Breeding method and cultivation method of colorful rice
By using hybridization breeding and artificial intelligence image recognition technology, five types of colored grain populations were constructed and automatically screened, solving the problem of insufficient accuracy in traditional artificial breeding and realizing efficient and accurate breeding of five-colored rice varieties.
Patent Information
- Application Number
- CN202511274540.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-08
- Publication Date
- 2025-11-11
AI Technical Summary
In the process of breeding five-colored rice, traditional methods rely on manual observation, which makes it difficult to unify the breeding standards, lacks accuracy, and is greatly affected by subjective factors.
A rice panicle population containing five colors of grains—black, brown, red, yellow, and white—was constructed using hybridization breeding. Automated separation and screening were performed using artificial intelligence image recognition technology. RGB thresholding and convolutional neural networks were used to identify the five colors of grains in rice panicle images, and multi-generational collaborative breeding was carried out in conjunction with agronomic traits.
This improved the accuracy and efficiency of breeding, reduced subjective errors, and enabled efficient and accurate breeding of multicolored rice varieties.
Smart Images
Figure CN120918097A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of rice breeding technology, and in particular to a method for breeding and cultivating five-colored rice per ear. Background Technology
[0002] Five-colored rice, a highly distinctive rice variety, is characterized by a gradient distribution of black, red, yellow, green, and purple colors on a single ear of rice, giving it exceptional ornamental value. Furthermore, five-colored rice is rich in various nutrients beneficial to the human body, such as amino acids, minerals (calcium, iron, zinc, etc.), and vitamins. Compared to ordinary rice varieties, its nutritional value is significantly higher, making it a promising candidate for the health food market.
[0003] Currently, the breeding of five-colored rice ears faces many challenges. Traditional breeding methods mainly rely on manual observation and screening. However, due to the complex and diverse colors and irregular distribution of five-colored rice ears, manual judgment is easily affected by subjective factors, making it difficult to unify breeding standards and resulting in insufficient accuracy in breeding. Summary of the Invention
[0004] The purpose of this invention is to provide a method for breeding and cultivating a single-ear five-colored rice variety, which can improve the accuracy of breeding a single-ear five-colored rice variety.
[0005] To achieve the above objectives, in a first aspect, the present invention provides a method for breeding five-colored rice in a single ear, comprising:
[0006] A rice panicle population containing five colors of grains—black, brown, red, yellow, and white—was constructed through hybridization breeding.
[0007] Artificial intelligence image recognition technology is used to automatically separate and screen seeds of five colors to obtain individual plants that meet the preset color ratio;
[0008] By combining agronomic traits, multi-generational synergistic breeding of individual plants was carried out until a five-colored rice variety that meets the standards for color stability and agronomic traits was obtained.
[0009] The specific steps for constructing a rice panicle population containing five types of grains—black, brown, red, yellow, and white—through hybridization breeding include:
[0010] Select indica white rice varieties or indica colored rice varieties as the female parent and japonica colored rice varieties as the male parent to introduce the target color gene;
[0011] Within 24 hours of emasculating the female parent, pollinate the male parent and harvest F0 generation hybrid seeds;
[0012] F0 generation seeds were planted to obtain F1 generation plants. False hybrids were eliminated and single plants with obvious color separation were retained as F2 generation parents.
[0013] F1 generation plants were planted to obtain F2 generation populations, and at least 200 plants were randomly selected as screening subjects.
[0014] The specific steps for using artificial intelligence image recognition technology to automatically separate and screen five types of colored seeds to obtain individual plants that meet the preset color ratios include:
[0015] Acquire images of rice ears;
[0016] The collected rice ear images were adjusted to a uniform size, and then median filtering was used for noise reduction and rice ear region segmentation.
[0017] Based on the characteristics of color space, an association model is used to identify five types of colored grains in rice ear images;
[0018] The output contains individual plants with five different colored seeds in a preset ratio.
[0019] The specific steps for identifying five color categories of grains in a rice ear image using an association model based on color space characteristics include:
[0020] A fast classification model based on RGB thresholds performs initial screening of five types of colored seeds by pre-setting the RGB value range of five types of colored seeds.
[0021] A high-precision classification model based on convolutional neural networks is used to extract deep features from images through the ResNet-50 architecture for fine screening of five types of colored seeds.
[0022] Among them, the high-precision classification model based on convolutional neural networks uses the ResNet-50 architecture to extract deep features from the image for fine screening of five types of color seeds.
[0023] First, data augmentation operations are performed on the original rice ear image, including rotation, scaling, and brightness adjustment.
[0024] Among these steps, the multi-generational synergistic selection of individual plants based on agronomic traits is included.
[0025] Based on the goal of high-yield breeding, the screening thresholds were set as plant height 80-100cm, ear length 18-22cm, and tiller number 15-20.
[0026] Secondly, the present invention also provides a cultivation method for a single-ear five-colored rice variety, wherein the five-colored rice variety bred using the aforementioned single-ear five-colored rice breeding method is characterized by comprising:
[0027] Select plots of land that meet specific soil conditions and carry out land preparation;
[0028] The seeds of the five-colored rice variety undergo sun drying, seed selection, disinfection, and germination treatment.
[0029] Sowing and seedling cultivation of seeds;
[0030] Transplanting should be carried out while controlling the specifications and quality of the seeds.
[0031] Fertilize and irrigate the seeds.
[0032] This invention discloses a breeding and cultivation method for a single-ear five-colored rice variety. It combines artificial intelligence image recognition technology to achieve automated separation and screening of five types of colored grains. Compared with manual observation, artificial intelligence image recognition has significant advantages such as objectivity, efficiency, and accuracy. It is not affected by subjective factors and can accurately analyze and judge images according to preset standards. Moreover, it can process large amounts of image data in a short time, greatly improving recognition efficiency and providing scientific decision-making basis for breeding experts. This improves the breeding efficiency and accuracy of single-ear five-colored rice and accelerates the breeding process of superior varieties. Attached Figure Description
[0033] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.
[0034] Figure 1 This is a flowchart of the breeding method for a single ear of five-colored rice according to the present invention.
[0035] Figure 2 This is a flowchart of the present invention for constructing a rice panicle population containing five types of grains: black, brown, red, yellow, and white, through hybridization breeding.
[0036] Figure 3 This invention utilizes artificial intelligence image recognition technology to automatically separate and screen five types of colored seeds, obtaining individual plants that meet a preset color ratio.
[0037] Figure 4 This is a flowchart of the present invention, which uses an association model to identify five types of colored grains in a rice ear image based on color space characteristics.
[0038] Figure 5 This is a flowchart of the cultivation method of the five-colored rice of the present invention. Detailed Implementation
[0039] The embodiments of the present invention are described in detail below. Examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, but should not be construed as limiting the present invention.
[0040] Firstly, please refer to Figures 1-4 ,in, Figure 1 This is a flowchart of the breeding method for a single ear of five-colored rice according to the present invention; Figure 2This is a flowchart of the present invention for constructing a rice panicle population containing five types of grains: black, brown, red, yellow, and white, through hybridization breeding; Figure 3 This invention utilizes artificial intelligence image recognition technology to automatically separate and screen five types of colored seeds to obtain individual plants that meet a preset color ratio. Figure 4 This is a flowchart of the present invention, which uses an association model to identify five types of colored grains in a rice ear image based on color space characteristics.
[0041] This invention provides a method for breeding five-colored rice in a single ear, comprising:
[0042] S1 uses hybridization breeding to construct a rice panicle population containing five types of grains: black, brown, red, yellow, and white.
[0043] The specific steps include:
[0044] S11 selects either an indica white rice variety or an indica colored rice variety as the female parent and a japonica colored rice variety as the male parent to introduce the target color gene.
[0045] S12 couples were pollinated with pollen from the male parent within 24 hours after the female parent was emasculated, and F0 generation hybrid seeds were harvested.
[0046] S13 plants were planted with F0 generation seeds to obtain F1 generation plants. False hybrids were eliminated and single plants with obvious color separation were retained as F2 generation parents.
[0047] S14 plants were planted to obtain F2 generation populations from F1 generation plants, and at least 200 plants were randomly selected as screening subjects.
[0048] In this embodiment of the invention, the goal is to obtain a five-colored rice variety with stable color and excellent agronomic traits. A rice panicle population containing five colors of grains—black, brown, red, yellow, and white—is constructed through hybridization breeding.
[0049] S2 utilizes artificial intelligence image recognition technology to automatically separate and screen five types of colored seeds, obtaining individual plants that meet the preset color ratio;
[0050] The specific steps include:
[0051] S21 captures images of rice ears;
[0052] In this embodiment of the invention, to eliminate ambient light interference, images of rice ears are acquired under a light intensity of 500-1000 lux and a solid-color non-reflective background.
[0053] S22 adjusts the acquired rice ear images to a uniform size, performs median filtering for noise reduction, and segments the rice ear regions;
[0054] In this embodiment of the invention, in order to unify the image input scale, the acquired image is resized to 1024×768 pixels, denoised by median filtering, and segmented into rice ear regions.
[0055] Based on the characteristics of the color space, S23 uses an association model to identify five types of colored grains in rice ear images;
[0056] The specific steps include:
[0057] S231 is a fast classification model based on RGB thresholds, which performs initial screening of five types of colored seeds by pre-setting the RGB value range of five types of colored seeds.
[0058] In this embodiment of the invention, the RGB value ranges of the five types of colored seeds were determined through the following correlation experiments:
[0059] Black: R = 26-31, G = 27-32, B = 28-33 (based on the spectral characteristics of melanin deposition);
[0060] Red: R = 200-230, G = 50-80, B = 50-80 (based on the absorption peak wavelength of anthocyanins);
[0061] The RGB value ranges for brown, yellow, and white were determined through experiments using standard color charts and entered into the breeding database.
[0062] S232 is a high-precision classification model based on convolutional neural networks. It uses the ResNet-50 architecture to extract deep features of images for fine screening of five types of colored seeds.
[0063] In this step, the original rice ear image is first subjected to data augmentation operations, including rotation, scaling, and brightness adjustment.
[0064] Specifically, the high-precision classification model of the convolutional neural network adopts the ResNet50 convolutional neural network model, and optimizes the recognition accuracy through the following associated training strategies: To solve the problem of imbalanced samples, data augmentation is performed on the original image by rotation (±15°), scaling (90%-110%), and brightness adjustment (-20% to +20%); To accelerate model convergence, ResNet-50 weights pre-trained on the ImageNet dataset are loaded, the parameters of the first 10 layers are frozen, and the remaining layers are fine-tuned; To improve the small sample class recognition rate, Focal Loss is used instead of traditional cross-entropy loss, and the weights of each class are dynamically adjusted.
[0065] The S24 output includes individual plants containing seeds of five different colors in a preset ratio.
[0066] S3 combined agronomic traits with multi-generational synergistic breeding of individual plants until a five-colored rice variety that meets the standards for color stability and agronomic traits was obtained.
[0067] Specifically, in the step of multi-generational collaborative breeding of single plants in combination with agronomic traits, the following correlation criteria are used to achieve simultaneous optimization of multiple traits: based on the goal of high-yield breeding, screening thresholds are set for plant height of 80-100cm, ear length of 18-22cm, and number of tillers of 15-20; to maintain the genetic stability of color, F3-F6 generations are planted continuously, and each generation is planted in a zone according to color line, with 30 plants per plot; to improve the breeding efficiency, the AI screening step described in claim 3 is repeated in each generation of planting to achieve a dynamic balance between color and agronomic traits.
[0068] Color stability was evaluated using the following correlation calculation method:
[0069] To quantify color ratio fluctuations, the coefficient of variation (CV) is calculated:
[0070]
[0071] Where σ is the standard deviation of color ratio and μ is the mean; based on long-term breeding experience, CV≤10% in continuous planting for 3-5 years is set as the stability criterion.
[0072] The breeding method for five-colored rice with a single ear of this invention combines artificial intelligence image recognition technology to achieve automated separation and screening of five types of colored grains. Compared with manual observation, artificial intelligence image recognition has significant advantages such as objectivity, efficiency, and accuracy. It is not affected by subjective factors and can accurately analyze and judge images according to preset standards. Moreover, it can process a large amount of image data in a short time, greatly improving recognition efficiency and providing scientific decision-making basis for breeding experts. This improves the breeding efficiency and accuracy of five-colored rice with a single ear and accelerates the breeding process of superior varieties of five-colored rice with a single ear.
[0073] Secondly, please refer to Figure 5 , Figure 5 This is a flowchart of the cultivation method of the five-colored rice of the present invention.
[0074] This invention also provides a method for cultivating five-colored rice per ear, comprising:
[0075] S100 selects plots of land that meet specific soil conditions and performs land preparation.
[0076] In this embodiment of the invention, fallow fields, rapeseed stubble fields, or early wheat stubble fields are preferred. The soil pH should be 5.5-7.0, organic matter content ≥2.5%, and drainage slope 1:500-1:1000. Land preparation employs a "two-rotary-one-harrow" process: the first rotary tillage depth is 12-15cm to crush straw residue; the second rotary tillage depth is 8-10cm, combined with the application of base fertilizer; finally, the field surface is harrowed to a level surface with a height difference ≤3cm.
[0077] S200 involves sun-drying, selecting, disinfecting, and germinating the seeds of the five-colored rice variety.
[0078] In this embodiment of the invention, the seeds are spread out to dry for 1-2 days 7 days before sowing to reduce the moisture content to 12%-14%; empty and shriveled seeds are removed by flotation with 1.13 specific gravity salt water; seeds are disinfected by soaking in 1% lime water for 15-24 hours at a water temperature of 15-20℃; and sowing is carried out when the buds are broken at a constant temperature of 30℃ and the buds are 1mm long.
[0079] S300 is used for sowing and seedling cultivation;
[0080] In this embodiment of the invention, sowing is carried out when the daily average temperature is consistently above 10℃, with 5-8 kg of well-rotted farmyard manure applied per square meter of seedbed, and a sowing rate of 60-80 g / m². 2 .
[0081] S400 controls the transplanting specifications and quality of seeds during transplanting;
[0082] In this embodiment of the invention, the transplanting specifications are as follows: row spacing 25cm, plant spacing 16cm, 200,000 holes per hectare; 10 seedlings per hole, seedling age controlled at 30-35 days. Transplanting quality: shallow planting: 2-3cm depth into the mud; uniform planting: missed planting rate ≤5%, floating seedling rate ≤3%.
[0083] S500 fertilizes and irrigates the seeds.
[0084] The present invention provides a cultivation method for a single-ear five-colored rice, which enables efficient cultivation of the single-ear five-colored rice and improves its quality.
[0085] The above-disclosed embodiments are merely one or more preferred embodiments of this application and should not be construed as limiting the scope of this application. Those skilled in the art can understand that all or part of the processes for implementing the above embodiments and equivalent changes made in accordance with the claims of this application still fall within the scope of this application.
Claims
1. A method for breeding a variety of five-colored rice per ear, characterized in that, include: A rice panicle population containing five colors of grains—black, brown, red, yellow, and white—was constructed through hybridization breeding. Artificial intelligence image recognition technology is used to automatically separate and screen seeds of five colors to obtain individual plants that meet the preset color ratio; By combining agronomic traits, multi-generational synergistic breeding of individual plants was carried out until a five-colored rice variety that meets the standards for color stability and agronomic traits was obtained.
2. The breeding method for five-colored rice per ear as described in claim 1, characterized in that, The specific steps for constructing a rice panicle population containing five types of grains—black, brown, red, yellow, and white—through hybridization breeding include: Select indica white rice varieties or indica colored rice varieties as the female parent and japonica colored rice varieties as the male parent to introduce the target color gene; Within 24 hours of emasculating the female parent, pollinate the male parent and harvest F0 generation hybrid seeds; F0 generation seeds were planted to obtain F1 generation plants. False hybrids were eliminated and single plants with obvious color separation were retained as F2 generation parents. F1 generation plants were planted to obtain F2 generation populations, and at least 200 plants were randomly selected as screening subjects.
3. The breeding method for five-colored rice with a single ear as described in claim 2, characterized in that, The specific steps for automating the separation and screening of five types of colored seeds using artificial intelligence image recognition technology to obtain individual plants that meet the preset color ratios include: Acquire images of rice ears; The collected rice ear images were adjusted to a uniform size, and then median filtering was used for noise reduction and rice ear region segmentation. Based on the characteristics of color space, an association model is used to identify five types of colored grains in rice ear images; The output contains individual plants with five different colored seeds in a preset ratio.
4. The breeding method for five-colored rice per ear as described in claim 3, characterized in that, Based on color space characteristics, the specific steps for identifying five color categories of grains in a rice ear image using an association model include: A fast classification model based on RGB thresholds performs initial screening of five types of colored seeds by pre-setting the RGB value range of five types of colored seeds. A high-precision classification model based on convolutional neural networks is used to extract deep features from images through the ResNet-50 architecture for fine screening of five types of colored seeds.
5. The breeding method for five-colored rice per ear as described in claim 4, characterized in that, In the high-precision classification model based on convolutional neural networks, the ResNet-50 architecture is used to extract deep features from images for fine screening of five color seeds. First, data augmentation operations are performed on the original rice ear image, including rotation, scaling, and brightness adjustment.
6. The breeding method for five-colored rice with a single ear as described in claim 5, characterized in that, In the process of multi-generational synergistic breeding of individual plants by combining agronomic traits... Based on the goal of high-yield breeding, the screening thresholds were set as plant height 80-100cm, ear length 18-22cm, and tiller number 15-20.
7. A cultivation method for a single-ear five-colored rice variety, using a five-colored rice variety bred using the breeding method for single-ear five-colored rice as described in any one of claims 1-6, characterized in that, include: Select plots of land that meet specific soil conditions and carry out land preparation; The seeds of the five-colored rice variety undergo sun drying, seed selection, disinfection, and germination treatment. Sowing and seedling cultivation of seeds; Transplanting should be carried out while controlling the specifications and quality of the seeds. Fertilize and irrigate the seeds.