Intelligent tea-leaf picker data acquisition method
Through the preliminary selection of germplasm resources and standardized breeding combined with image recognition equipment data collection, the machine picking model is trained, and the problem of insufficient data in intelligent tea picking machines when identifying and picking different tea tree varieties is solved, efficient mechanized picking of famous and high-quality teas in tea gardens is realized, and labor intensity and production costs are reduced.
Patent Information
- Application Number
- CN202510341591.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-21
- Publication Date
- 2025-07-08
AI Technical Summary
When existing intelligent tea picking machines identify and pick different tea tree varieties, the data collection volume is small and the tea tree germplasm resources are numerous, resulting in low recognition and picking efficiency, which is difficult to meet the mechanized picking needs of famous and excellent teas in tea gardens.
Through primary selection of germplasm resources, standardized short-pile cutting breeding, image recognition equipment data collection and data processing, combined with manual and mechanical picking data training, the machine picking model is gradually improved, including data preprocessing, model training and error correction, and a tea tree germplasm resource database suitable for machine picking is established.
It improves the identification and picking accuracy of intelligent tea picking machines, reduces labor intensity and production costs, and realizes efficient mechanized picking of famous and excellent teas in tea gardens.
Smart Images
Figure CN120277343A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of tea tree breeding, and in particular relates to a data collection method for an intelligent tea picking machine. Background Art
[0002] Tea is an important cash crop in my country and plays a vital role in the development of the national economy. With the development of society, tea bud picking requires a large amount of labor. Some tea-growing areas are short of tea pickers, especially the famous and high-quality tea cannot be harvested in time and can only be abandoned. Mechanized tea picking has gradually become the future development trend of tea production in my country. In particular, mechanized picking can improve the operating efficiency of tea production. The application of advanced famous and high-quality tea picking robots to replace backward manual picking or indiscriminate mechanical harvesting operations has become an inevitable trend in the harvest of famous and high-quality tea in tea gardens. It is of great significance to improve planting benefits and reduce labor intensity and production costs.
[0003] At present, mechanized tea picking has been tried in some tea-growing areas with intelligent tea picking machines for single bud picking. The key to the operation of intelligent tea picking machines is to detect tea buds through machine vision and locate the picking points before picking and grabbing. For the intelligent tea picking machines to identify tea buds and make picking movements, they need image data of the growth status of tea buds of different varieties (lines) suitable for machine picking. Only by training to identify a large number of samples of different varieties (lines) suitable for machine picking and passing trial picking training can the intelligent machine gradually establish a database of identification and picking models for intelligent tea picking machines. At present, the market is mainly concentrated in the mechanical and operating fields of intelligent tea picking machines. Due to the large number of tea germplasm resources and the short growth cycle of adult tea tree buds, the amount of data collected is small. Summary of the invention
[0004] The purpose of the present invention is to solve the deficiencies of the above-mentioned technology and to provide a data collection method for an intelligent tea picking machine that can identify and train the picking of different varieties (lines).
[0005] In order to achieve the above object, the present invention provides a data collection method for an intelligent tea picking machine, which is characterized by the following steps:
[0006] (I) Preliminary selection of germplasm resources: First, tea professionals select 30-50 representative tea germplasm resources that are initially suitable for machine harvesting from local tea areas or tea tree resource gardens, and identify varieties (lines), wild tea germplasms, etc.;
[0007] (2) Propagation of germplasm resources by short shoot cuttings: The initially selected germplasm resources are subjected to short shoot cuttings in spring, summer, and autumn. Standardized cuttings are used. The cuttings of the initially selected tea germplasm resources follow the following standards. When collecting tea branches, the thickness of tea branches of the same variety (line) is similar, with a tolerance of ±0.1 cm. The standard length of cuttings has a tolerance of ±0.5 cm. The cuttings are treated with the same rooting agent and fungicide formula. Standard plug trays (5X10) are used for cutting, and the plug trays are placed in the same environment for rooting culture. The part of the cutting exposed above the substrate is at the same horizontal plane, and the tops of the cuttings are at the same level with an error of no more than ±0.3 cm. The propagation quantity can be determined according to the specific situation, and each tea germplasm should have no less than 4 plug trays each time; Observe and record the germination status of each portion every 10 days. When 50% of the cuttings of any tea germplasm have sprouted and grown into new shoots with one bud and one leaf, remove the plug tray from the propagation area and place it on the workbench;
[0008] Step (3) Design of the data collection workbench: The data collection workbench includes an image recognition device, a data storage module, a data processing module, a mechanical picking module, and an intelligent control module. The image recognition device is connected to the data storage module through a data signal, and then connected to the data processing module, the intelligent control module, and the mechanical picking module through a signal;
[0009] Step (4): Data collection: Place the seedling-raising plug tray in step (2) on the data collection workbench for tea bud data collection. (a) Turn on the image acquisition device of the workbench, and use the horizontal, top, and oblique image recognition devices to obtain data on the agronomic traits and growth status of tea buds, and save the recognized and collected data to the storage module (S100); (b) Manually pick 50% of the tea buds on the plug cuttings of the plug tray. The speed of the manual picking process is 1 / 5 slower than the normal picking speed, and the image acquisition device collects the manual picking data throughout the process; The data collected and saved in step (b) is processed by the data processing module and then input into the intelligent control module of the data collection workbench as the manual picking control model data (S100a); (c) Turn on the mechanical picking part on the data collection workbench. Under the assistance of manual work and the control of the designed intelligent control module, the mechanical picking part on the workbench picks the remaining 50% of the tea buds on the plug cuttings. Adjust the mechanical picking speed of the machine to be 1 / 10 slower than normal to obtain the data of the machine control module for machine picking (S100b). Repeat steps (b) and (c) for training to gradually improve the picking speed and accuracy of manual and tea-picking machines;
[0010] Step (5): When the cuttings are cultured and germinated and the new shoots grow to one bud and two leaves or one bud and three leaves, start repeating the operation steps of steps (3) and (4) to obtain data S200 and S300;
[0011] Step (Six): For each of the 30 - 50 initially selected tea germplasm resources collected and bred, obtain the agronomic traits of the tea bud growth, the data of hand - picking and machine - picking tea at different bud ages (S101, S102;......; S150), (S101a, S102a;......; S150a), (S101b, S102b;......; S150b), (S201, S202;......; S250), (S201a, S202a;......; S250a), (S201b, S202b;......; S250b), (S301, S302;......; S350), (S301a, S302a;......; S350a), (S301b, S302b;......; S350b) according to the operation steps (Three) and (Four) as the data for training the picking machine control model to identify and pick.
[0012] In the present invention, through the standardized short - spike breeding of the initially selected tea germplasm in the tea area, when the cuttings germinate to the stage of one bud and one leaf, one bud and two leaves or one bud and three leaves, the short - spike breeding plug trays are placed on the data acquisition workbench, and the tea bud growth status, data such as manual and mechanical picking processes (such as picking force, angle, speed, etc.) are collected step by step through the image recognition device to obtain tea bud data, manual picking, machine - picking data, etc. The data is analyzed, processed, trained, corrected and improved; through a large amount of data collection, the data of the machine - picking tea bud model is gradually improved, and the data is analyzed and processed through the comparison algorithm, and the data is input into the tea - picking machine control module as the data for the manipulator of the tea - picking machine control module.
[0013] The specific methods for data analysis, processing, training, correction and improvement are as follows:
[0014] (1) Data pre - processing: Denoise the data collected through the image recognition device (such as tea bud color, growth posture, etc.) and operation data of manual and mechanical picking action characteristics (such as picking force, angle, speed, etc.), and analyze the above data using computer vision technology.
[0015] (2) Control model training and optimization: Input the pre - processed data into the model for training, adjust the model parameters to minimize the error, and dynamically adjust the model parameters according to the actual picking effect feedback.
[0016] (3) Data correction and control system improvement: Error analysis: Compare the effects of machine picking and manual picking, and analyze the error sources (such as recognition errors, action deviations, etc.). Record the picking failure cases (such as missed picking, mis - picking) for model correction, re - input the error data into the model for training, and correct the model parameters. Continuously collect new data and update the model to adapt to different tea garden environments and tea bud growth states.
[0017] Specific image processing and training are well-known techniques in their respective fields and will not be described in detail here.
[0018] A data acquisition method for an intelligent tea-picking machine provided by the present invention can be applied to advanced famous and high-quality tea-picking robots, thereby replacing backward manual picking or undifferentiated mechanical harvesting operations, providing technical guarantees for the harvest of famous and high-quality tea in tea gardens, and having very practical and important significance for improving planting efficiency, reducing labor intensity and production costs. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] Figure 1 It is a data acquisition platform for a data acquisition method of an intelligent tea-picking machine of the present invention;
[0020] Figure 2 It is a schematic structural diagram of the data acquisition steps of a data acquisition method of an intelligent tea-picking machine of the present invention;
[0021] Figure 3 It is a schematic structural diagram of the tea-picking data acquisition steps from one bud with one leaf to one bud with three leaves.
[0022] Among them: image recognition device 1, data storage module 2, data processing module 3, mechanical picking module 4, intelligent control module 5, agronomic and growth status acquisition data S100, manual picking control model data S100a, machine-picking machine control module data S100b, one-bud-two-leaf status acquisition data S200, one-bud-three-leaf status acquisition data S300, storage module S500. SPECIFIC EMBODIMENTS
[0023] The following describes the preferred embodiments of the present invention with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only for the purpose of illustrating and explaining the present invention, and are not intended to limit the present invention.
[0024] Embodiment 1.
[0025] In order to enable those skilled in the art to better understand the solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.
[0026] A data acquisition method for an intelligent tea-picking machine provided in this embodiment has the following steps: Step (1) Primary selection of germplasm resources: First, tea professionals initially select 30-50 representative tea germplasm resources that are initially suitable for machine picking from local tea areas or tea germplasm resource nurseries, and identify tea germplasm such as varieties, strains, and wild ones; the primary selection criteria for tea germplasm suitable for machine picking can refer to the methods of the prior art
[0027] Step (II) Short spike propagation of germplasm resources: The initially selected germplasm resources are subjected to short spike cuttings in spring, summer, and autumn. Standardized cuttings are used. The cuttings of the initially selected tea germplasm resources follow the following standards. When collecting tea branches, the thickness of tea branches of the same variety (line) is similar, with a difference of ±0.1 cm. The standard length of cuttings has a difference of ±0.5 cm. The cuttings are treated with the same rooting agent and fungicide formula. Standard plug trays (5X10) are used for cutting, and the plug trays are placed in the same environment for rooting and cultivation. The part of the cuttings exposed above the substrate is at the same horizontal plane, and the tops of the cuttings are at the same horizontal level with an error not exceeding ±0.3 cm. The propagation quantity can be determined according to the specific situation, and each tea germplasm should have no less than 4 plug trays each time; Observe and record the sprouting status of each portion every 10 days. When 50% of the cuttings of any one tea germplasm have sprouted and grown to new shoots with one bud and one leaf, take the plug tray out of the propagation area and place it on the workbench;
[0028] Step (III) Design of data collection workbench: As Figure 1 shown, the data collection workbench includes an image recognition device 1, a data storage module 2, a data processing module 3, a mechanical picking module 4, and an intelligent control module 5. The image recognition device 1 is connected to the data storage module 2 through a data signal, and then signal-connected to the data processing module 3, the intelligent control module 5, and the mechanical picking module 4;
[0029] Step (IV): Data collection: Place the seedling-raising plug tray in step (II) on the data collection workbench to collect tea bud data. (a) Turn on the image acquisition device of the workbench, and use the horizontal, top, and oblique image recognition devices to obtain data on the agronomic traits and growth status of tea buds, and save the recognized and collected data to the storage module (S100); (b) Manually pick 50% of the tea buds on the cuttings of the plug tray. The speed during the manual picking process is 1 / 5 slower than the normal picking speed, and the image acquisition device collects the manual picking data throughout the process; The data collected and saved in step (b) is processed by the data processing module and then input into the intelligent control module of the data collection workbench as data for the manual picking control model (S100a); (c) Turn on the mechanical picking part on the data collection workbench. Under the assistance of manual work and the control of the designed intelligent control module, the mechanical picking part on the workbench picks the remaining 50% of the tea buds on the cuttings of the plug tray. Adjust the mechanical picking speed of the machine to 1 / 10 slower than normal to obtain data for the machine control module of mechanical picking (S100b). Repeat steps (b) and (c) for training to gradually improve the picking speed and accuracy of manual and tea-picking machines;
[0030] Step (V): When the germinated cuttings in the cutting cultivation and breeding germplasm grow to one bud and two leaves or one bud and three leaves, start repeating the operation steps in steps (III) and (IV) to obtain data S200, S300;
[0031] Step (Six): For each of the 30 - 50 preliminarily selected tea germplasm resources collected and bred, obtain the agronomic traits of the tea bud growth, the data of different bud ages of hand - picked and machine - picked tea (S101, S102;......; S150), (S101a, S102a;......; S150a), (S101b, S102b;......; S150b), (S201, S202;......; S250), (S201a, S202a;......; S250a), (S201b, S202b;......; S250b), (S301, S302;......; S350), (S301a, S302a;......; S350a), (S301b, S302b;......; S350b) according to the operation steps (Three) and (Four) as the data for training the picking machine control model to identify and pick.
[0032] Since there are rich tea germplasm resources and a large number of varieties (lines), the number of varieties (lines) suitable for machine picking requires professional personnel to initially select suitable picking germplasm resources. Design a tea - picking machine to mechanically pick tea buds under the control of the control module. The initial data of the tea - picking machine in the control module needs to be collected. Since the growth cycle of tea buds is short and the growth characteristics of tea buds of different tea germplasm resources are different, it is necessary to pre - collect a large amount of data on the growth characteristics of tea buds and picking skill data of different germplasm resources. This embodiment relies on an image recognition device to collect the tea germplasm resources suitable for mechanical picking initially selected by professional personnel. By collecting the data on the growth characteristics of tea buds and picking skill data, and through a comparison algorithm to analyze and process the data, the data is input into the control module of the tea - picking machine as the data for the manipulator of the tea - picking machine control module.
[0033] In this embodiment, first, the tea germplasm resources suitable for machine picking initially selected by professional personnel are standardized and short - spike propagated according to the foregoing steps. When the propagated tea germplasm resources sprout to one bud and one leaf, one bud and two leaves, and one bud and three leaves, use Figure 1 the data acquisition workbench shown to collect data S100 on the horizontal level, growth status, etc. of tea buds; in the second step above, the tea buds in the plug trays are hand - picked at a speed 1 / 5 slower than the normal picking speed, and all the acquisition data S100a of the entire process of hand - picking are obtained by using an image recognition device; 50% of the remaining tea buds in the plug trays are picked by the tea - picking machine at a frequency 1 / 10 of the normal picking speed under the operation of the design module, and all the acquisition data S100b of the entire process of machine - picking are obtained by using an image recognition device.
[0034] Save the obtained data S100, S100a, and S100b to the storage module S500, as Figure 2As shown, the initial data S100, S100a, S100b (including other germplasm resources) are obtained and input into the data processing and analysis module S401 through the storage module S500. After the data is processed, it is saved to the tea picking machine storage, application, and training module S402, and the tea picking machine training continues. During this process, the non-standard parts of the tea picking machine are continuously corrected manually at S403, and the data at S402 is revised; through a large amount of data collection, training, and correction at S404, the data of the tea picking machine control module control system S405 that meets the preset conditions and completes the recognition training of tea trees and tea buds is gradually obtained, such as Figure 3 as shown.
[0035] Using the same steps, the collected and bred tea germplasm resources are used to collect data on their growth status, manual picking, and machine picking, and are classified and saved according to the names of the germplasm resources.
[0036] The tea picking machine of the present invention has been continuously improved through the identification of tea buds of different tea germplasms, manual picking, machine training, and correction, and gradually meets the preset conditions.
[0037] In summary, the tea germplasm resources that meet machine picking screened by professional technicians are used for standard short spike breeding. Through the image recognition device, data on the growth status of tea buds of tea germplasm resources, manual picking, and machine picking processes are collected. By training the machine picking and correcting the data, the data of the tea picking machine control module control system is gradually obtained.
[0038] The specific methods for the data described in this embodiment to be analyzed, processed, trained, corrected, and improved are as follows:
[0039] (1) Data preprocessing: Denoise the data collected by the image recognition device (such as the color and growth posture of tea buds) and operation data of manual picking and mechanical picking action characteristics (such as picking force, angle, speed, etc.), and analyze the above data using computer vision technology.
[0040] (2) Control model training and optimization: Input the preprocessed data into the model for training, adjust the model parameters to minimize the error, and dynamically adjust the model parameters according to the feedback of the actual picking effect.
[0041] (3) Data correction and control system improvement: Error analysis: Compare the effects of machine picking and manual picking, and analyze the sources of errors (such as recognition errors, action deviations, etc.). Record the cases of picking failures (such as missed picking, mispicking) for model correction, re-input the error data into the model for training, and correct the model parameters. Continuously collect new data and update the model to adapt to different tea garden environments and tea bud growth states.
[0042] Specific image processing and training are well-known technologies in each specialty and will not be described in detail here.
[0043] The above-described embodiments merely represent the implementation manners of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several changes and improvements can still be made, and these all fall within the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the appended claims.
[0044] Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
Claims
1. A method for collecting data of an intelligent tea picking machine Its characteristics are as follows: Step (1): Initial selection of germplasm resources: First, tea professionals initially select 30 - 50 representative tea germplasm resources, recognized varieties (lines), wild tea germplasm, etc. that are initially suitable for mechanical harvesting from local tea areas or tea germplasm nurseries; Step (2): Short - spike propagation of the initially selected germplasm resources: The initially selected germplasm resources are subjected to short - spike cutting in spring, summer, and autumn. Standardized cutting is adopted. When propagating the initially selected tea germplasm resources, the following standards are followed. When collecting tea branches, the thickness of tea branches of the same variety (line) is similar, with a tolerance of ±0.1 cm. The standard length of cuttings has a tolerance of ±0.5 cm. The cuttings are treated with the same rooting agent and fungicide formula. Standard plug trays (5X10) are used for cutting, and the cuttings are placed in the same environment for rooting culture. The part of the cutting exposed above the substrate is at the same horizontal plane, and the tops of the cuttings are within a horizontal error of no more than ±0.3 cm. The propagation quantity can be determined according to specific circumstances, with each tea germplasm having no less than 4 plug trays each time. Observe and record the germination status of each germplasm every 10 days. When 50% of the cuttings of any tea germplasm germinate and grow into new shoots with one bud and one leaf, remove the plug tray from the propagation area and place it on the workbench; Step (3): Design of the data collection workbench: The data collection workbench includes an image recognition device (1), a data storage module (2), a data processing module (3), a mechanical harvesting module (4), and an intelligent control module (5). The image recognition device (1) is connected to the data storage module (2) through a data signal, and then signal - connected to the data processing module (3), the intelligent control module (5), and the mechanical harvesting module (4); Step (4): Data collection: Place the plug trays from step (2) on the data collection workbench for tea bud data collection. (a) Turn on the image acquisition device of the workbench, and obtain data on the agronomic traits and growth status of tea buds using horizontal, top, and oblique image recognition devices, and save the recognized and collected data to the storage module (S100); (b) Manually pick 50% of the tea buds on the plug tray cuttings. The speed of the manual picking process is 1 / 5 slower than normal picking, and the image acquisition device collects the manual picking data throughout the process. The data collected and saved in step (b) is processed by the data processing module and then input into the intelligent control module of the data collection workbench as data for the manual picking control model (S100a); (c) Turn on the mechanical harvesting part on the data collection workbench. Under the assistance of manual operation and the control of the designed intelligent control module, the mechanical harvesting part on the workbench picks the remaining 50% of the tea buds on the plug tray cuttings. Adjust the mechanical harvesting speed of the machine - picked part to 1 / 10 slower than normal, and obtain data for the machine - picked machine control module (S100b). Repeat steps (b) and (c) for training to gradually improve the picking speed and accuracy of manual and tea - picking machines; Step (5): When the germinated cuttings in the cutting culture grow into new shoots with one bud and two leaves or one bud and three leaves, start repeating the operation steps of step (4) to obtain data S200 and S300; Step (VI): For each of the 30 - 50 initially selected tea germplasm resources collected and bred, use the agronomic traits of the tea bud growth of each tea germplasm resource obtained according to operation step (IV), the data of different bud ages of hand - picked and machine - picked tea (S101, S102;......; S150), (S101a, S102a;......; S150a), (S101b, S102b;......; S150b), (S201, S202;......; S250), (S201a, S202a;......; S250a), (S201b, S202b;......; S250b), (S301, S302;......; S350), (S301a, S302a;......; S350a), (S301b, S302b;......; S350b) as the data for training the picking mechanical control model to identify and pick.