation is triggered.

[0092] Through automated control, the cleaning system and the weighing system work together to complete material recovery and weight detection.

[0093] During the implementation of automated control, each operational step can be equipped with log recording and anomaly handling mechanisms to ensure that the process is traceable and monitorable, which not only significantly improves the efficiency of high-throughput detection, but also ensures the consistency and accuracy of data. Attached Figure Description

[0094] Figure 1(a) Schematic diagram of the overall structure of this embodiment (including the outer frame);

[0095] Figure 1(b) is a schematic diagram of the overall structure of this embodiment (excluding the outer frame and the first camera, first viewpoint);

[0096] Figure 1(c) is a schematic diagram of the overall structure of this embodiment (a second view of Figure 1(b), excluding the material bottle);

[0097] Figure 1(d) is a schematic diagram of the overall structure of this embodiment (the third view of Figure 1(b), excluding the material clearing pipe 402);

[0098] Figure 1(e) is a schematic diagram of the overall structure of this embodiment (the fourth view of Figure 1(b), excluding the feeding system);

[0099] Figure 2 This is a schematic diagram of the direct vibration hopper of the feeding system;

[0100] Figure 3 This is a schematic diagram of the flexible vibratory feeder.

[0101] Figure 4 This is a structural diagram of the cleaning system and the weighing system;

[0102] Figure 5 This is a schematic diagram of the connection structure between the cleaning system and the weighing system;

[0103] Figure 6 This is a schematic diagram of the supporting system;

[0104] Figure 7 It is a software architecture diagram;

[0105] In the diagram: 1. Feeding mechanism; 2. Image acquisition system; 3. Vibration system; 4. Cleaning system; 5. Weighing system; 6. Support system; 7. Container.

[0106] 101, 102, and 103, for connecting the first camera;

[0107] Ring light source 201, second camera 202, support frame 203;

[0108] Built-in driver 301, material tray of flexible vibratory feeder 302, indicator light and interface of flexible vibratory feeder 303, switch of flexible vibratory feeder 304, vibration isolation fixture 305;

[0109] Cyclone separator 401, cleaning pipe 402, exhaust port at the top of the cyclone separator (for connecting the air intake port of the blower) 403, support frame 404;

[0110] Weighing table 501;

[0111] Vibration isolation plate 601, small threaded hole 602, countersunk hole 603, large threaded hole 604, stand 605, foot cup 606. Detailed Implementation The invention will be further explained below in conjunction with its specific application in poplar seeds:

[0112] This embodiment is a further optimization of the technical solution described in the "Multi-type Small Particle Seed Image Acquisition System Based on Suction Cleaning" (announcement number CN120001547B).

[0113] Referring to Figures 1(a) to 1(e), the data acquisition device of this embodiment includes:

[0114] The seed feeding system 1 includes a vibrating hopper 101 and a first camera, which is used to uniformly disperse and transport a quantitative amount of poplar seeds to the image acquisition area.

[0115] Image acquisition system 2 includes a ring light source 201, a second camera 202 and a support frame 203, used for high-resolution image acquisition of seeds;

[0116] Vibration system 3 includes a flexible vibrating plate 302 and a built-in driver 301, used to control the movement of the seed before and after image acquisition;

[0117] The cleaning system 4 includes a cyclone separator 401, a cleaning channel 402, a blower 403, and a support frame 404, which are used to remove the seeds of the acquired images.

[0118] The weighing system 5 includes a high-precision electronic scale and a weighing platform 501, used to obtain the weight of the seeds with an accuracy of 0.001g;

[0119] The support system / frame 6, including vibration damping plate 601, platform 605 and foot cup 606, is used to fix the device and reduce vibration interference.

[0120] In practice, the data acquisition method of this device can be implemented through programming. The program coordinates and controls the seed feeding system, image acquisition system, vibration system, cleaning system, and weighing system in the data acquisition device, thereby improving the automation level of poplar image acquisition and other processes.

[0121] This invention controls mechanical vibration through machine vision feedback during image acquisition. This achieves the ideal posture of small seeds, ensuring they are "separate and do not overlap," preventing seeds from mating, overlapping, or piling up, thus ensuring the complete outline of the seeds is captured.

[0122] refer to Figure 2 The seed feeding system 1 includes a direct vibration hopper 101 and a first camera. The hopper is used to store the seeds to be tested, and the industrial camera is a 20-megapixel camera used to count the number of seeds. A quantitative amount of seeds is transferred to the vibration system 3 by vibration conveying. In this example, the first camera is also equipped with a corresponding ring light source.

[0123] In the image acquisition system 2, the second camera 202 is a 65-megapixel camera. After geometric calibration, it achieves an imaging accuracy of 20μm / pixel and is fixed on the support frame 203.

[0124] The ring light source 201 provides uniform illumination, while the flexible vibrating disk 302 integrates a backlight source for coordinated use, supporting image acquisition under combined front and backlight illumination conditions. Both the front and back lights have adjustable brightness levels. By adjusting the brightness ratio of the front and back lights, various imaging effects can be achieved under different illumination modes to suit different target characteristics. This includes:

[0125] Backlighting is achieved using only a flexible vibrating disk 302 integrated backlight source to create a silhouette effect, maximizing the brightness contrast between the target outline and the background to improve outline extraction accuracy. The backlight brightness is adjusted according to the target size and material transmittance. Frontlighting is achieved using only a ring light source 201 to collect the color and texture features of the target surface. Alternating backlighting and frontlighting are achieved by using the flexible vibrating disk 302 integrated backlight source and the ring light source 201. The color and texture information extracted from the frontlighting image is obtained by processing the reliable outline obtained from the backlighting image to acquire the comprehensive features of the target. This eliminates the insufficiency of color and texture features under single backlighting conditions and the inadequacy of outline information extraction due to shadows around the target under single frontlighting conditions.

[0126] refer to Figure 3 In the vibration system 3, the vibration isolation fixture 305 (such as rubber shock-absorbing pads) is located below the flexible vibrating plate 302 to reduce vibration transmission; the flexible vibrating plate 302 controls the seeds to gather, disperse, and move towards the edge according to the received commands, so that the seeds achieve the effect of "separating each seed from its boundary". With the help of the image acquisition system to acquire images, multiple tests have proven that the effect of "separating each seed from its boundary" can be achieved for about 120 seeds with a few vibrations. Subsequently, the cleaning system 4 is used to complete the seed removal. Multiple tests have proven that the cleaning rate can reach 100%.

[0127] refer to Figure 4 In the cleaning system 4, the cyclone separator 401 and the blower separate seeds through airflow and centrifugal force, and the weighing platform 501 below is equipped with a container 7 to collect the seeds. When the blower is running, the seeds in the flexible vibrating plate 302 are removed by negative pressure suction. At the same time, the container bottle is also sucked away from the weighing platform 501 and tightly connected to the cyclone separator 401. The seeds fall into the container. When the blower stops, the container falls back to the weighing platform 501, and the weighing is completed by a high-precision electronic scale. The support frame 404 is used to adjust the height of the cyclone separator 401.

[0128] In the weighing system 5, the high-precision electronic scale and the weighing platform 501 work together with the material cleaning system 4 to complete the weighing.

[0129] refer to Figure 6 In the support system (frame) 6, the vibration isolation plate 601 is connected to each subsystem through the small threaded hole 602, the countersunk hole 603 and the large threaded hole 604; the platform 605 and the foot cup 606 provide stable support and adjust the height of the device.

[0130] In this invention, the seed roundness and area thresholds are obtained using machine vision methods to acquire the seed roundness and area, and then the thresholds are calculated. These parameters are used to determine the effectiveness of seed dispersion.

[0131] The feeding system uses machine vision to sense the number of seeds and controls the vibration of the direct vibration hopper 101 to achieve quantitative feeding and prevent the "separation of seeds" effect from being impossible due to too many seeds in the flexible vibrating plate 302.

[0132] Compared to some cleaning devices that may cause mechanical damage to seeds, the cyclone separator 401 mainly relies on airflow and centrifugal force for separation, resulting in less physical damage to the seeds. It can better maintain the integrity and viability of the seeds, which is beneficial to improving the germination rate and sowing quality.

[0133] After the cleaning system 4 performs the cleaning operation, the weighing system 5 weighs the seeds collected in the corresponding batch, and the weight is compared with the initial weight to calculate the loss rate.

[0134] The apparatus in this example is further described below:

[0135] In seed feeding system 1: First, seeds are placed into the hopper. The vibrating hopper 101 is then activated, causing the seeds to move within the hopper through vibration. A first camera is positioned above the discharge port at the end of the hopper's tray. Machine vision technology is used to count the total number of seeds discharged. When the number of discharged seeds reaches a preset value, the vibrating hopper 101 stops vibrating and discharging, thus achieving quantitative feeding. At this point, feeding system 1 evenly disperses a quantitative amount of poplar seeds and delivers them to the tray 302 of the flexible vibrating feeder.

[0136] In image acquisition system 2: the ring light source 201 is easy to install and provides a uniform and stable lighting environment to ensure image quality. The second camera 202 is a 6500W pixel camera, ensuring image quality and achieving an imaging accuracy of 20µm / pixel. The support frame 203 supports the entire structure. The vibration system 3 is placed on the vibration isolation plate 601 to reduce vibration interference, enhance system stability, and improve image quality.

[0137] The flexible vibratory feeder of the vibration system 3 includes a built-in vibration driver 301, a feed tray 302, an indicator light and interface 303, a switch 304, and a vibration isolation fixture 305. The built-in driver 301 causes the flexible vibratory feeder to vibrate. The feed tray 302 is used to load seeds. The indicator light and interface 303 displays the working status of the flexible vibratory feeder and allows connection to other machines. The switch 304 turns the flexible vibratory feeder on and off. The vibration isolation fixture 305 is located below the entire flexible vibratory feeder to reduce vibration interference and enhance system stability.

[0138] In addition, the flexible vibrating plate has a backlight, which, together with the ring light 201, can acquire images of the seed when using front light and backlight respectively. Different effects can also be obtained by adjusting the ratio of front light and backlight.

[0139] After the seeds are fed from the feeding mechanism to the material tray 302 of the flexible vibratory feeder, the built-in driver 301 connected to the flexible vibratory feeder starts working according to the vibration command until the seeds are dispersed.

[0140] Then, the camera is used to acquire images. After the image acquisition is completed, the built-in driver 301 starts working according to another vibration command, controlling the seeds to gather and finally move to the edge of the flexible vibrating plate to reach the discharge port, so that the cleaning system 4 can complete the cleaning.

[0141] In the cleaning system 4, the cyclone separator 401, compared to some cleaning equipment that may cause mechanical damage to seeds, mainly relies on airflow and centrifugal force for separation, resulting in less physical damage to the seeds. It better preserves the integrity and vigor of the seeds, which is beneficial for improving seed germination rate and sowing quality. A container (material bottle) 7 is placed below the cyclone separator to collect the separated seeds. The seeds are discharged from the flexible vibrating plate through the cleaning pipe 402. When the blower is running, the seeds in the material tray 302 of the flexible vibrating plate are removed by negative pressure suction. At the same time, the container is also sucked away from the weighing platform 501 and tightly connected to the cyclone separator 401, with the seeds falling into the container 7. When the blower stops, the container falls back to the weighing platform 501. Cleaning is complete.

[0142] The weighing system 5 includes a high-precision electronic scale and a weighing platform 501, which is used to obtain the weight of the seeds with an accuracy of 0.001g. By comparing the seed weight before and after collection with the number of seeds collected, the thousand-seed weight of the seeds can be obtained.

[0143] In support system 6, the seed feeding system 1, image acquisition system 2, vibration system 3, cleaning system 4, and weighing system 5 are all placed on vibration isolation plate 601 to reduce vibration interference and enhance system stability. Small threaded hole 602 is used to connect and fix different camera brackets, countersunk hole 603 is used to connect and fix vibration isolation plate 601 to platform 605, large threaded hole 604 is used to connect and fix vibration system 3, and platform 605 and foot cup 606 are used to support the entire device.

[0144] When using,

[0145] First, a suitable amount of seeds from a batch are laid flat on a flexible vibrating plate, and the roundness and area of ​​each seed are measured. Then, the threshold values ​​for roundness and area are calculated.

[0146] Next, for the same type of seed, experiments were conducted at multiple vibration frequency points to obtain satisfactory vibration results.

[0147] After obtaining the above parameters during the job preparation process, image acquisition can begin in batches:

[0148] First, the seeds are introduced into the direct vibration hopper, the quantitative feeding system works, and the required number of seeds enter the flexible vibrating plate;

[0149] Then, the flexible vibrating plate operates, and the seeds vibrate; the images captured by the second camera determine whether the seeds are separated. If the seeds are separated, the vibration stops, and images are captured under three combined lighting modes; after completion, the material is removed.

[0150] After the material removal is completed, the next round of image acquisition will begin: the quantitative feeding system will feed a new batch of seeds into the flexible vibrating plate, and the above process will be repeated until the entire batch of seeds has completed the image acquisition task.

[0151] Based on the images acquired by this device, the visual detection results can determine the number of seeds, and combined with the seed weighing data, the thousand-seed weight information can be obtained.

Claims

1. A device for collecting traits of small seeds, comprising: The system includes a feeding system, an image acquisition system, a vibration system, and a cleaning system. All systems are connected to the same rack; The discharge port of the feeding system is above the vibrating plate of the vibration system; the camera lens of the image acquisition system is vertically facing the surface of the vibrating plate from top to bottom; the feed port of the cleaning system is connected to the discharge port of the vibrating plate through a pipe; the discharge port of the cleaning system is connected to the material bottle. Its characteristic is that in the image acquisition system, there are two light sources corresponding to the camera, namely a front light source and a back light source, and the brightness of the front light source and the back light source can be adjusted respectively; The backlight is a flat panel light source that is attached to the surface of the vibratory feeder; The positive light source is a ring light source, which is located above the vibrating plate; the ring light source is coaxial with the camera lens, and the projection of the camera lens is within the ring of light of the ring light source. The cleaning system includes a cyclone separator and a blower; the air inlet of the cyclone separator is the feed inlet of the cleaning system; the exhaust port at the top of the cyclone separator is connected to the air intake port of the blower; the discharge port at the bottom of the cyclone separator is below the material bottle; and the weighing platform of the weighing system is below the material bottle. The top opening of the bottle corresponds to the cross-sectional shape of the discharge port of the cyclone separator; when the fan is working, the bottle is drawn upward due to the negative pressure inside the cyclone separator, so that the top opening of the bottle is in close contact with the discharge port of the cyclone separator; when the fan is stopped, the bottle falls onto the weighing platform.

2. The small seed trait collection device according to claim 1, characterized in that: The feeding system includes a vertical vibration hopper and a first camera; the vertical vibration hopper stores the seeds to be tested; the lens of the first camera is vertically facing the material tray of the vertical vibration hopper, and the material tray outlet is above the vibrating plate of the vibration system; The first camera captures images of the seeds on the feed tray, and the number of seeds is identified by machine vision. Once the number of seeds reaches a preset value, a quantitative amount of seeds is dropped into the flexible vibrating plate of the vibration system through the feed tray outlet.

3. The small seed trait collection device according to claim 1, characterized in that: There is a limiting structure between the bottom of the cyclone separator and the weighing platform to prevent the material bottle from moving up or down.

4. The small seed trait collection device according to claim 2, characterized in that: The acquisition method of the acquisition device includes the following steps: (i) job preparation and (ii) image acquisition; In step (a): take a portion of the seeds from the same batch and spread them evenly on a vibrating plate; S1. First, use the camera of the image acquisition system to acquire seed images; then use machine vision software to measure the roundness and area of ​​each seed; then calculate the threshold for the roundness and area of ​​the seed. a. Measure the area of ​​the seed: The formula for calculating the pixel area A of the seed in the image is: ， Where x represents the column coordinate of the pixel. y represents the row coordinate of the pixel. R represents the seed region. 1 indicates: the count contributed by each pixel, with each pixel counted as 1; The formula for calculating the roundness C of a seed is: ， in, Indicates the area of ​​the region. Represents the area of ​​the convex hull; b. Seed area threshold: = ×a, where a represents the seed area threshold coefficient; Seed roundness threshold: = ×c, where c represents the seed's roundness threshold coefficient; in , These represent the average pixel area of ​​the seed and the average roundness of the seed, respectively. S2. Obtain the vibration parameters of the vibratory feeder using the following method: The vibratory feeder is vibrated by multiple vibratory motors distributed at its bottom. Through experiments, different combinations of seed state and corresponding vibratory motor working modes were obtained: combination a corresponds to dispersed seeds, combination b corresponds to gathered seeds, and combination c corresponds to seeds moving towards the discharge port at the edge of the vibratory feeder. In step (ii), one round of image acquisition is as follows: First, the seeds are introduced into the direct vibration hopper, the feeding system works, and the required number of seeds are fed into the flexible vibrating plate; Then, control the vibratory feeder to operate in combination mode a; based on the images captured by the second camera, if it is determined that the seeds are separated, stop the vibration, and then capture images again; After image acquisition is completed, the vibratory feeder is controlled to operate in combination b; after determining that the seeds have gathered based on the images acquired by the second camera, the operation is switched to combination c; after determining that the seeds have moved to the discharge port of the vibratory feeder based on the images acquired by the second camera, the vibration is stopped; the cleaning system is started to remove the seeds from the vibratory feeder. After cleaning, turn off the blower and weigh the collected seeds; Repeat step (ii) until the entire batch of seeds has completed the image acquisition task.

5. The small seed morphology collection device according to claim 4, characterized in that, in step S1, the AreaCenter operator of Halcon machine vision software is used to calculate the pixel area A of the seed in the image.

6. The small seed trait collection device according to claim 4, characterized in that... The seeds are poplar seeds; In step S1, the area threshold coefficient of poplar seeds is 0.8, and the roundness threshold coefficient of poplar seeds is 0.

9.

7. The small seed trait collection device according to claim 4, characterized in that: In step (ii), the image acquisition system acquires images under combined front and backlight illumination, and the brightness of both the front and back lights can be adjusted; the combined illumination methods include: a. Backlighting is applied using only the backlight source to achieve a silhouette effect and increase the brightness contrast between the seed target outline and the background; the brightness of the backlight source is adjusted according to the size and transmittance of the seed target. b. Illuminate the seed target surface with only a positive light source to capture the color and texture features; c. Backlight and front light sources are used alternately for backlighting and front lighting. The color and texture information of the seed target are extracted from the front lighting image by using the reliable contour obtained by processing the backlight illumination image.

8. The small seed trait collection device according to claim 1, characterized in that: In step (ii), the cleaning system is started. When the blower is working, the inner cavity of the cyclone separator is under negative pressure. The bottle is drawn upward to the outlet of the cyclone separator and held there. At this time, the negative pressure of the cyclone separator on the outlet of the vibrating plate is stronger. The seeds are sucked into the cyclone separator and fall into the bottle after separation. After that, the blower stops working and the bottle falls freely onto the weighing platform for weighing.

9. The small-particle seed trait collection device according to claim 1, characterized in that: Image acquisition system: The distance between the ring light source and the camera lens surface is 40mm, and the distance between the camera lens surface and the flat light source surface is 210mm. The outer diameter of the camera lens is 61 mm, and the inner diameter of the ring light source is 200 mm.

10. The small-particle seed trait collection device according to claim 1, characterized in that the material... The inner diameter of the bottle opening is 47mm; a plastic connector is fitted onto the outlet of the cyclone separator, and the outer diameter of the connector is 0.5mm smaller than the inner diameter of the bottle opening.

Images