A rice tillering stage image recognition method

Abstract

The application discloses a rice tillering stage image recognition method, and specifically comprises the following steps: step S1, original image acquisition and processing, which is used for acquiring a clear image of a rice plant and providing preliminary processing for the image; step S2, plant image skeletonization, which generates a skeleton image in the form of a line segment from an image containing a rice plant, corresponding to stems and leaves and tillers of the plant; and step S3, skeleton pattern classification and recognition, which uses a basic algorithm based on deep learning to classify the rice skeleton image and determine whether tillering occurs; the rice tillering stage recognition based on machine vision technology realizes automatic agricultural operation reminding and agricultural operation decision making for the tillering growth period of the rice, replaces manual field inspection, has high timeliness and precision, meets the working mode requirements of low cost and unmanned operation, uses a camera as a front-end tool for recognition, is matched with corresponding back-end software algorithms and programs, and realizes complete self-tracking and self-judgment in the tillering stage.

Description

Technical Field

[0001] This invention belongs to the field of smart agriculture technology. Specifically, this invention relates to an image recognition method for rice tillering stage. Background Technology

[0002] With the development of modern agricultural technology, unmanned farms are rapidly emerging. To achieve full-cycle, all-encompassing unmanned farm applications for rice cultivation, precise determination of the rice growth stage is crucial. Accurately determining the rice growth stage allows for the correlation of relevant agricultural activities, guiding autonomous machine decision-making and providing effective support for the accurate and rational arrangement of related agricultural tasks such as pest and disease control, fertilization, and irrigation. The tillering stage is a critical growth period for rice, involving numerous related agricultural activities and requiring close monitoring.

[0003] The tillering stage of rice signifies the gradual development of its root system, the unfolding of its leaves, and the increasing number of tillers. This stage is crucial for rice cultivation and management, directly impacting the later quality of the plant and its final yield. Therefore, during this period, close attention must be paid to irrigation regulation, fertilization arrangements, pest and disease control, and responses to abnormal environmental conditions, with appropriate implementation of supporting agricultural practices. In traditional methods, determining the timing of the tillering stage and arranging agricultural activities rely on manual field inspections combined with human experience. This model, entirely dependent on human initiative and the accuracy of experience, is ill-suited to the demands of building unmanned farms requiring timely data collection, precise management, and standardized operations. Summary of the Invention

[0004] This invention provides a method for image recognition of rice tillering stage to solve the technical problems existing in the background art.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: a method for image recognition of rice tillering stage, specifically including the following steps: Step S1: Raw image acquisition and processing, used to obtain clear images of rice plants and provide preliminary image processing; Step S2: Skeletonization of plant images. The image containing rice plants is used to generate a skeleton image in the form of line segments, corresponding to the stems, leaves and tillers of the plant. Step S3: Skeleton pattern classification and recognition. Using a deep learning-based algorithm, the rice skeleton image is classified to determine whether tillering occurs.

[0006] Preferably, step S1 specifically includes the following steps: Step S1-1: Obtain a clear image of the rice plant by tilting the fixed camera. The fixed camera can adjust the shooting position and image shooting focal length to capture images of the quality required for image recognition. Step S1-2: Simultaneously triggered with step S1-1 is the illuminance sensor installed on the fixed camera. The illuminance sensor is synchronized with the fixed camera's shooting time and is used to correct the color and sharpness of the captured image. Step S1-3: The time of image capture is compared with the time calculated by the rice growth period model. By comparing the time with the rice tillering period time calculated by the rice growth period model, images that are significantly outside the normal tillering period time range are removed. Steps S1-4: The original color RGB image will be processed into grayscale to convert it into a grayscale image with clearer contrast. During the processing, data collected by the illuminance sensor will be used to provide more accurate calibration for the image grayscale processing.

[0007] Preferably, step S2 specifically includes the following steps: Step S2-1: Processing the image of rice plants after grayscale processing to generate a skeleton. Specifically, a skeleton extraction algorithm is used to generate the skeleton of the rice plant stem and leaves. Step S2-2: Incomplete rice plant skeleton images and mutually occluded plant skeleton images are defined as images that cannot be used to clearly identify individual plants. They are then classified with independent plant skeleton images to ensure that the image recognition object is a clear and independent rice plant skeleton image. Step S2-3: Segment and extract the skeleton image of an independent rice plant.

[0008] Preferably, step S3 specifically includes the following steps: Step S3-1: Determine the main stem and tillers of the plant based on the skeleton image of the independent rice plant. That is, determine the main stem and tillers of the plant based on the branching method and number of branches of the skeleton line segments. Step S3-2: Based on the extracted main stem and tillers of the plant, perform tiller feature extraction and classification; for rice that has not entered the tillering stage, there is only a main stem and no tillers, so only one image content of the main stem and tiller feature category can be extracted; for rice that has entered the tillering stage, there are at least two image contents that match the main stem and tiller feature categories. Step S3-3: Based on the tillering judgment result, output the tillering identification result.

[0009] The beneficial effects of adopting the above technical solutions are: 1. The rice tillering stage image recognition method of the present invention uses rice tillering stage recognition based on machine vision technology to realize automated agricultural reminders and agricultural decisions for this growth stage of rice tillering, replacing manual field inspections. It has both high timeliness and accuracy, and meets the requirements of low-cost and unmanned operation mode. The camera is used as the front-end tool for recognition, and is equipped with corresponding back-end software algorithms and programs to realize complete autonomous tracking and autonomous judgment of the tillering stage. Attached Figure Description

[0010] Figure 1 This is a flowchart of the rice tillering stage image recognition method of the present invention; Figure 2 This is a schematic diagram of the original image acquisition and processing; Figure 3 This is a schematic diagram of the skeletonization of a plant image; Figure 4 This is a schematic diagram of skeleton pattern classification and recognition; Detailed Implementation

[0011] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings, in order to help those skilled in the art to have a more complete, accurate and in-depth understanding of the concept and technical solutions of the present invention, and to facilitate its implementation.

[0012] like Figures 1 to 4 As shown, this invention is an image recognition method for rice tillering stage. It uses machine vision technology to identify the rice tillering stage, realizing automated agricultural reminders and decisions for this growth stage of rice, replacing manual field inspections. It has both high timeliness and accuracy, and meets the requirements of low-cost, unmanned operation mode. The camera is used as the front-end tool for recognition, and is equipped with corresponding back-end software algorithms and programs to achieve complete autonomous tracking and judgment of the tillering stage.

[0013] The specific working method is described below using specific embodiments: Example 1: Step S1: Original image acquisition and processing, used to obtain clear images of rice plants and provide preliminary processing for the images; Step S1 specifically includes the following steps: Step S1-1: Obtain a clear image of the rice plant by tilting the fixed camera. The fixed camera can adjust the shooting position and image shooting focal length to capture images of the quality required for image recognition. Step S1-2: Simultaneously triggered with step S1-1 is the illuminance sensor installed on the fixed camera. The illuminance sensor is synchronized with the fixed camera's shooting time and is used to correct the color and sharpness of the captured image. Step S1-3: The time of image capture is compared with the time calculated by the rice growth period model. By comparing the time with the rice tillering period time calculated by the rice growth period model, images that are significantly outside the normal tillering period time range are removed. Steps S1-4: The original color RGB image will be processed into grayscale to convert it into a grayscale image with clearer contrast. During the processing, data collected by the illuminance sensor will be used to provide more accurate calibration for the image grayscale processing.

[0014] Step S2: Skeletonization of plant images. The image containing rice plants is used to generate a skeleton image in the form of line segments, corresponding to the stems, leaves and tillers of the plant. Step S2 specifically includes the following steps. Step S2-1: Processing the image of rice plants after grayscale processing to generate a skeleton. Specifically, a skeleton extraction algorithm is used to generate the skeleton of the rice plant stem and leaves. Step S2-2: Incomplete rice plant skeleton images and mutually occluded plant skeleton images are defined as images that cannot be used to clearly identify individual plants. They are then classified with independent plant skeleton images to ensure that the image recognition object is a clear and independent rice plant skeleton image. Step S2-3: Segment and extract the skeleton image of an independent rice plant.

[0015] Step S3: Skeleton pattern classification and recognition. Using a deep learning-based algorithm, the rice skeleton image is classified to determine whether tillering occurs.

[0016] Step S3 specifically includes the following steps. Step S3-1: Determine the main stem and tillers of the plant based on the skeleton image of the independent rice plant. That is, determine the main stem and tillers of the plant based on the branching method and number of branches of the skeleton line segments. Step S3-2: Based on the extracted main stem and tillers of the plant, perform tiller feature extraction and classification; for rice that has not entered the tillering stage, there is only a main stem and no tillers, so only one image content of the main stem and tiller feature category can be extracted; for rice that has entered the tillering stage, there are at least two image contents that match the main stem and tiller feature categories. Step S3-3: Based on the tillering judgment result, output the tillering identification result.

[0017] The present invention has been described above by way of example with reference to the accompanying drawings. Obviously, the specific implementation of the present invention is not limited to the above-described manner. Any non-substantial improvements made using the inventive concept and technical solution of the present invention, or the direct application of the inventive concept and technical solution of the present invention to other situations without modification, are all within the protection scope of the present invention.

Claims

1. A method for image recognition of rice tillering stage, characterized in that: Specifically, it includes the following steps: Step S1: Raw image acquisition and processing, used to obtain clear images of rice plants and provide preliminary image processing; Step S2: Skeletonization of plant images. The image containing rice plants is used to generate a skeleton image in the form of line segments, corresponding to the stems, leaves and tillers of the plant. Step S3: Skeleton pattern classification and recognition. Using a deep learning-based algorithm, the rice skeleton image is classified to determine whether tillering occurs. Step S1 specifically includes the following steps: Step S1-1: Obtain a clear image of the rice plant by tilting the fixed camera. The fixed camera can adjust the shooting position and image shooting focal length to capture images of the quality required for image recognition. Step S1-2: Simultaneously triggered with step S1-1 is the illuminance sensor installed on the fixed camera. The illuminance sensor is synchronized with the fixed camera's shooting time and is used to correct the color and sharpness of the captured image. Step S1-3: The time of image capture is compared with the time calculated by the rice growth period model. By comparing the time with the rice tillering period time calculated by the rice growth period model, images that are significantly outside the normal tillering period time range are removed. Step S1-4: The original color RGB image will be processed into grayscale to convert it into a grayscale image with clearer contrast. During the processing, data collected by the illuminance sensor will be combined to provide more accurate calibration for the image grayscale processing. Step S2 specifically includes the following steps. Step S2-1: Processing the image of rice plants after grayscale processing to generate a skeleton. Specifically, a skeleton extraction algorithm is used to generate the skeleton of the rice plant stem and leaves. Step S2-2: Incomplete rice plant skeleton images and mutually occluded plant skeleton images are defined as images that cannot be used to clearly identify individual plants. They are then classified with independent plant skeleton images to ensure that the image recognition object is a clear and independent rice plant skeleton image. Step S2-3: Segment and extract the skeleton image of an independent rice plant; Step S3 specifically includes the following steps. Step S3-1: Determine the main stem and tillers of the plant based on the skeleton image of the independent rice plant. That is, determine the main stem and tillers of the plant based on the branching method and number of branches of the skeleton line segments. Step S3-2: Based on the extracted main stem and tillers of the plant, perform tiller feature extraction and classification; for rice that has not entered the tillering stage, there is only a main stem and no tillers, so only one image content of the main stem and tiller feature category can be extracted; for rice that has entered the tillering stage, there are at least two image contents that match the main stem and tiller feature categories. Step S3-3: Based on the tillering judgment result, output the tillering identification result.

Images