WO2022218185A1 - 用于植物病症诊断的方法和植物病症诊断系统 - Google Patents

用于植物病症诊断的方法和植物病症诊断系统 Download PDF

Info

Publication number
WO2022218185A1
WO2022218185A1 PCT/CN2022/085172 CN2022085172W WO2022218185A1 WO 2022218185 A1 WO2022218185 A1 WO 2022218185A1 CN 2022085172 W CN2022085172 W CN 2022085172W WO 2022218185 A1 WO2022218185 A1 WO 2022218185A1
Authority
WO
WIPO (PCT)
Prior art keywords
species
candidate
information
disease
plant
Prior art date
Application number
PCT/CN2022/085172
Other languages
English (en)
French (fr)
Inventor
徐青松
李青
Original Assignee
杭州睿胜软件有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 杭州睿胜软件有限公司 filed Critical 杭州睿胜软件有限公司
Priority to US18/286,754 priority Critical patent/US20240193778A1/en
Publication of WO2022218185A1 publication Critical patent/WO2022218185A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0012—Biomedical image inspection
    • G06T7/0014—Biomedical image inspection using an image reference approach
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0012—Biomedical image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30181—Earth observation
    • G06T2207/30188—Vegetation; Agriculture
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A40/00—Adaptation technologies in agriculture, forestry, livestock or agroalimentary production
    • Y02A40/10—Adaptation technologies in agriculture, forestry, livestock or agroalimentary production in agriculture

Definitions

  • the present disclosure relates to the field of computer technology, and in particular, to a method for diagnosing plant diseases and a system for diagnosing plant diseases.
  • One of the objectives of the present disclosure is to provide a method for diagnosing plant diseases, the method comprising: acquiring a plant image; determining identification information according to the plant image, wherein the identification information includes at least one of species information and disease information and extracting diagnostic information in the content management system according to the determined identification information, and outputting the diagnostic information.
  • At least part of the diagnostic information is changed with different plant images.
  • the at least part of the diagnostic information includes a reference map, the reference map corresponding to at least the condition information, and the reference map is similar to the plant image.
  • extracting diagnosis information in the content management system according to the determined identification information, and outputting the diagnosis information includes: in the content management system, determining a corresponding candidate reference gallery according to the disease information; In the reference gallery, based on the similarity with the plant image and/or the matching degree with the species information, determine the extracted one or more reference images and each of the one or more reference images. and outputting the one or more reference pictures, so that the one or more reference pictures are arranged in descending order of priority.
  • the candidate reference gallery based on the similarity with the plant image and/or the matching degree with the species information, it is determined that the extracted one or more reference images and the one or more reference images are The priority corresponding to each reference image in the multiple reference images includes: taking the preset number of candidate reference images with the highest similarity to the plant image in the candidate reference gallery as the first reference atlas, and using the candidate reference images as the first reference image set.
  • All other candidate reference images in the reference gallery are used as the second reference atlas; in the first reference atlas, a first reference image that matches the species information at the first species classification level is determined, wherein the determined first reference image
  • the reference map has a first priority; in the second reference map set, a second reference map that matches the species information at the first species classification level is determined, wherein the determined second reference map has a second priority, and the second priority is lower than the first priority; in the first reference atlas, a third reference map that matches the species information at a second species taxonomy level higher than the first species taxonomy level is determined, wherein all The determined third reference map has a third priority, and the third priority is lower than the second priority; and in the second reference map set, a fourth reference is determined that matches the species information at the second species taxonomy level diagram, wherein the determined fourth reference diagram has a fourth priority, and the fourth priority is lower than the third priority.
  • the candidate reference gallery based on the similarity with the plant image and/or the matching degree with the species information, it is determined that the extracted one or more reference images and the one or more reference images are
  • the priority corresponding to each reference picture in the or multiple reference pictures further includes: when neither the first reference picture set nor the second reference picture set is determined to be at a species classification level lower than or equal to the preset species classification level When the reference image matches the species information, a preset default image corresponding to the disease information is determined as the reference image.
  • the method further includes: cropping an edge region of the original image forming the reference image, so that the scale of the reference image obtained after cropping is consistent with a preset display scale, and the reference image The image features corresponding to the disease information are located in the middle area of the reference image.
  • extracting diagnosis information in the content management system according to the determined identification information, and outputting the diagnosis information includes: in the content management system, according to the determined identification information, extracting corresponding information according to a preset output field When the complete diagnostic data is extracted, the diagnostic information is generated according to the diagnostic data, and the diagnostic information is output.
  • extracting diagnostic information in the content management system according to the determined identification information, and outputting the diagnostic information further includes: when complete diagnostic data is not extracted, in the content management system, according to the determined The identification information is retrieved, and the corresponding diagnosis documents are retrieved; and the diagnosis information is generated according to the diagnosis documents, and the diagnosis information is output.
  • extracting diagnostic information in the content management system according to the determined identification information, and outputting the diagnostic information includes: in the content management system, generating the diagnostic information according to a preset output format, and outputting the diagnostic information.
  • the diagnostic data includes diagnostic summary data including a condition name corresponding to the condition name field in the preset output field and a at least one.
  • the diagnostic data includes diagnostic detail data including a symptom analysis corresponding to the symptom analysis field in the preset output field, a solution corresponding to the solution field in the preset output field, and a At least one of the preventive measures corresponding to the preventive measures field in the preset output field.
  • determining the identification information according to the plant image includes: when the current diagnosis mode is a passive diagnosis mode, determining a candidate species and a candidate disorder corresponding to at least part of the candidate species according to the plant image; for the candidate species having the corresponding candidate disorder , screening the candidate diseases of the candidate species according to the first preset condition; and after performing screening, if there are remaining diseases, the remaining diseases are used as disease information, and the candidate species corresponding to the remaining diseases is used as species information .
  • determining a candidate species and a candidate disease corresponding to at least part of the candidate species according to the plant image includes: judging whether the candidate species is in a preset white list of species; when the candidate species is in the preset white list of species When the candidate disease is selected, the disease candidate corresponding to the candidate species is determined according to the plant image and the candidate species; when the candidate species is not in the preset white list of species, the candidate disease corresponding to the candidate species is not determined for the candidate species candidate disease.
  • screening out the candidate conditions of the candidate species according to the first preset condition includes: when there are at least two candidate conditions corresponding to the same candidate species, according to the diagnostic confidence of the candidate conditions from high to high From the lowest order, screening is performed on the candidate symptoms of the candidate species according to the first preset condition, until the remaining symptoms are screened out or all the candidate symptoms corresponding to the candidate species are screened out.
  • screening the candidate disease of the candidate species according to the first preset condition includes: judging whether the candidate species is in the preset white list of species; when the candidate species is not in the preset species When in the white list, the candidate disease corresponding to the candidate species is screened out.
  • the screening of the candidate symptoms of the candidate species according to the first preset condition includes: comparing the diagnostic confidence of the candidate symptoms with the first preset confidence; when the diagnostic confidence of the candidate symptoms is less than When the first preset reliability is used, the candidate disease is screened out.
  • the screening of the candidate symptoms of the candidate species according to the first preset condition includes: comparing the diagnostic accuracy of the candidate symptoms with the preset accuracy; when the diagnostic accuracy of the candidate symptoms is less than the With a preset accuracy, the candidate condition is screened out.
  • screening the candidate disease of the candidate species according to the first preset condition includes: judging whether the candidate species is in the first candidate species blacklist corresponding to the candidate disease; when the candidate species is When in the first candidate species blacklist, the candidate condition is screened out.
  • determining the identification information according to the plant image further includes: when the current diagnosis mode is an active diagnosis mode, determining a candidate species and candidate disease information corresponding to the candidate species according to the plant image, wherein the candidate disease information includes a candidate The disease or the candidate disease is not detected; the disease information is screened from the candidate disease information according to the second preset condition.
  • screening disease information from candidate disease information according to a second preset condition includes: comparing species confidence of the candidate species with a second preset confidence, and comparing candidate species corresponding to the candidate species The diagnostic confidence of the disease information and the third preset reliability; the species confidence of the candidate species is greater than or equal to the second preset reliability, and the diagnostic confidence of the candidate disease information is greater than or equal to the In the third preset reliability, the candidate species is screened as the first undetermined species, and the candidate disease information is screened as the first undetermined disease information.
  • screening disease information from candidate disease information according to a second preset condition includes: judging whether the first undetermined species is in the second candidate species blacklist corresponding to the first undetermined disease information; When the species is in the second candidate species blacklist, the first pending disease information is screened out.
  • screening disease information from the candidate disease information according to a second preset condition includes: For each plant image, the first undetermined disease information corresponding to the plant image and having the maximum diagnostic confidence is screened as the second undetermined disease information of the plant image; The disease information is filtered out from the disease information.
  • determining the identification information according to the plant image further includes: when the current diagnosis mode is the active diagnosis mode, determining the candidate species corresponding to each plant image according to the plurality of plant images; The candidate species with the largest species confidence is screened as the second undetermined species corresponding to the plant image; among the second undetermined species corresponding to each plant image, the second undetermined species with the largest number is screened as the species information, or The second undetermined species with the largest number and the largest species confidence is screened as species information; disease information corresponding to the species information is determined.
  • determining the identification information according to the plant image includes: using a pre-trained species identification model to determine the species information according to the plant image; wherein the species identification model is a neural network model.
  • determining the identification information according to the plant image includes: using a pre-trained disease diagnosis model to determine disease information according to the plant image; wherein the disease diagnosis model is a neural network model.
  • a system for diagnosing plant diseases includes a processor and a memory, and the memory stores instructions, and when the instructions are executed by the processor, realize The steps of the method for the diagnosis of plant disorders as described above.
  • a computer-readable storage medium on which instructions are stored, and when the instructions are executed, the above-mentioned method for diagnosing plant diseases is implemented A step of.
  • a computer program product comprising instructions that, when executed by the processor, implement the steps of the method for plant disease diagnosis as described above .
  • FIG. 1 shows a schematic flowchart of a method for diagnosing plant diseases according to an exemplary embodiment of the present disclosure
  • FIG. 2 shows a schematic flowchart of a method for diagnosing plant diseases according to another exemplary embodiment of the present disclosure
  • FIG. 3 shows a schematic flowchart of a method for diagnosing plant diseases according to yet another exemplary embodiment of the present disclosure
  • FIG. 4 shows a schematic flowchart of step S300 according to a specific example of the present disclosure
  • FIG. 5 shows a schematic diagram of a diagnostic card according to a specific example of the present disclosure
  • FIG. 6 shows a schematic diagram of displayed diagnostic information according to a specific example of the present disclosure
  • FIG. 7 shows a block diagram of a plant disease diagnosis system according to an exemplary embodiment of the present disclosure.
  • FIG. 1 shows a schematic flowchart of a method for diagnosing plant diseases according to an exemplary embodiment of the present disclosure.
  • the method can be implemented in an application (app) installed on a smart terminal such as a mobile phone and a tablet computer. As shown in FIG. 1 , the method may include: step S100 , acquiring a plant image.
  • plant images uploaded by the user may be obtained directly.
  • corresponding prompt information may be generated and output to prompt the user to upload a plant image.
  • the prompt information may also include specific requirements for plant images, such as prompting the user to upload an image of the entire plant, a partial image of a plant's stem, leaf, etc., or a partial image of a site with obvious lesions, etc.
  • preprocessing such as marking can also be performed on multiple plant images, such as marking the whole plant image, the partial plant image (including marking the parts of the plant in the plant image), etc., so as to better to identify species information and/or condition information.
  • Step S200 determining identification information according to the plant image, wherein the identification information includes at least one of species information and disease information.
  • the identification of the plant image may include the identification of at least one of a species and a condition.
  • the diagnosis of the disorder can be performed in, for example, a passive diagnosis mode or an active diagnosis mode.
  • the passive diagnosis mode both species information and disease information can be determined according to the plant image when only certain preset conditions are met, and only the species information can be determined according to the plant image when the above preset conditions are not met.
  • the active diagnosis mode both species information and disease information can be determined from plant images.
  • condition information described herein may also include relevant information indicating that the plant is not suffering from any disease.
  • determining the identification information according to the plant image may include:
  • Step S211 when the current diagnosis mode is the passive diagnosis mode, determine candidate species and candidate diseases corresponding to at least some of the candidate species according to the plant image.
  • the candidate species and the candidate diseases corresponding to this plant image can be determined according to each plant image; the candidate species and the candidate diseases corresponding to the plant image can also be determined according to a plurality of associated plant images.
  • the candidate species and candidate conditions corresponding to the images can be determined according to each plant image; the candidate species and the candidate diseases corresponding to the plant image can also be determined according to a plurality of associated plant images.
  • a corresponding candidate condition may be determined for each candidate species.
  • the corresponding candidate conditions may also be determined only for some of the candidate species to simplify processing.
  • the diagnosis of disorders may be performed only for candidate species with high confidence in some species, so as to generate candidate disorders corresponding to these candidate species, waiting for further screening.
  • the species confidence refers to the probability that the species corresponding to the plant image is the candidate species.
  • determining candidate species and candidate conditions corresponding to at least part of the candidate species based on plant images may include:
  • the candidate disease corresponding to the candidate species is not determined for the candidate species.
  • the species included in the preset white list of species are generally common species or important species, and the diagnosis of the diseases of these species generally has high accuracy and reliability. That is, symptoms can be determined only for these species, thereby reducing processing difficulty, and outputting inaccurate, unreliable, or unimportant symptoms to the user can be avoided.
  • a pre-trained species identification model can be used to determine candidate species or species information from plant images.
  • the species identification model may be a neural network model, specifically a convolutional neural network model or a residual network model.
  • the convolutional neural network model is a deep feedforward neural network, which uses a convolution kernel to scan the plant image, extracts the features to be recognized in the plant image, and then recognizes the features to be recognized of the plant.
  • the original plant images can be directly input into the convolutional neural network model without preprocessing the plant images.
  • the convolutional neural network model has higher recognition accuracy and recognition efficiency.
  • the residual network model Compared with the convolutional neural network model, the residual network model has more identity mapping layers, which can avoid the phenomenon of accuracy saturation or even decline caused by the increase of the network depth (the number of layers in the network).
  • the identity mapping function of the identity mapping layer in the residual network model needs to satisfy: the sum of the identity mapping function and the input of the residual network model is equal to the output of the residual network model. After the introduction of identity mapping, the residual network model has more obvious changes to the output, so it can greatly improve the identification accuracy and identification efficiency of plant species.
  • training the species recognition model may include:
  • the training ends when the first training accuracy rate is greater than or equal to the first preset accuracy rate, and a trained species identification model is obtained.
  • a large number of plant images may be included, and each plant image is correspondingly marked with a species.
  • the species identification model can also output multiple candidate species, wherein each candidate species can have its corresponding species confidence for further analysis and screening.
  • the trained species recognition model can also be tested, which can specifically include:
  • the first training set and/or the species identification model are adjusted for retraining.
  • the plant images in the first test set and the first training set are not exactly the same, so the first test set can be used to test whether the species recognition model also has a good recognition effect on plant images other than the first training set .
  • the first model accuracy rate of the species identification model is calculated by comparing the species outputted according to the plant images in the first test set and the labeled species.
  • the calculation method of the first model accuracy rate may be the same as the calculation method of the first training accuracy rate.
  • the first training set can be adjusted, for example, plants marked with species in the first training set can be added.
  • the second preset accuracy rate may be set equal to the first preset accuracy rate.
  • a pre-trained disease diagnosis model can be used to determine candidate diseases or disease information according to plant images.
  • condition information may include a candidate condition or no candidate condition was detected.
  • the disease diagnosis model may be a neural network model, specifically a convolutional neural network model or a residual network model.
  • training the condition diagnosis model may include:
  • the training ends when the second training accuracy rate is greater than or equal to the third preset accuracy rate, and a trained disease diagnosis model is obtained.
  • the second sample set may include a large number of plant images, and each plant image is correspondingly marked with disease information, for example, the disease information may be the disease of the plant in the plant image, or the disease related to the of plants corresponding to no disease detected.
  • the plant images in the second sample set may be at least partially identical to the plant images in the first sample set.
  • the plant image is input into the disease diagnosis model to generate the output disease information, and then according to the comparison result between the output disease information and the labeled disease information, the relevant parameters in the disease diagnosis model can be adjusted, that is, the disease diagnosis model is trained. , until the training ends when the second training accuracy rate of the disease diagnosis model is greater than or equal to the third preset accuracy rate, thereby obtaining the trained disease diagnosis model.
  • the disease diagnosis model can output multiple candidate disease information, wherein each candidate disease information can have its corresponding diagnostic confidence for further analysis and screening.
  • the diagnostic confidence refers to the probability that the disease information corresponding to the plant image is the candidate disease information.
  • disease diagnosis model can also be tested, which can specifically include:
  • the second training set and/or the disease diagnosis model are adjusted for retraining.
  • the plant images in the second test set and the second training set are not exactly the same, so the second test set can be used to test whether the disease diagnosis model has a good diagnostic effect on plant images other than the second training set.
  • the second model accuracy rate of the disease diagnosis model is calculated by comparing the output disease information generated according to the plant images in the second test set and the labeled disease information.
  • the calculation method of the second model accuracy rate may be the same as the calculation method of the second training accuracy rate.
  • the disease information marked in the second training set can be added. the number of plant images, or adjust the disease diagnosis model itself, or adjust both of the above, and then retrain the disease diagnosis model to improve its diagnosis.
  • the fourth preset accuracy rate may be set equal to the third preset accuracy rate.
  • the identification and diagnosis of species and diseases can also be implemented by the same pre-trained model, that is, the model can integrate the functions of the above-mentioned species identification model and disease diagnosis model.
  • determining at least one of the species information and the disease information according to the plant image may further include: Step S212 , for the candidate species having the corresponding candidate disease, according to the first preset condition, the candidate disease of the candidate species Perform screening.
  • the main purpose of the user is not to diagnose the condition itself, but to determine the species information of plants, etc., for example.
  • only diseases with high accuracy and reliability can be output, so as to help users discover plant diseases in time, and avoid additional troubles to users due to inaccurate output diseases.
  • symptoms with low accuracy and reliability among the candidate symptoms can be screened out according to the first preset condition.
  • candidate species when candidate species are determined from plant images, it may be determined that there are one or more candidate species.
  • one or more candidate conditions may be determined for each candidate species, or, as described above, candidate conditions may only be determined for some of the candidate species. Further, for each candidate species having a corresponding candidate disease, the candidate disease corresponding to the candidate species may be screened out according to the first preset condition.
  • the determined candidate species include species 1, species 2 and species 3.
  • Species 1, 2, and 3 are listed in descending order of species confidence. For example, species 1 has a species confidence of 0.8, species 2 has a species confidence of 0.75, and species 3 has a species confidence of 0.7.
  • the symptoms that can be output are determined according to the first preset condition for each species.
  • the diagnostic confidence for Condition 2-1 is 95%, the diagnostic confidence for Condition 2-2 is 90%, and the diagnostic confidence for Condition 2-3 is 82%, then during screening for Species 2 , which can be screened in the order of diagnosis confidence from high to low, and in the order of disease 2-1, disease 2-2 and disease 2-3. If no remaining symptoms that can be used for output are found after screening for disease 2-1, continue screening for disease 2-2. If after screening the disease 2-1, the remaining symptoms that can be used for output have been found, then the screening of the disease 2-2 may not be continued, so as to simplify the whole processing process. Of course, in some cases, it is also possible that after screening all candidate symptoms, still no remaining symptoms that can be used for output are found, the screening is stopped, and no symptoms are output in subsequent steps.
  • the first preset condition involved may be related to a variety of factors, such as the diagnostic confidence of the current diagnosis, the type of candidate species, the diagnostic accuracy of a certain type of disease, and the candidate species and One or more of the degree of matching among the candidate conditions, etc.
  • the screening of the candidate diseases of the candidate species according to the first preset condition may include: judging whether the candidate species is in the preset white list of species; and when the candidate species is not in the preset white list of species, screening the candidate species. Eliminate candidate conditions corresponding to candidate species.
  • screening may be performed first according to a preset whitelist of species, so as to reduce the amount of data to be processed in subsequent screening.
  • the species included in the preset white list of species are generally common species or important species, and the diagnosis of the diseases of these species generally has high accuracy and reliability.
  • only the symptoms corresponding to these species may be output in subsequent steps without being screened out, so as to avoid outputting inaccurate, unreliable or unimportant symptoms to the user as much as possible to avoid causing the user to additional trouble.
  • the screening of the candidate disease of the candidate species according to the first preset condition may include: comparing the diagnostic confidence of the candidate disease with the first preset confidence; and when the diagnostic confidence of the candidate disease is less than the first When the reliability is preset, candidate symptoms are screened out.
  • the diagnostic confidence can represent the reliability of the obtained condition in a single diagnosis process.
  • the first preset reliability may be set to 70%. That is to say, when the diagnostic confidence of the candidate disease is less than 70%, the candidate disease will be screened out and not output, so as to avoid the user's troubles caused by the output diagnosis information inconsistent with the actual situation.
  • performing screening on the candidate disease of the candidate species according to the first preset condition may include: comparing the diagnostic accuracy of the candidate disease with the preset accuracy; and when the diagnostic accuracy of the candidate disease is less than the preset accuracy When the candidate condition is screened out.
  • diagnostic accuracy reflects the overall accuracy of identifying a particular type of condition.
  • the diagnostic accuracy can be obtained according to the ratio of the number of correct diagnoses to the total number of times in a certain total number of diagnoses. For some diseases that are difficult to diagnose, their diagnostic accuracy is often low, so by screening out candidate diseases related to these diseases, the output of inaccurate diseases can be avoided as much as possible.
  • the screening of the candidate diseases of the candidate species according to the first preset condition may include: judging whether the candidate species is in the first candidate species blacklist corresponding to the candidate symptoms; and when the candidate species is in the first candidate species When the species is in the blacklist, the candidate diseases are screened out.
  • a corresponding blacklist of candidate species can be preset, so as to screen out the candidate diseases, so as to improve the accuracy and reliability of the output.
  • the above-mentioned specific methods on how to screen out candidate conditions that satisfy the first preset condition can be combined with each other. For example, in a specific example, as long as the candidate condition satisfies that the diagnostic confidence is less than the first preset confidence, the candidate species is not in the preset white list of species, the diagnostic accuracy is less than the preset accuracy, and the candidate species is in the Any condition in the first candidate species blacklist, the candidate disease will be excluded.
  • determining the identification information according to the plant image may further include:
  • Step S213 after performing screening, if there are remaining symptoms, the remaining symptoms are used as disease information, and the candidate species corresponding to the remaining symptoms are used as species information.
  • the remaining disease can be used as disease information, and the candidate species corresponding to the remaining disease can be used as species information for subsequent processing.
  • the disease information can be empty (or the plant is healthy), correspondingly, in the subsequent steps, no disease can be output, or the symptom field (Symptom) can be filled with the content "healthy” (Health)” output, so as not to cause additional trouble to the user.
  • determining the identification information according to the plant image may further include:
  • Step S221 when the current diagnosis mode is the active diagnosis mode, the candidate species and the candidate disease information corresponding to the candidate species are determined according to the plant image.
  • candidate species and candidate disease information corresponding to each candidate species can be determined according to the plant image, so as to obtain the health status of the plant as comprehensively as possible for further analysis and processing by the user.
  • the candidate disease information may include candidate disease or no candidate disease detected.
  • the candidate species can be determined by the above-mentioned species identification model, and the candidate disease information can be determined by the above-mentioned disease diagnosis model, which will not be repeated here.
  • determining the identification information according to the plant image may further include:
  • Step S222 screening out the disease information from the candidate disease information according to the second preset condition.
  • the active diagnosis mode Compared with the passive diagnosis mode, in the active diagnosis mode, more candidate disease information can be included in the final disease information for user reference, and at the same time, the accuracy and reliability of the generated disease information can be appropriately reduced. Require.
  • screening the disease information from the candidate disease information according to the second preset condition may include:
  • the candidate species is screened as the first undetermined species, and the candidate disease is selected as the first undetermined species.
  • the information is screened as the first undetermined disease information.
  • screening out the disease information from the candidate disease information according to the second preset condition may further include: judging whether the first pending species is in the second candidate species blacklist corresponding to the first pending disease information; When the species is in the second candidate species blacklist, the first pending disease information is screened out.
  • a blacklist of candidate species corresponding to the disease information can be preset according to the mutually exclusive relationship between such species and diseases, so as to further screen out the first pending disease information, on the one hand, the amount of data to be processed can be reduced. , to improve the processing efficiency, and on the other hand, it also helps to further improve the accuracy and reliability of the output.
  • screening out the disease information from the candidate disease information according to the second preset condition may include: for each plant image, respectively, selecting the first pending disease information corresponding to the plant image and having the maximum diagnostic confidence as the one of the plant image. second undetermined disease information; and screening disease information from the second undetermined disease information of all plant images according to a third preset condition.
  • the second undetermined disease information corresponding to the plant image with the greatest diagnostic confidence is screened out.
  • part of the disease information in the candidate disease information may have been screened out according to one or some methods in the embodiments in the active identification mode described above, so the diagnostic confidence of the second pending disease information is not the same. It is not necessarily the disease information with the largest diagnostic confidence among all the candidate disease information corresponding to the plant image.
  • the second undetermined disease information corresponding to each plant image is aggregated, and disease information is filtered out.
  • the disease information is generally disease information with the greatest accuracy and reliability. Specifically, the disease information can be screened out by referring to the method for screening out certain disease information according to the first preset condition described above. Alternatively, disease information can also be screened out according to other preset conditions.
  • the disease information may be the disease of the plant and its related information, so as to help the user diagnose the disease and take further corresponding measures.
  • the disease information may also be that the disease is not detected, which means that the plant is currently in a relatively healthy state.
  • species information when determining disease information, species information may also be uniquely determined according to a plant image, and then disease information may be determined according to the species information.
  • determining the identification information according to the plant image may further include:
  • Step S231 when the current diagnosis mode is the active diagnosis mode, determine the candidate species corresponding to each plant image according to the plurality of plant images;
  • Step S232 for each plant image, screen the candidate species with the largest species confidence as the second undetermined species corresponding to the plant image;
  • Step S233 among the second pending species corresponding to each plant image, screen the second pending species with the largest number as species information, or screen the second pending species with the largest number and the largest species confidence as species information.
  • the determined candidate species corresponding to plant image A include species M, species N and species P
  • the candidate species corresponding to plant image B include species N and species P
  • the candidate species corresponding to plant image C include species N and species Q.
  • the screened second undetermined species include species M corresponding to plant image A, species N corresponding to plant image B, and species N corresponding to plant image C. It can be seen that in the second undetermined species corresponding to each plant image, the number of species M is 1, and the number of species N is 2. In this case, species N will be used as species information.
  • the determined candidate species corresponding to plant image A include species M, species N and For species P, the candidate species corresponding to plant image B include species N and species P, and the candidate species corresponding to plant image C include species P and species N.
  • the screened second undetermined species include species M corresponding to plant image A, species N corresponding to plant image B, and species P corresponding to plant image C. It can be seen that in the second undetermined species corresponding to each plant image, the number of species M, species N and species P are all 1.
  • the species with the largest species confidence among species M, N, and P is used as species information. For example, if the species confidence of species M is greater than that of species N, and the species confidence of species N is greater than that of species P, then species M will be used as species information.
  • determining the identification information according to the plant image may further include:
  • Step S234 determining disease information corresponding to the species information.
  • the disease information corresponding to the resultant species is further determined. Specifically, after the species information is determined, based on the species information, the disease diagnosis model can be used to determine candidate disease information according to the plant image, and then according to the first preset condition, the second preset condition or the third preset condition described above Preset conditions and the like are used to screen candidate disease information, thereby obtaining disease information for output.
  • the disease information may be the disease of the plant and its related information, so as to help the user diagnose the disease and take further corresponding measures.
  • the disease information may also be that the disease is not detected, which means that the plant is currently in a relatively healthy state.
  • the method for diagnosing plant diseases may further include: step S300 , extracting diagnosis information from the content management system according to at least one of the determined species information and disease information, and outputting the diagnosis information.
  • the content management system may be a software system located between the systems or processes of the front-end and back-end of the WEB.
  • Content management systems may be used to submit, modify, publish, etc. content such as text files, pictures, data in databases, forms, and the like.
  • the content management system can also provide content crawling tools to automatically capture content from third parties such as text files, HTML web pages, Web services, databases, etc., and put them into the corresponding content library of the content management system after analysis and processing. middle.
  • the content management system can also assist the WEB front end to provide content to users in a personalized way, that is, to provide a personalized portal framework to better push the content to users based on WEB technology.
  • descriptive content about plants and their diseases may be stored, and these descriptive content may be text or pictures, for example, may include various fields, articles, etc., so as to It enables users to obtain introductions about plants and their diseases, such as interesting stories, uses of plants, maintenance methods, and descriptions of diseases, in the diagnostic information extracted and output from the content management system.
  • One-to-one correspondence with each species information can include a species name (UID1) to distinguish different species.
  • a one-to-one correspondence with each condition information can include a condition name (UID2 or ComnonName) to distinguish different conditions.
  • the relevant information of multiple species can be output to users in the form of one species corresponding to one card. Users can switch to display various species and their related information by sliding cards on the interactive interface.
  • At least part of the diagnostic information can be changed with different plant images.
  • the output diagnosis information can be adaptively changed according to the plant image input by the user, which realizes a more flexible output and helps to make the output diagnosis information match the user's input. Matching, thereby improving the user experience and reducing the confusion caused to the user due to the mismatch of input and output.
  • extracting diagnostic information in the content management system according to the determined identification information, and outputting the diagnostic information may include: step S310 , in the content management system, according to the determined identification information , extract the corresponding diagnostic data according to the preset output field; and step S320 , when complete diagnostic data is extracted, generate diagnostic information according to the diagnostic data, and output the diagnostic information.
  • the preset output fields may be set by the user through an interactive interface according to their own needs, or the preset output fields may also be relatively fixed several fields.
  • corresponding diagnostic data extracted according to the determined identification information may be filled in a corresponding template having a preset output format to form diagnostic information. Diagnostic information can be organized or laid out in the form of cards, labels, etc., and output to the user. Users can select, switch or move cards, labels, etc. on the interactive interface, and read the specific content contained in one or some cards and labels according to their needs, so as to obtain relevant information.
  • the diagnostic data may include diagnostic summary data and/or diagnostic detail data.
  • diagnostic summary data and the diagnostic detailed data different fields can be set to store the data extracted from the content management system in the corresponding fields.
  • the diagnosis summary data may include at least one of a condition name corresponding to the condition name field in the preset output field and a diagnosis summary corresponding to the diagnosis summary field in the preset output field.
  • the condition name and diagnosis summary may be displayed as a diagnosis card as shown in FIG. 5 .
  • “Black Spot” is the disease name, and "Your plant get black spot on the leaves. It is due to xxx,xxx,xxxx,xxxxx,xxx,xxx,xxxx,xxxxx,xxxxxxx,xxxxxxx” is Diagnostic summary.
  • other buttons (such as "Check for causes") may also be included in the diagnosis card shown in FIG. 5 to link to other more detailed diagnosis information.
  • the diagnostic detail data may include a symptom analysis corresponding to the symptom analysis field in the preset output field, a solution corresponding to the solution field in the preset output field, and preventive measures in the preset output field Field at least one of the corresponding precautions.
  • FIG. 6 it is a schematic diagram of displayed diagnostic information in a specific example.
  • the plant image from the user is displayed at the top of the interface
  • the diagnosis card including the diagnosis summary data can be located under the plant image
  • further diagnosis detailed data can be located under the diagnosis card.
  • the complete diagnostic data may not be extracted in the content management system.
  • the identification information determined according to the plant image may be output according to the old method of displaying fixed content, or the identification information determined according to the determined identification information may be output. , to obtain diagnostic information from the relevant literature.
  • extracting diagnostic information in the content management system according to the determined identification information, and outputting the diagnostic information may further include: Step S331, when complete diagnostic data is not extracted, in the content management system , according to the determined identification information, retrieve the corresponding diagnosis document; step S332 , generate diagnosis information according to the diagnosis document, and output the diagnosis information.
  • a reference map (the map located at the bottom of FIG. 6 ) can be included.
  • the reference image corresponds to at least disease information, and the reference image is similar to the plant image.
  • the output diagnostic information can no longer be fixed, but the relevant pictures used for explanation in the output diagnostic information can be replaced according to the plant image input by the user, so that these pictures used for explanation are the same as those used for explanation.
  • the plant images taken by the user are more similar, so that the user does not feel that the image in the output diagnostic information is too different from the plant image taken by the user, so as to avoid causing trouble to the user and improving the user experience.
  • extracting diagnosis information in the content management system according to the determined identification information, and outputting the diagnosis information may include: in the content management system, determining a corresponding candidate reference gallery according to the disease information; , based on the similarity with the plant image and/or the match with the species information, determine the extracted one or more reference images and the priority corresponding to each reference image in the one or more reference images; and One or more reference pictures are output, so that the one or more reference pictures are arranged in descending order of priority.
  • each reference image in the content management system may be marked with UID1 of corresponding species information (UID1 may include species, variety, variety, genus, family, etc.) and UID2 of disease information.
  • UID1 may include species, variety, variety, genus, family, etc.
  • UID2 of disease information
  • the reference images can be classified, filtered, and the like.
  • one or more reference pictures corresponding to each disease information can be formed into candidate reference libraries corresponding to the corresponding disease information respectively.
  • the type of plant corresponding to the reference image can be determined according to the UID1 marked on each reference image.
  • the user can better identify the disease of the plant, especially when the plant image taken by the user is unclear or the shooting position is not good.
  • the reference pictures with higher priority may be displayed preferentially or arranged in front of the displayed reference pictures, so as to facilitate the user's viewing.
  • the candidate reference gallery based on the similarity with the plant image and/or the matching degree with the species information, it is determined that the extracted one or more reference images and the one or more reference images are Or the priority corresponding to each reference picture in the multiple reference pictures may include:
  • the first reference atlas determining a first reference map that matches the species information at a first species taxonomy level, wherein the determined first reference map has a first priority
  • a second reference map that matches the species information at the first species taxonomy level is determined, wherein the determined second reference map has a second priority, and the second priority is lower than the first a priority;
  • a third reference map that matches the species information at a second species taxonomy level higher than the first species taxonomy level is determined, wherein the determined third reference map has a third priority, and the third priority is lower than the second priority;
  • a fourth reference map that matches the species information at the second species taxonomy level is determined, wherein the determined fourth reference map has a fourth priority, and the fourth priority is lower than the third Three priority.
  • six candidate reference images that are closest to the image features of the plant images uploaded by the user may be determined from the candidate reference gallery as the first atlas, and other candidate reference images in the candidate reference gallery are used as the second. Atlas.
  • the candidate reference gallery based on the similarity with the plant image and/or the matching degree with the species information, it is determined that the extracted one or more reference images and the one or more reference images are extracted
  • the priority corresponding to each reference picture in the reference pictures may also include:
  • the prediction corresponding to the disease information Set the default image to be the reference image. For example, if you still cannot find a matching result reference image in the first atlas or the second atlas when you find the classification level of family, you can use the preset default image in the reference gallery as the reference image and do not search further. .
  • the display ratio of the pictures in the displayed diagnostic information is between 3:2 and 1:1, so as to have a better display effect.
  • the scale of the reference images filtered from the candidate reference gallery may not be suitable for the above display scale. Often, such figures can be stretched or cropped to fit the display scale. However, considering that when the reference image is stretched, the characteristics of some diseases may be deformed, which is not conducive to the user's good identification of the symptoms. Therefore, in an exemplary embodiment of the present disclosure, a cutting method can be used to process it. Reference image.
  • the method for diagnosing plant diseases may further include: cropping the edge area of the original picture forming the reference picture, so that the scale of the reference picture obtained after cropping is consistent with the preset display scale, and the reference picture
  • the image features corresponding to the disease information in are located in the middle area of the reference image.
  • the images whose image features corresponding to the disease information are located in the edge area are removed or ignored, and these images will not be included in the content management system.
  • the pictures can be processed, such as cropping, when they are stored in the content management system.
  • the reference image of the selected content management system may be processed, such as cropping, before being output.
  • the plant disease system 900 may include a processor 910 and a memory 920.
  • the memory 920 stores instructions.
  • the above-described method for plant disease diagnosis can be implemented. A step of.
  • the processor 910 can perform various actions and processes according to the instructions stored in the memory 920 .
  • the processor 910 may be an integrated circuit chip with signal processing capability.
  • the aforementioned processors may be general purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), off-the-shelf programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.
  • DSPs digital signal processors
  • ASICs application specific integrated circuits
  • FPGAs off-the-shelf programmable gate arrays
  • Various methods, steps and logic block diagrams disclosed in the embodiments of the present disclosure can be implemented or executed.
  • the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc., and may be an X86 architecture or an ARM architecture, or the like.
  • the memory 920 stores executable instructions that are executed by the processor 910 for the above-described method for vegetative disease diagnosis.
  • Memory 920 may be volatile memory or nonvolatile memory, or may include both volatile and nonvolatile memory.
  • the nonvolatile memory may be read only memory (ROM), programmable read only memory (PROM), erasable programmable read only memory (EPROM), electrically erasable programmable read only memory (EEPROM), or flash memory.
  • Volatile memory may be random access memory (RAM), which acts as an external cache.
  • RAM Random Access Memory
  • SRAM static random access memory
  • DRAM dynamic random access memory
  • SDRAM synchronous dynamic random access memory
  • DDRSDRAM double data rate synchronous dynamic Random Access Memory
  • ESDRAM Enhanced Synchronous Dynamic Random Access Memory
  • SLDRAM Synchronous Link Dynamic Random Access Memory
  • DR RAM Direct Memory Bus Random Access Memory
  • the plant disease diagnosis system may also include a content management system.
  • descriptive content about plants and their diseases may be stored, and these descriptive content may be text or pictures, for example, may include various fields, articles, etc., so as to It enables users to obtain introductions about plants and their diseases, such as interesting stories, uses of plants, maintenance methods, and descriptions of diseases, in the diagnostic information extracted and output from the content management system.
  • the content management system may also be independent of the plant disease diagnosis system, and the plant disease diagnosis system may be communicatively connected with the content management system to obtain relevant content.
  • a computer-readable storage medium is provided, and instructions are stored on the computer-readable storage medium.
  • the above-described methods for diagnosing plant diseases can be implemented. step.
  • computer-readable storage media in embodiments of the present disclosure may be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. It should be noted that computer-readable storage media described herein are intended to include, but not be limited to, these and any other suitable types of memory.
  • the present disclosure also proposes a computer program product, which can include instructions that, when executed by a processor, can implement the steps of the method for plant disease diagnosis as described above.
  • the instructions may be any set of instructions to be executed directly by one or more processors, such as machine code, or any set of instructions to be executed indirectly, such as scripts.
  • the terms "instructions,” “applications,” “processes,” “steps,” and “programs” are used interchangeably herein. Instructions may be stored in object code format for direct processing by one or more processors, or in any other computer language, including scripts or collections of self-contained source code modules that are interpreted on demand or compiled ahead of time.
  • the instructions may include instructions that cause, for example, one or more processors to function as the various neural networks herein. The functions, methods, and routines of the instructions are explained in more detail elsewhere in this document.
  • each block in the flowchart or block diagrams may represent a module, segment, or portion of code that contains one or more logic for implementing the specified logic Executable instructions for the function.
  • the functions noted in the blocks may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved.
  • each block of the block diagrams and/or flowchart illustrations, and combinations of blocks in the block diagrams and/or flowchart illustrations can be implemented in dedicated hardware-based systems that perform the specified functions or operations , or can be implemented in a combination of dedicated hardware and computer instructions.
  • the various example embodiments of the present disclosure may be implemented in hardware or special purpose circuits, software, firmware, logic, or any combination thereof. Certain aspects may be implemented in hardware, while other aspects may be implemented in firmware or software that may be executed by a controller, microprocessor or other computing device. While aspects of the embodiments of the present disclosure are illustrated or described as block diagrams, flowcharts, or using some other graphical representation, it is to be understood that the blocks, apparatus, systems, techniques, or methods described herein may be taken as non-limiting Examples are implemented in hardware, software, firmware, special purpose circuits or logic, general purpose hardware or controllers or other computing devices, or some combination thereof.
  • the word "exemplary” means “serving as an example, instance, or illustration” rather than as a “model” to be exactly reproduced. Any implementation illustratively described herein is not necessarily to be construed as preferred or advantageous over other implementations. Furthermore, the present disclosure is not to be bound by any expressed or implied theory presented in the preceding technical field, background, brief summary or detailed description.
  • the word “substantially” is meant to encompass any minor variation due to design or manufacturing imperfections, tolerances of devices or elements, environmental influences, and/or other factors.
  • the word “substantially” also allows for differences from a perfect or ideal situation due to parasitics, noise, and other practical considerations that may exist in an actual implementation.
  • connection means that one element/node/feature is electrically, mechanically, logically or otherwise directly connected to another element/node/feature (or direct communication).
  • coupled means that one element/node/feature can be mechanically, electrically, logically or otherwise linked, directly or indirectly, with another element/node/feature to allow interaction, even though the two features may not be directly connected. That is, “coupled” is intended to encompass both direct and indirect connections of elements or other features, including connections utilizing one or more intervening elements.
  • first,” “second,” and the like may also be used herein for reference purposes only, and are thus not intended to be limiting.
  • the terms “first,” “second,” and other such numerical terms referring to structures or elements do not imply a sequence or order unless the context clearly dictates otherwise.
  • providing is used broadly to encompass all ways of obtaining an object, thus “providing something” includes, but is not limited to, “purchasing,” “preparing/manufacturing,” “arranging/arranging,” “installing/ Assembly”, and/or “Order” objects, etc.

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Medical Informatics (AREA)
  • Biomedical Technology (AREA)
  • Quality & Reliability (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Radiology & Medical Imaging (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Data Mining & Analysis (AREA)
  • Computational Linguistics (AREA)
  • Molecular Biology (AREA)
  • Public Health (AREA)
  • Software Systems (AREA)
  • Mathematical Physics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Biophysics (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • Pathology (AREA)
  • Databases & Information Systems (AREA)
  • Primary Health Care (AREA)
  • Epidemiology (AREA)
  • Medical Treatment And Welfare Office Work (AREA)

Abstract

本公开涉及一种用于植物病症诊断的方法和植物病症诊断系统。所述方法包括：获取植物图像；根据植物图像确定识别信息，其中，所述识别信息包括物种信息和病症信息中的至少一种；以及根据所确定的识别信息，在内容管理系统中提取诊断信息，并输出所述诊断信息。

Description

用于植物病症诊断的方法和植物病症诊断系统 技术领域
本公开涉及计算机技术领域，具体来说，涉及一种用于植物病症诊断的方法和植物病症诊断系统。
背景技术
在植物的生长过程中，常常遭遇疾病、虫害等问题的困扰。目前，通常是由专业的管理人员来发现并处理这些问题。然而，如果相关人员没有能够发现这些病症，那么很可能对植物的生长造成严重的不良影响。
发明内容
本公开的目的之一是提供一种用于植物病症诊断的方法，所述方法包括：获取植物图像；根据植物图像确定识别信息，其中，所述识别信息包括物种信息和病症信息中的至少一种；以及根据所确定的识别信息，在内容管理系统中提取诊断信息，并输出所述诊断信息。
在一些实施例中，针对不同的植物图像，在所确定的识别信息相同的情况下，至少部分诊断信息是随着不同的植物图像而改变的。
在一些实施例中，所述至少部分诊断信息包括参考图，所述参考图至少与所述病症信息对应，且所述参考图与所述植物图像相似。
在一些实施例中，根据所确定的识别信息，在内容管理系统中提取诊断信息，并输出所述诊断信息包括：在内容管理系统中，根据所述病症信息确定对应的候选参考图库；在候选参考图库中，基于与所述植物图像的相似度和/或与所述物种信息的匹配度，确定被提取的一幅或多幅参考图以及与所述一幅或多幅参考图中的每幅参考图对应的优先级；以及输出所述一幅或多幅参考图，使得所述一幅或多幅参考图按照优先级由高至低的顺序被依次排列。
在一些实施例中，在候选参考图库中，基于与所述植物图像的相似度和/或与所述物种信息的匹配度，确定被提取的一幅或多幅参考图以及与所述一幅或多幅参考图中的每幅参考图对应的优先级包括：将候选参考图库中的与所述植物图像的相似度最高的预设数量的候选参考图作为第一参考图集，并将候选参考图库中的所有其他候选参考图作为第二参考图集；在第一参考图集中，确定在第一物种分类级别上与所述物种信息匹配的第一参考图，其中，所确定的第一参考图具有第一优先级；在第二参考图集中，确定在第一物种分类级别上与所述物种信息匹配的第二参考图，其中，所确定的第二参考图具有第二优先级，且第二优先级低于第一优先级；在第一参考图集中，确定在高于第一物种分类级别的第二物种分类级别上与所述物种信息匹配的第三参考图，其中，所确定的第三参考图具有第三优先级，且第三优先级低于第二优先级；以及在第二参考图集中，确定在第二物种分类级别上与所述物种 信息匹配的第四参考图，其中，所确定的第四参考图具有第四优先级，且第四优先级低于第三优先级。
在一些实施例中，在候选参考图库中，基于与所述植物图像的相似度和/或与所述物种信息的匹配度，确定被提取的一幅或多幅参考图以及与所述一幅或多幅参考图中的每幅参考图对应的优先级还包括：当在第一参考图集和第二参考图集中均未能确定在低于或等于预设物种分类级别的物种分类级别上与所述物种信息匹配的参考图时，将与所述病症信息对应的预设默认图确定为参考图。
在一些实施例中，所述方法还包括：对形成参考图的原始图的边缘区域进行裁切，以使得裁切后所得的参考图的比例与预设显示比例相符，且所述参考图中的与所述病症信息对应的图像特征位于所述参考图的中部区域内。
在一些实施例中，根据所确定的识别信息，在内容管理系统中提取诊断信息，并输出所述诊断信息包括：在内容管理系统中，根据所确定的识别信息，按照预设输出字段提取相应的诊断数据；在提取到完整的诊断数据时，根据诊断数据生成诊断信息，并输出诊断信息。
在一些实施例中，根据所确定的识别信息，在内容管理系统中提取诊断信息，并输出所述诊断信息还包括：在未提取到完整的诊断数据时，在内容管理系统中，根据所确定的识别信息，检索相应的诊断文献；以及根据诊断文献生成诊断信息，并输出诊断信息。
在一些实施例中，根据所确定的识别信息，在内容管理系统中提取诊断信息，并输出所述诊断信息包括：在内容管理系统中，按照预设输出格式生成诊断信息，并输出诊断信息。
在一些实施例中，诊断数据包括诊断概要数据，所述诊断概要数据包括与预设输出字段中的病症名称字段相应的病症名称以及与预设输出字段中的诊断摘要字段相应的诊断摘要中的至少一者。
在一些实施例中，诊断数据包括诊断详细数据，所述诊断详细数据包括与预设输出字段中的症状分析字段相应的症状分析、与预设输出字段中的解决方案字段相应的解决方案以及与预设输出字段中的预防措施字段相应的预防措施中的至少一者。
在一些实施例中，根据植物图像确定识别信息包括：在当前诊断模式为被动诊断模式时，根据植物图像确定候选物种和与至少部分候选物种对应的候选病症；针对具有对应的候选病症的候选物种，根据第一预设条件对所述候选物种的候选病症执行筛除；以及在执行筛除后，如果存在剩余病症，则将剩余病症作为病症信息，将与剩余病症对应的候选物种作为物种信息。
在一些实施例中，根据植物图像确定候选物种和与至少部分候选物种对应的候选病症包括：判断候选物种是否在预设物种白名单中；当所述候选物种在所述预设物种白名单中时，根据植物图像和所述候选物种确定与所述候选物种对应的候选病症；当所述候选物种不在所述预设物种白名单中时，不 针对所述候选物种确定与所述候选物种对应的候选病症。
在一些实施例中，根据第一预设条件对所述候选物种的候选病症执行筛除包括：当存在与同一个候选物种对应的至少两种候选病症时，按照候选病症的诊断置信度由高到低的顺序，根据第一预设条件对所述候选物种的候选病症执行筛除，直至筛选出剩余病症或者筛除了与所述候选物种对应的所有候选病症。
在一些实施例中，根据第一预设条件对所述候选物种的候选病症执行筛除包括：判断所述候选物种是否在预设物种白名单中；当所述候选物种不在所述预设物种白名单中时，筛除与所述候选物种对应的候选病症。
在一些实施例中，根据第一预设条件对所述候选物种的候选病症执行筛除包括：比较候选病症的诊断置信度与第一预设置信度；当所述候选病症的诊断置信度小于所述第一预设置信度时，筛除所述候选病症。
在一些实施例中，根据第一预设条件对所述候选物种的候选病症执行筛除包括：比较候选病症的诊断准确度与预设准确度；当所述候选病症的诊断准确度小于所述预设准确度时，筛除所述候选病症。
在一些实施例中，根据第一预设条件对所述候选物种的候选病症执行筛除包括：判断所述候选物种是否在与候选病症对应的第一候选物种黑名单中；当所述候选物种在所述第一候选物种黑名单中时，筛除所述候选病症。
在一些实施例中，根据植物图像确定识别信息还包括：在当前诊断模式为主动诊断模式时，根据植物图像确定候选物种和与所述候选物种对应的候选病症信息，其中，候选病症信息包括候选病症或未检测到候选病症；根据第二预设条件从候选病症信息中筛选出病症信息。
在一些实施例中，根据第二预设条件从候选病症信息中筛选出病症信息包括：比较所述候选物种的物种置信度和第二预设置信度，以及比较与所述候选物种对应的候选病症信息的诊断置信度和第三预设置信度；在所述候选物种的物种置信度大于或等于所述第二预设置信度，且所述候选病症信息的诊断置信度大于或等于所述第三预设置信度时，将所述候选物种筛选为第一待定物种，以及将所述候选病症信息筛选为第一待定病症信息。
在一些实施例中，根据第二预设条件从候选病症信息中筛选出病症信息包括：判断第一待定物种是否在与第一待定病症信息对应的第二候选物种黑名单中；当第一待定物种在所述第二候选物种黑名单中时，筛除所述第一待定病症信息。
在一些实施例中，植物图像至少有两幅，且根据每幅植物图像确定有候选物种和与候选物种对应的候选病症信息；根据第二预设条件从候选病症信息中筛选出病症信息包括：分别针对每幅植物图像，将与植物图像对应的、具有最大诊断置信度的第一待定病症信息筛选为植物图像的第二待定病症信息；根据第三预设条件从所有植物图像的第二待定病症信息中筛选出病症信息。
在一些实施例中，根据植物图像确定识别信息还包括：在当前诊断模式为主动诊断模式时，根据多幅植物图像分别确定与每幅植物图像对应的候选物种；分别针对每幅植物图像，将具有最大物种置信度的候选物种筛选为与该植物图像对应的第二待定物种；在与各植物图像对应的第二待定物种中，将具有最大数目的第二待定物种筛选为物种信息，或者将具有最大数目的且具有最大物种置信度的第二待定物种筛选为物种信息；确定与所述物种信息对应的病症信息。
在一些实施例中，根据植物图像确定识别信息包括：利用预先训练好的物种识别模型，根据植物图像确定物种信息；其中，所述物种识别模型为神经网络模型。
在一些实施例中，根据植物图像确定识别信息包括：利用预先训练好的病症诊断模型，根据植物图像确定病症信息；其中，所述病症诊断模型为神经网络模型。
根据本公开的另一方面，提出了一种植物病症诊断系统，所述植物病症诊断系统包括处理器和存储器，所述存储器上存储有指令，当所述指令被所述处理器执行时，实现如上所述的用于植物病症诊断的方法的步骤。
根据本公开的又一方面，提出了一种计算机可读存储介质，所述计算机可读存储介质上存储有指令，当所述指令被执行时，实现如上所述的用于植物病症诊断的方法的步骤。
根据本公开的再一方面，提出了一种计算机程序产品，所述计算机程序产品包括指令，当所述指令被所述处理器执行时，实现如上所述的用于植物病症诊断的方法的步骤。
通过以下参照附图对本公开的示例性实施例的详细描述，本公开的其它特征及其优点将会变得更为清楚。
附图说明
构成说明书的一部分的附图描述了本公开的实施例，并且连同说明书一起用于解释本公开的原理。
参照附图，根据下面的详细描述，可以更加清楚地理解本公开，其中：
图1示出了根据本公开的一示例性实施例的用于植物病症诊断的方法的流程示意图；
图2示出了根据本公开的另一示例性实施例的用于植物病症诊断的方法的流程示意图；
图3示出了根据本公开的又一示例性实施例的用于植物病症诊断的方法的流程示意图；
图4示出了根据本公开的一具体示例的步骤S300的流程示意图；
图5示出了根据本公开的一具体示例的诊断卡片的示意图；
图6示出了根据本公开的一具体示例的所显示的诊断信息的示意图；
图7示出了根据本公开的示例性实施例的植物病症诊断系统的框图。
注意，在以下说明的实施方式中，有时在不同的附图之间共同使用同一附图标记来表示相同部分或具有相同功能的部分，而省略其重复说明。在一些情况中，使用相似的标号和字母表示类似项，因此，一旦某一项在一个附图中被定义，则在随后的附图中不需要对其进行进一步讨论。
为了便于理解，在附图等中所示的各结构的位置、尺寸及范围等有时不表示实际的位置、尺寸及范围等。因此，本公开并不限于附图等所公开的位置、尺寸及范围等。
具体实施方式
下面将参照附图来详细描述本公开的各种示例性实施例。应注意到：除非另外具体说明，否则在这些实施例中阐述的部件和步骤的相对布置、数字表达式和数值不限制本公开的范围。
以下对至少一个示例性实施例的描述实际上仅仅是说明性的，决不作为对本公开及其应用或使用的任何限制。也就是说，本文中的结构及方法是以示例性的方式示出，来说明本公开中的结构和方法的不同实施例。然而，本领域技术人员将会理解，它们仅仅说明可以用来实施的本公开的示例性方式，而不是穷尽的方式。此外，附图不必按比例绘制，一些特征可能被放大以示出具体组件的细节。
对于相关领域普通技术人员已知的技术、方法和设备可能不作详细讨论，但在适当情况下，所述技术、方法和设备应当被视为授权说明书的一部分。
在这里示出和讨论的所有示例中，任何具体值应被解释为仅仅是示例性的，而不是作为限制。因此，示例性实施例的其它示例可以具有不同的值。
图1示出了根据本公开的一示例性实施例的用于植物病症诊断的方法的流程示意图，该方法可以在例如手机、平板电脑等智能终端上安装的应用程序(app)中实现。如图1所示，该方法可以包括：步骤S100，获取植物图像。
在一些示例中，可以直接获取由用户上传的植物图像。在另一些示例中，可以在接收到用户指令后，生成并输出相应的提示信息，以提示用户上传植物图像。进一步的，在提示信息中，还可以包括对植物图像的具体要求，例如提示用户上传整株植物的图像，植物的茎、叶等部位的局部图像，或者有明显病变的部位的局部图像等。在这种情况下，还可以对多幅植物图像进行标记等预处理，例如分别标记出整株植物图像、局部植物图像(包括标记出在该植物图像中的植物的部位)等，以便更好地识别物种信息和/或病症信息。
步骤S200，根据植物图像确定识别信息，其中，识别信息包括物种信息和病症信息中的至少一种。
具体而言，对植物图像的识别可以包括对物种和病症中的至少一者的识别。其中，对病症的诊断可以在例如被动诊断模式或主动诊断模式下进行。在被动诊断模式下，可以在仅满足一定的预设条件时，根据植物图像来确定 物种信息和病症信息两者，而在不满足上述预设条件的情况下，可以根据植物图像确定仅物种信息。在主动诊断模式下，可以根据植物图像来确定物种信息和病症信息两者。还需要注意的是，本文中所述的病症信息也可以包括指示植物没有遭受任何疾病的相关信息。
在一些实施例中，如图2和图3所示，根据植物图像确定识别信息可以包括：
步骤S211，在当前诊断模式为被动诊断模式时，根据植物图像确定候选物种和与至少部分候选物种对应的候选病症。
在确定候选物种和与候选物种对应的候选病症时，可以根据每幅植物图像来确定与这幅植物图像对应的候选物种和候选病症；也可以根据多幅相关联的植物图像来确定与这些植物图像对应的候选物种和候选病症。
对于一幅植物图像而言，有可能存在多个候选物种，而对每一个候选物种而言，可能存在多种可能的候选病症。在一些实施例中，可以针对每个候选物种都确定对应的候选病症。在另一些实施例中，也可以只针对部分候选物种来确定对应的候选病症，以简化处理。
例如，在一些示例中，可以只针对部分物种置信度较高的候选物种进行病症的诊断，以生成与这些候选物种对应的候选病症，等待进一步的筛选。其中，物种置信度是指该植物图像对应的物种是该候选物种的概率。
或者，在另一些示例中，根据植物图像确定候选物种和与至少部分候选物种对应的候选病症可以包括：
判断候选物种是否在预设物种白名单中；
当候选物种在预设物种白名单中时，根据植物图像和候选物种确定与候选物种对应的候选病症；以及
当候选物种不在预设物种白名单中时，不针对候选物种确定与候选物种对应的候选病症。
其中，被包括在预设物种白名单中的物种一般为常见物种或重要物种，并且对这些物种的病症的诊断一般具有较高的准确性和可靠性。也就是说，可以只针对这些物种来确定病症，从而降低处理难度，并且可以避免输出不准确、不可靠或不重要的病症给用户。
可以利用预先训练好的物种识别模型，来根据植物图像确定候选物种或物种信息。其中，物种识别模型可以是神经网络模型，具体可以是卷积神经网络模型或残差网络模型。
卷积神经网络模型为深度前馈神经网络，其利用卷积核扫描植物图像，提取出植物图像中待识别的特征，进而对植物待识别的特征进行识别。另外，在对植物图像进行识别的过程中，可以直接将原始的植物图像输入卷积神经网络模型，而无需对植物图像进行预处理。卷积神经网络模型相比于其他的识别模型，具备更高的识别准确率以及识别效率。
残差网络模型相比于卷积神经网络模型多了恒等映射层，可以避免随着 网络深度(网络中叠层的数量)的增加而导致的准确率饱和、甚至下降的现象。残差网络模型中恒等映射层的恒等映射函数需要满足：恒等映射函数与残差网络模型的输入之和等于残差网络模型的输出。引入恒等映射以后，残差网络模型对输出的变化更加明显，因此可以大大提高植物物种的识别准确率和识别效率。
在一些实施例中，训练物种识别模型可以包括：
获取具有第一预设数量的标注有物种的植物图像的第一样本集；
从第一样本集中确定第一比例的植物图像作为第一训练集；
利用第一训练集训练物种识别模型；以及
在第一训练准确率大于或者等于第一预设准确率时训练结束，得到训练后的物种识别模型。
具体地，在第一样本集中，可以包括大量的植物图像，并且每幅植物图像都对应标注有物种。将植物图像输入物种识别模型以产生输出的物种，然后根据输出的物种和标注的物种之间的比较结果，可以对物种识别模型中的相关参数进行调节，即对物种识别模型进行训练，直至物种识别模型的第一训练准确率大于或者等于第一预设准确率时训练结束，从而得到训练后的物种识别模型。根据一幅植物图像，物种识别模型也可以输出多个候选物种，其中每个候选物种可以具有其相应的物种置信度，以待进一步的分析筛选。
进一步的，还可以对训练得到的物种识别模型进行测试，具体可以包括：
从第一样本集中确定第二比例的植物图像作为第一测试集；
利用第一测试集确定训练后的物种识别模型的第一模型准确率；以及
在第一模型准确率小于第二预设准确率时，调整第一训练集和/或物种识别模型进行重新训练。
一般情况下，第一测试集和第一训练集中的植物图像并不完全相同，因而可以用第一测试集来测试物种识别模型是否对第一训练集之外的植物图像也有很好的识别效果。在测试过程中，通过比较根据第一测试集中的植物图像所产生的输出的物种和标注的物种，来计算物种识别模型的第一模型准确率。在一些示例中，第一模型准确率的计算方法可以与第一训练准确率的计算方法相同。当测试得到的第一模型准确率小于第二预设准确率时，表明物种识别模型的识别效果还不够好，因而可以调整第一训练集，例如可以增加第一训练集中的标注有物种的植物图像的数量，或者调整物种识别模型本身，或者对上述两者均进行调整，然后重新训练物种识别模型来改善其识别效果。在一些实施例中，第二预设准确率可以被设置为等于第一预设准确率。
同理，可以利用预先训练好的病症诊断模型，来根据植物图像确定候选病症或病症信息。需要注意的是，病症信息可以包括候选病症或未检测到候选病症。病症诊断模型可以是神经网络模型，具体可以是卷积神经网络模型或残差网络模型。
在一些实施例中，训练病症诊断模型可以包括：
获取第二预设数量的标注有病症信息的植物图像的第二样本集；
从第二样本集中确定第三比例的植物图像作为第二训练集；
利用第二训练集训练病症诊断模型；以及
在第二训练准确率大于或者等于第三预设准确率时训练结束，得到训练后的病症诊断模型。
具体地，在第二样本集中，可以包括大量的植物图像，并且每幅植物图像都对应标注有病症信息，该病症信息例如可以是这幅植物图像中的植物所患的病症，或者是与健康的植物对应的未检测到病症。该第二样本集中的植物图像可以与第一样本集中的植物图像有至少部分相同。将植物图像输入病症诊断模型以产生输出的病症信息，然后根据输出的病症信息和标注的病症信息之间的比较结果，可以对病症诊断模型中的相关参数进行调节，即对病症诊断模型进行训练，直至病症诊断模型的第二训练准确率大于或者等于第三预设准确率时训练结束，从而得到训练后的病症诊断模型。根据一幅植物图像，病症诊断模型可以输出多个候选病症信息，其中每个候选病症信息可以具有其相应的诊断置信度，以待进一步的分析筛选。其中，诊断置信度是指与植物图像对应的病症信息为该候选病症信息的概率。
进一步的，还可以对病症诊断模型进行测试，具体可以包括：
从第二样本集中确定第四比例的植物图像作为第二测试集；
利用第二测试集确定训练后的病症诊断模型的第二模型准确率；以及
在第二模型准确率小于第四预设准确率时，调整第二训练集和/或病症诊断模型进行重新训练。
一般情况下，第二测试集和第二训练集中的植物图像并不完全相同，因而可以用第二测试集来测试病症诊断模型是否对第二训练集之外的植物图像也有很好的诊断效果。在测试过程中，通过比较根据第二测试集中的植物图像所产生的输出的病症信息和标注的病症信息，来计算病症诊断模型的第二模型准确率。在一些示例中，第二模型准确率的计算方法可以与第二训练准确率的计算方法相同。当测试得到的第二模型准确率小于第四预设准确率时，表明病症诊断模型的诊断效果还不够好，因而可以调整第二训练集，具体例如可以增加第二训练集中的标注有病症信息的植物图像的数量，或者调整病症诊断模型本身，或者对上述两者均进行调整，然后重新训练病症诊断模型来改善其诊断效果。在一些实施例中，第四预设准确率可以被设置为等于第三预设准确率。当然，在一些实施例中，对物种和病症的识别和诊断也可以由同一个预先训练好的模型来实现，即该模型可以将上述物种识别模型和病症诊断模型的功能整合在一起。
返回图2和图3，根据植物图像确定物种信息和病症信息中的至少一种还可以包括：步骤S212，针对具有对应的候选病症的候选物种，根据第一预设条件对候选物种的候选病症执行筛除。
一般而言，在被动诊断模式下，用户的主要目的并非诊断病症本身，而 是例如为了确定植物的物种信息等。在这种情况下，可以只输出准确性和可靠性较高的病症，从而在帮助用户及时发现植物病症的同时，避免因输出的病症不够准确而给用户带来额外的困扰。具体地，可以根据第一预设条件来筛除候选病症中的准确性和可靠性较低的病症。
具体地，在根据植物图像确定候选物种时，可能确定有一个或多个候选物种。此时，可以针对每个候选物种确定一个或多个候选病症，也可以如上文所述的，只针对部分候选物种来确定候选病症。进一步的，对于有对应的候选病症的各个候选物种，可以根据第一预设条件来筛除对应于该候选物种的候选病症。
例如，在一具体示例中，根据某一幅或某一些植物图像，确定的候选物种有物种1、物种2和物种3。物种1、物种2和物种3是按照物种置信度由高至低的顺序排列的。例如，物种1的物种置信度为0.8，物种2的物种置信度为0.75，物种3的物种置信度为0.7。在被动诊断模式下，针对每个物种来根据第一预设条件确定可以输出的病症。而在对某一个物种来筛除候选病症时，可以按照该物种的候选病症的诊断置信度由高到低的顺序，根据第一预设条件逐条确定是否筛除某一候选病症。例如，与物种2对应的候选病症有3种，而与物种3对应的候选病症有2种。如果病症2-1的诊断置信度为95％、病症2-2的诊断置信度为90％、以及病症2-3的诊断置信度为82％，那么，在针对物种2进行筛除的过程中，可以按照诊断置信度由高到低的顺序，按照病症2-1、病症2-2和病症2-3的顺序进行筛除。如果在对病症2-1进行筛除后，没有找到可以用于输出的剩余病症，则继续对病症2-2进行筛除。如果在对病症2-1进行筛除后，已经找到了可以用于输出的剩余病症，那么，可以不再继续对病症2-2进行筛除，以简化整个处理过程。当然，在一些情况下，也可能在执行对所有候选病症的筛除后，仍然没有找到可以用于输出的剩余病症，则停止筛除，在后续步骤中也不输出任何病症。
在筛除过程中，所涉及的第一预设条件可能与多种因素有关，这些因素例如本次诊断的诊断置信度、候选物种的种类、对某一类病症的诊断准确度以及候选物种和候选病症之间的匹配程度等当中的一个或多个。
在一实施例中，根据第一预设条件对候选物种的候选病症执行筛除可以包括：判断候选物种是否在预设物种白名单中；以及当候选物种不在预设物种白名单中时，筛除与候选物种对应的候选病症。
具体地，可以在确定候选物种和候选病症之后，首先根据预设物种白名单来执行筛除，以减少后续筛除时所要处理的数据的量。其中，被包括在预设物种白名单中的物种一般为常见物种或重要物种，并且对这些物种的病症的诊断一般具有较高的准确性和可靠性。在本实施例中，只有与这些物种对应的病症才有可能不被筛除而在后续步骤中被输出，从而尽可能避免输出不准确、不可靠或不重要的病症给用户，以避免引起用户额外的困扰。
在一实施例中，根据第一预设条件对候选物种的候选病症执行筛除可以 包括：比较候选病症的诊断置信度与第一预设置信度；以及当候选病症的诊断置信度小于第一预设置信度时，筛除候选病症。
其中，诊断置信度可以表征在单次诊断过程中，所获得的病症的可靠性。在一具体示例中，第一预设置信度可以设为70％。也就是说，当候选病症的诊断置信度小于70％时，该候选病症将被筛除而不会被输出，以避免输出的诊断信息与实际情况不符而给用户带来困扰。
在一实施例中，根据第一预设条件对候选物种的候选病症执行筛除可以包括：比较候选病症的诊断准确度与预设准确度；以及当候选病症的诊断准确度小于预设准确度时，筛除候选病症。
其中，诊断准确度反映了整体上识别某一特定类型的病症的准确性。诊断准确度可以根据在一定的诊断总次数中，正确诊断的次数与总次数的比值而得到。对于一些诊断难度较大的病症而言，其诊断准确度往往较低，那么通过筛除与这些病症相关的候选病症，可以尽可能避免输出不准确的病症。
在一实施例中，根据第一预设条件对候选物种的候选病症执行筛除可以包括：判断候选物种是否在与候选病症对应的第一候选物种黑名单中；以及当候选物种在第一候选物种黑名单中时，筛除候选病症。
对于某一特定物种而言，它可能根本不会或只有很小的概率遭遇某些特定类型的病症。因此，可以根据这样的物种与病症之间的相互排斥的关系，预先设置相应的候选物种黑名单，从而对候选病症进行筛除，以提高输出的准确性和可靠性。
需要注意的是，上述关于如何筛除满足第一预设条件的候选病症的具体方法可以相互结合。例如，在一具体示例中，只要候选病症满足诊断置信度小于第一预设置信度、候选物种不在预设物种白名单中、诊断准确度小于预设准确度以及候选物种在与候选病症对应的第一候选物种黑名单中的任何一个条件，该候选病症就会被排除。
如图2和图3所示，根据植物图像确定识别信息还可以包括：
步骤S213，在执行筛除后，如果存在剩余病症，则将剩余病症作为病症信息，将与剩余病症对应的候选物种作为物种信息。
在筛除满足第一预设条件的候选病症后，可能存在以下几种情况：
(1)在筛除后，仅存在一种剩余病症，那么，可以将该剩余病症作为病症信息，将与剩余病症对应的候选物种作为物种信息，以待后续处理。
(2)在筛除后，存在至少两种剩余病症。那么，可以将所有的剩余病症包含在病症信息中，将与每种剩余病症分别对应的候选物种包含在物种信息中，以待后续处理；或者，按照剩余病症中的诊断置信度由高到低的顺序，或者按照其它的顺序，将其中的一条或几条病症包含在病症信息中，并将与包含在病症信息中的剩余病症对应的一个或多个候选物种包含在物种信息中，以待后续处理。
(3)在筛除后，没有筛选出任何剩余病症。那么，考虑到当前识别模式 为被动识别模式，病症信息可以为空(或植物为健康)，相应地，在后续步骤中，可以不输出任何病症，或者在症状字段(Symptom)中填充内容“健康(Health)”输出，以免给用户带来额外的困扰。
在本公开的示例性实施例中，如图2所示，根据植物图像确定识别信息还可以包括：
步骤S221，在当前诊断模式为主动诊断模式时，根据植物图像确定候选物种和与候选物种对应的候选病症信息。
一般而言，在主动诊断模式下，用户可能已经发现了植物的问题，而希望进行确诊。因此，在主动模式下，可以根据植物图像确定候选物种和与各个候选物种对应的候选病症信息，以尽可能全面地获取植物的健康状态，供用户进行进一步的分析和处理。其中，候选病症信息可以包括候选病症或未检测到候选病症。
其中，候选物种可以通过上文所述的物种识别模型来确定，而候选病症信息可以通过上文所述的病症诊断模型来确定，在此不再赘述。
返回图2，根据植物图像确定识别信息还可以包括：
步骤S222，根据第二预设条件从候选病症信息中筛选出病症信息。
相比于被动诊断模式，在主动诊断模式下，可以将更多的候选病症信息包含在最终的病症信息中以供用户参考，同时可以适当降低对所生成的病症信息的准确性和可靠性的要求。
在主动诊断模式下，所涉及的候选物种和对应的候选病症信息的数据量可能较大，为了简化处理，根据第二预设条件从候选病症信息中筛选出病症信息可以包括：
比较候选物种的物种置信度和第二预设置信度，以及比较与候选物种对应的候选病症信息的诊断置信度和第三预设置信度；以及
在候选物种的物种置信度大于或等于第二预设置信度，且候选病症信息的诊断置信度大于或等于第三预设置信度时，将候选物种筛选为第一待定物种，以及将候选病症信息筛选为第一待定病症信息。
也就是说，在根据第二预设条件进行筛选时，可以仅保留物种置信度大于或等于第二预设置信度，且诊断置信度大于或等于第三预设置信度的候选物种和对应的候选病症信息，以减小待处理的数据量，以及提高所生成的病症信息的准确性和可靠性。
进一步的，根据第二预设条件从候选病症信息中筛选出病症信息还可以包括：判断第一待定物种是否在与第一待定病症信息对应的第二候选物种黑名单中；以及当第一待定物种在第二候选物种黑名单中时，筛除第一待定病症信息。
如上文所述，对于某一特定物种而言，可能根本不会或只有很小的概率遭遇某些特定种类的病症。因此，可以根据这样的物种与病症之间的相排斥的关系，预先设置与病症信息对应的候选物种黑名单，从而进一步对第一待 定病症信息进行筛除，一方面可以减少待处理的数据量、提高处理效率，另一方面也有助于进一步提高输出的准确性和可靠性。
在一些情况下，植物图像可能至少有两幅，且根据每幅植物图像都确定有候选物种和与候选物种对应的候选病症信息。那么，根据第二预设条件从候选病症信息中筛选出病症信息可以包括：分别针对每幅植物图像，将与植物图像对应的、具有最大诊断置信度的第一待定病症信息筛选为植物图像的第二待定病症信息；以及根据第三预设条件从所有植物图像的第二待定病症信息中筛选出病症信息。
也就是说，对于每幅植物图像，筛选出与该植物图像对应的具有最大诊断置信度的第二待定病症信息。需要注意的是，候选病症信息中的部分病症信息可能已经根据前面描述的主动识别模式下的实施例中的某一或某些方法被筛除，因而该第二待定病症信息的诊断置信度并不一定是与该植物图像对应的所有候选病症信息中，具有最大诊断置信度的病症信息。然后，将与每幅植物图像对应的第二待定病症信息汇总，并从中筛选出病症信息。病症信息一般是具有最大的准确性和可靠性的病症信息，具体而言，可以参考上文所描述的根据第一预设条件筛除某些病症信息的方法来筛选出病症信息。或者，也可以根据其它的预设条件来筛选出病症信息。
其中，病症信息可以是植物所患的病症及其相关信息，以帮助用户对病症进行确诊，并进一步采取相应的措施。或者，病症信息也可能是未检测到病症，即表明植物当前处于比较健康的状态。
根据本公开的又一示例性实施例，在确定病症信息时，也可以根据植物图像先唯一确定出物种信息，然后根据该物种信息来确定病症信息。具体而言，如图3所示，根据植物图像确定识别信息还可以包括：
步骤S231，在当前诊断模式为主动诊断模式时，根据多幅植物图像分别确定与每幅植物图像对应的候选物种；
步骤S232，分别针对每幅植物图像，将具有最大物种置信度的候选物种筛选为与该植物图像对应的第二待定物种；
步骤S233，在与各植物图像对应的第二待定物种中，将具有最大数目的第二待定物种筛选为物种信息，或者将具有最大数目的且具有最大物种置信度的第二待定物种筛选为物种信息。
例如，在一种情况下，如果对于植物图像A、植物图像B和植物图像C，按照物种置信度由高到低的顺序，所确定的与植物图像A对应的候选物种包括物种M、物种N和物种P，与植物图像B对应的候选物种包括物种N和物种P，与植物图像C对应的候选物种包括物种N和物种Q。那么，所筛选出的第二待定物种包括与植物图像A对应的物种M、与植物图像B对应的物种N和与植物图像C对应的物种N。可见，在与各植物图像对应的第二待定物种中，物种M的数目为1，而物种N的数目为2。在这种情况下，物种N将被作为物种信息。
在另一种情况下，如果对于植物图像A、植物图像B和植物图像C，按照物种置信度由高到低的顺序，所确定的与植物图像A对应的候选物种包括物种M、物种N和物种P，与植物图像B对应的候选物种包括物种N和物种P，与植物图像C对应的候选物种包括物种P和物种N。那么，所筛选出的第二待定物种包括与植物图像A对应的物种M、与植物图像B对应的物种N和与植物图像C对应的物种P。可见，在与各植物图像对应的第二待定物种中，物种M、物种N和物种P的数目都是1。在这种情况下，将物种M、物种N和物种P中具有最大物种置信度的物种作为物种信息。例如，如果物种M的物种置信度大于物种N的物种置信度，且物种N的物种置信度大于物种P的物种置信度，那么物种M将被作为物种信息。
返回图3，根据植物图像确定识别信息还可以包括：
步骤S234，确定与物种信息对应的病症信息。
也就是说，在确定了结果物种的前提下，进一步确定与结果物种对应的病症信息。具体而言，在确定物种信息之后，可以基于该物种信息，利用病症诊断模型来根据植物图像确定候选病症信息，然后根据上文所述的第一预设条件、第二预设条件或第三预设条件等来对候选病症信息进行筛选，从而得到用于输出的病症信息。
其中，病症信息可以是植物所患的病症及其相关信息，以帮助用户对病症进行确诊，并进一步采取相应的措施。或者，病症信息也可能是未检测到病症，即表明植物当前处于比较健康的状态。
返回图1，用于植物病症诊断的方法还可以包括：步骤S300，根据所确定的物种信息和病症信息中的至少一种，在内容管理系统中提取诊断信息，并输出所述诊断信息。
其中，内容管理系统(CMS)可以是一种位于WEB前端和后端的系统或流程之间的软件系统。可以使用内容管理系统来对例如文本文件、图片、数据库中的数据、表格等内容进行提交、修改、发布等。内容管理系统还可以提供内容抓取工具，自动抓取来自第三方的例如文本文件、HTML网页、Web服务、数据库等的内容，并经分析处理后放到该内容管理系统自身的相应的内容库中。内容管理系统也可以辅助WEB前端将内容以个性化的方式提供给用户，即提供个性化的门户框架，以基于WEB技术将内容更好地推送给用户。在本公开的实施例中的内容管理系统中，可以存储有对植物及其病症的描述性内容，这些描述性内容可以是文字的或者是图片的，例如可以包括各种字段、文章等，从而使得用户能够在从内容管理系统中所提取并输出的诊断信息中获得关于植物及其病症的介绍，例如有趣的故事、植物的用途、养护方法和对病症的描述等。
与每种物种信息一一对应的可以包括物种名称(UID1)，以区分不同的物种。类似地，与每种病症信息一一对应的可以包括病症名称(UID2或ComnonName)，以区分不同的病症。在内容管理系统中提取相关的诊断信息 时，可以根据UID1和UID2来进行检索。当内容管理系统中预先存储了大量的数据时，可以涵盖大多数诊断情形，为用户提供相应的诊断信息。
基于内容管理系统，可以以一个物种对应一张卡片的形式，将多个物种的相关信息输出给用户。用户可以在交互界面上通过滑动卡片等方式来切换显示各个物种及其相关信息。
在一些实施例中，针对不同的植物图像，在所确定的识别信息相同的情况下，至少部分诊断信息是可以随着不同的植物图像而改变的。这样，即使所得到的识别信息是相同的，但是输出的诊断信息可以根据用户输入的植物图像而产生适应性的变化，实现了更加灵活的输出，有助于使得输出的诊断信息与用户的输入相匹配，从而改善用户体验，减少因输入输出的不匹配给用户带来的困惑。
在一些实施例中，如图4所示，根据所确定的识别信息，在内容管理系统中提取诊断信息，并输出诊断信息可以包括：步骤S310，在内容管理系统中，根据所确定的识别信息，按照预设输出字段提取相应的诊断数据；以及步骤S320，在提取到完整的诊断数据时，根据诊断数据生成诊断信息，并输出诊断信息。
在一些实施例中，预设输出字段可以由用户根据其自身需求通过交互界面来设置，或者，预设输出字段也可以是相对固定的若干个字段。在内容管理系统中，根据所确定的识别信息提取到的相应的诊断数据可以被填充在具有预设输出格式的相应的模板中，以形成诊断信息。诊断信息可以以卡片、标签等形式被组织或布局，并输出给用户。用户可以在交互界面上通过选择、切换或移动卡片、标签等，根据需要读取某个或某些卡片、标签中所包含的具体内容，从而获得相关信息。
在一些实施例中，诊断数据可以包括诊断概要数据和/或诊断详细数据。在诊断概要数据和诊断详细数据中，可以设置不同的字段，以将从内容管理系统中提取到的数据存储在相应的字段中。
在一些实施例中，诊断概要数据可以包括与预设输出字段中的病症名称字段相应的病症名称以及与预设输出字段中的诊断摘要字段相应的诊断摘要中的至少一者。在一些实施例中，病症名称和诊断摘要可以被显示为如图5所示的诊断卡片。其中，“Black Spot”为病症名称，而“Your plant get black spot on the leaves.It is due to xxx,xxx,xxxx,xxxx,xxxxx,xxxxx,xxx,xxx,xxxx,xxxx,xxxxx,xxxxx”为诊断摘要。此外，在图5所示的诊断卡片中还可以包括其他的按钮(例如“Check for causes”)，以链接到其他更详细的诊断信息。
在一些实施例中，诊断详细数据可以包括与预设输出字段中的症状分析字段相应的症状分析、与预设输出字段中的解决方案字段相应的解决方案以及与预设输出字段中的预防措施字段相应的预防措施中的至少一者。通过将症状分析、解决方案以及预防措施存储在不同的字段中，可以基于内容管理系统方便地生成诊断信息，也方便用户的查看。
如图6所示，为一具体示例中的所显示的诊断信息的示意图。其中，来自用户的植物图像被显示在界面的最上方，包括诊断概要数据的诊断卡片可以位于植物图像下方，而更进一步的诊断详细数据可以位于诊断卡片的下方。
在一些实施例中，在内容管理系统中可能无法提取到完整的诊断数据,此时，可以按照旧有的显示固定内容的方式来输出根据植物图像确定的识别信息，或者根据所确定的识别信息，从相关文献中获取诊断信息。
例如，如图4所示，根据所确定的识别信息，在内容管理系统中提取诊断信息，并输出诊断信息还可以包括：步骤S331，在未提取到完整的诊断数据时，在内容管理系统中，根据所确定的识别信息，检索相应的诊断文献；步骤S332，根据诊断文献生成诊断信息，并输出诊断信息。
如图6所示，在可以跟随植物图像而适应性改变的至少部分诊断信息中，可以包括参考图(位于图6下方的图)。参考图至少与病症信息对应，且参考图与植物图像相似。这样，所输出的诊断信息可以不再是固定的，而是可以根据用户输入的植物图像来对输出的诊断信息中的用来解释说明的相关图片进行替换，使这些用于解释说明的图片与用户拍摄的植物图像更加相似，从而不会让用户觉得输出的诊断信息中的图与自己拍摄的植物图像有着过大的差别，避免引起用户的困扰，以提高用户体验。
在一些实施例中，根据所确定的识别信息，在内容管理系统中提取诊断信息，并输出诊断信息可以包括：在内容管理系统中，根据病症信息确定对应的候选参考图库；在候选参考图库中，基于与植物图像的相似度和/或与物种信息的匹配度，确定被提取的一幅或多幅参考图以及与一幅或多幅参考图中的每幅参考图对应的优先级；以及输出一幅或多幅参考图，使得一幅或多幅参考图按照优先级由高至低的顺序被依次排列。
其中，内容管理系统中的每幅参考图可以被标记有对应的物种信息的UID1(UID1可以包括种、变种、品种、属、科等)和病症信息的UID2。基于UID1和UID2，可以对参考图进行分类、筛选等。例如，可以根据UID2，将与每种病症信息对应的一幅或多幅参考图组成分别与相应的病症信息对应的候选参考图库。而在参考图库中选取所需的参考图时，可以根据每幅参考图所标记的UID1，确定该参考图相应的植物的种类。通过显示参考图，可以帮助用户更好地识别植物的病症，尤其是在用户所拍摄的植物图像不清楚或者拍摄部位不好的情况下。
在通常情况下，与植物图像的相似度越高，与物种信息的匹配度越高的参考图，将具有更高的优先级。优先级较高的参考图可以被优先显示或者排列在被显示的多幅参考图的前面的位置处，以方便用户的查看。
在一些实施例中，在候选参考图库中，基于与所述植物图像的相似度和/或与所述物种信息的匹配度，确定被提取的一幅或多幅参考图以及与所述一幅或多幅参考图中的每幅参考图对应的优先级可以包括：
将候选参考图库中的与所述植物图像的相似度最高的预设数量的候选参 考图作为第一参考图集，并将候选参考图库中的所有其他候选参考图作为第二参考图集；
在第一参考图集中，确定在第一物种分类级别上与所述物种信息匹配的第一参考图，其中，所确定的第一参考图具有第一优先级；
在第二参考图集中，确定在第一物种分类级别上与所述物种信息匹配的第二参考图，其中，所确定的第二参考图具有第二优先级，且第二优先级低于第一优先级；
在第一参考图集中，确定在高于第一物种分类级别的第二物种分类级别上与所述物种信息匹配的第三参考图，其中，所确定的第三参考图具有第三优先级，且第三优先级低于第二优先级；以及
在第二参考图集中，确定在第二物种分类级别上与所述物种信息匹配的第四参考图，其中，所确定的第四参考图具有第四优先级，且第四优先级低于第三优先级。
在一具体示例中，可以从候选参考图库中确定与用户上传的植物图像的图像特征最接近的六张候选参考图作为第一图集，而将候选参考图库中的其他候选参考图作为第二图集。首先，在第一图集中，查找与植物图像中的植物的种匹配的第一参考图，该第一参考图具有最高的第一优先级；然后，在第二图集中，查找与植物图像中的植物的种匹配的第二参考图，该第二参考图具有较第一优先级低的第二优先级；然后，在第一图集中，查找与植物图像中的植物的属匹配的第三参考图，该第三参考图具有较第二优先级低的第三优先级；然后，在第二图集中，查找与植物图像中的植物的属匹配的第四参考图，该第四参考图具有较第三优先级低的第四优先级；然后，在第一图集中，查找与植物图像中的植物的科匹配的第五参考图，该第五参考图具有较第四优先级低的第五优先级；最后，在第二图集中，查找与植物图像中的植物的科匹配的第六参考图，该第六参考图具有最低的第六优先级。在显示参考图时，可以按照第一参考图、第二参考图、第三参考图、第四参考图、第五参考图、第六参考图的顺序进行显示，其中，第一参考图被显示在最显眼的位置处。
在一些实施例中，在候选参考图库中，基于与所述植物图像的相似度和/或与所述物种信息的匹配度，确定被提取的一幅或多幅参考图以及与一幅或多幅参考图中的每幅参考图对应的优先级还可以包括：
当在第一参考图集和第二参考图集中均未能确定在低于或等于预设物种分类级别的物种分类级别上与所述物种信息匹配的参考图时，将与病症信息对应的预设默认图确定为参考图。例如，如果查找到科这一分类级别时，仍然不能在第一图集或第二图集中找到匹配的结果参考图，那么可以将参考图库中的预设默认图作为参考图，不再进一步查找。
通常情况下，所显示的诊断信息中的图片的显示比例在3：2至1：1之间，从而具有较好的显示效果。然而，从候选参考图库中筛选出来的参考图 的比例可能不适合上述显示比例。通常，可以对这样的图进行拉伸或裁切，以适应显示比例。然而，考虑到对参考图进行拉伸时，可能导致某些病症的特征发生变形，不利于用户很好地识别病症，因此在本公开的示例性实施例中，可以采用裁切的方式来处理参考图。具体而言，用于植物病症诊断的方法还可以包括：对形成参考图的原始图的边缘区域进行裁切，以使得裁切后所得的参考图的比例与预设显示比例相符，且参考图中的与病症信息对应的图像特征位于参考图的中部区域内。
具体而言，可以基于区域识别模型，在为内容管理系统选取参考图素材时，就将与病症信息对应的图像特征位于边缘区域的图片去除或忽略，这些图片将不被收录在内容管理系统中。或者，可以在将图片存储在内容管理系统中时，对其进行例如裁切等处理。又或者，可以在根据植物图像确定了要被输出的参考图后，对选择内容管理系统的参考图进行例如裁切等处理后再输出。当然，在其他一些实施例中，也可以预先判定病症的特征在形成参考图的原始图中所在的位置，并在裁切过程中避开这些位置。
根据本公开的另一个方面，还提出了一种植物病症诊断系统。如图7所示，植物病症系统900可以包括处理器910和存储器920，存储器920上存储有指令，当指令被处理器910执行时，可以实现如上文所描述的用于植物病症诊断的方法中的步骤。
其中，处理器910可以根据存储在存储器920中的指令执行各种动作和处理。具体地，处理器910可以是一种集成电路芯片，具有信号的处理能力。上述处理器可以是通用处理器、数字信号处理器(DSP)、专用集成电路(ASIC)、现成可编程门阵列(FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。可以实现或者执行本公开实施例中公开的各种方法、步骤及逻辑框图。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等，可以是X86架构或者是ARM架构等。
存储器920存储有可执行指令，该指令在被处理器910执行上文所述的用于植物病症诊断的方法。存储器920可以是易失性存储器或非易失性存储器，或可包括易失性和非易失性存储器两者。非易失性存储器可以是只读存储器(ROM)、可编程只读存储器(PROM)、可擦除可编程只读存储器(EPROM)、电可擦除可编程只读存储器(EEPROM)或闪存。易失性存储器可以是随机存取存储器(RAM)，其用作外部高速缓存。通过示例性但不是限制性说明，许多形式的RAM可用，例如静态随机存取存储器(SRAM)、动态随机存取存储器(DRAM)、同步动态随机存取存储器(SDRAM)、双倍数据速率同步动态随机存取存储器(DDRSDRAM)、增强型同步动态随机存取存储器(ESDRAM)、同步连接动态随机存取存储器(SLDRAM)和直接内存总线随机存取存储器(DR RAM)。应注意，本文描述的方法的存储器旨在包括但不限于这些和任意其它适合类型的存储器。
在一些实施例中，植物病症诊断系统还可以包括内容管理系统。在本公 开的实施例中的内容管理系统中，可以存储有对植物及其病症的描述性内容，这些描述性内容可以是文字的或者是图片的，例如可以包括各种字段、文章等，从而使得用户能够在从内容管理系统中所提取并输出的诊断信息中获得关于植物及其病症的介绍，例如有趣的故事、植物的用途、养护方法和对病症的描述等。当然，内容管理系统也可以独立于植物病症诊断系统之外，植物病症诊断系统可以与内容管理系统通信地连接以获取相关内容。
根据本公开的另一个方面，提出了一种计算机可读存储介质，计算机可读存储介质上存储有指令，当指令被执行时，可以实现上文所描述的用于植物病症诊断的方法中的步骤。
类似地，本公开实施例中的计算机可读存储介质可以是易失性存储器或非易失性存储器，或可包括易失性和非易失性存储器两者。应注意，本文描述的计算机可读存储介质旨在包括但不限于这些和任意其它适合类型的存储器。
本公开也提出了一种计算机程序产品，该计算机程序产品可以包括指令，当指令被处理器执行时，可以实现如上所述的用于植物病症诊断的方法的步骤。
指令可以是将由一个或多个处理器直接地执行的任何指令集，诸如机器代码，或者间接地执行的任何指令集，诸如脚本。本文中的术语“指令”、“应用”、“过程”、“步骤”和“程序”在本文中可以互换使用。指令可以存储为目标代码格式以便由一个或多个处理器直接处理，或者存储为任何其他计算机语言，包括按需解释或提前编译的独立源代码模块的脚本或集合。指令可以包括引起诸如一个或多个处理器来充当本文中的各神经网络的指令。本文其他部分更加详细地解释了指令的功能、方法和例程。
需要说明的是，附图中的流程图和框图，图示了按照本公开各种实施例的系统、方法和计算机程序产品的可能实现的体系架构、功能和操作。在这点上，流程图或框图中的每个方框可以代表一个模块、程序段、或代码的一部分，所述模块、程序段、或代码的一部分包含一个或多个用于实现规定的逻辑功能的可执行指令。也应当注意，在有些作为替换的实现中，方框中所标注的功能也可以以不同于附图中所标注的顺序发生。例如，两个接连地表示的方框实际上可以基本并行地执行，它们有时也可以按相反的顺序执行，这依所涉及的功能而定。也要注意的是，框图和/或流程图中的每个方框、以及框图和/或流程图中的方框的组合，可以用执行规定的功能或操作的专用的基于硬件的系统来实现，或者可以用专用硬件与计算机指令的组合来实现。
一般而言，本公开的各种示例实施例可以在硬件或专用电路、软件、固件、逻辑，或其任何组合中实施。某些方面可以在硬件中实施，而其他方面可以在可以由控制器、微处理器或其他计算设备执行的固件或软件中实施。当本公开的实施例的各方面被图示或描述为框图、流程图或使用某些其他图形表示时，将理解此处描述的方框、装置、系统、技术或方法可以作为非限 制性的示例在硬件、软件、固件、专用电路或逻辑、通用硬件或控制器或其他计算设备，或其某些组合中实施。
在说明书及权利要求中的词语“前”、“后”、“顶”、“底”、“之上”、“之下”等，如果存在的话，用于描述性的目的而并不一定用于描述不变的相对位置。应当理解，这样使用的词语在适当的情况下是可互换的，使得在此所描述的本公开的实施例，例如，能够在与在此所示出的或另外描述的那些取向不同的其他取向上操作。
如在此所使用的，词语“示例性的”意指“用作示例、实例或说明”，而不是作为将被精确复制的“模型”。在此示例性描述的任意实现方式并不一定要被解释为比其它实现方式优选的或有利的。而且，本公开不受在上述技术领域、背景技术、发明内容或具体实施方式中所给出的任何所表述的或所暗示的理论所限定。
如在此所使用的，词语“基本上”意指包含由设计或制造的缺陷、器件或元件的容差、环境影响和/或其它因素所致的任意微小的变化。词语“基本上”还允许由寄生效应、噪声以及可能存在于实际的实现方式中的其它实际考虑因素所致的与完美的或理想的情形之间的差异。
另外，前面的描述可能提及了被“连接”或“耦接”在一起的元件或节点或特征。如在此所使用的，除非另外明确说明，“连接”意指一个元件/节点/特征与另一种元件/节点/特征在电学上、机械上、逻辑上或以其它方式直接地连接(或者直接通信)。类似地，除非另外明确说明，“耦接”意指一个元件/节点/特征可以与另一元件/节点/特征以直接的或间接的方式在机械上、电学上、逻辑上或以其它方式连结以允许相互作用，即使这两个特征可能并没有直接连接也是如此。也就是说，“耦接”意图包含元件或其它特征的直接连结和间接连结，包括利用一个或多个中间元件的连接。
另外，仅仅为了参考的目的，还可以在本文中使用“第一”、“第二”等类似术语，并且因而并非意图限定。除非上下文明确指出，否则涉及结构或元件的词语“第一”、“第二”和其它此类数字词语并没有暗示顺序或次序。
还应理解，“包括/包含”一词在本文中使用时，说明存在所指出的特征、整体、步骤、操作、单元和/或组件，但是并不排除存在或增加一个或多个其它特征、整体、步骤、操作、单元和/或组件以及/或者它们的组合。
在本公开中，术语“提供”从广义上用于涵盖获得对象的所有方式，因此“提供某对象”包括但不限于“购买”、“制备/制造”、“布置/设置”、“安装/装配”、和/或“订购”对象等。
虽然已经通过示例对本公开的一些特定实施例进行了详细说明，但是本领域的技术人员应该理解，以上示例仅是为了进行说明，而不是为了限制本公开的范围。在此公开的各实施例可以任意组合，而不脱离本公开的精神和范围。本领域的技术人员还应理解，可以对实施例进行多种修改而不脱离本公开的范围和精神。本公开的范围由所附权利要求来限定。

Claims (29)

  1. 一种用于植物病症诊断的方法，其特征在于，所述方法包括：
    获取植物图像；
    根据所获取的植物图像确定识别信息，其中，所述识别信息包括物种信息和病症信息中的至少一种；以及
    根据所确定的识别信息，在内容管理系统中提取诊断信息，并输出所述诊断信息。
  2. 根据权利要求1所述的方法，其特征在于，针对不同的植物图像，在所确定的识别信息相同的情况下，至少部分诊断信息是随着不同的植物图像而改变的。
  3. 根据权利要求2所述的方法，其特征在于，所述至少部分诊断信息包括参考图，所述参考图至少与所述病症信息对应，且所述参考图与所述植物图像相似。
  4. 根据权利要求3所述的方法，其特征在于，根据所确定的识别信息，在内容管理系统中提取诊断信息，并输出所述诊断信息包括：
    在内容管理系统中，根据所述病症信息确定对应的候选参考图库；
    在所述候选参考图库中，基于与所述植物图像的相似度和/或与所述物种信息的匹配度，确定被提取的一幅或多幅参考图以及与所述一幅或多幅参考图中的每幅参考图对应的优先级；以及
    输出所述一幅或多幅参考图，使得所述一幅或多幅参考图按照优先级由高至低的顺序被依次排列。
  5. 根据权利要求4所述的方法，其特征在于，在所述候选参考图库中，基于与所述植物图像的相似度和/或与所述物种信息的匹配度，确定被提取的一幅或多幅参考图以及与所述一幅或多幅参考图中的每幅参考图对应的优先级包括：
    将所述候选参考图库中的与所述植物图像的相似度最高的预设数量的候选参考图作为第一参考图集，并将所述候选参考图库中的所有其他候选参考图作为第二参考图集；
    在所述第一参考图集中，确定在第一物种分类级别上与所述物种信息匹配的第一参考图，其中，所确定的第一参考图具有第一优先级；
    在所述第二参考图集中，确定在第一物种分类级别上与所述物种信息匹配的第二参考图，其中，所确定的第二参考图具有第二优先级，且所述第二优先级低于所述第一优先级；
    在所述第一参考图集中，确定在高于第一物种分类级别的第二物种分类级别上与所述物种信息匹配的第三参考图，其中，所确定的第三参考图具有第三优先级，且所述第三优先级低于所述第二优先级；以及
    在所述第二参考图集中，确定在第二物种分类级别上与所述物种信息匹配的第四参考图，其中，所确定的第四参考图具有第四优先级，且所述第四优先级低于所述第三优先级。
  6. 根据权利要求5所述的方法，其特征在于，在所述候选参考图库中，基于与所述植物图像的相似度和/或与所述物种信息的匹配度，确定被提取的一幅或多幅参考图以及与所述一幅或多幅参考图中的每幅参考图对应的优先级还包括：
    当在所述第一参考图集和所述第二参考图集中均未能确定在低于或等于预设物种分类级别的物种分类级别上与所述物种信息匹配的参考图时，将与所述病症信息对应的预设默认图确定为参考图。
  7. 根据权利要求3所述的方法，其特征在于，所述方法还包括：
    对形成参考图的原始图的边缘区域进行裁切，以使得裁切后所得的参考图的比例与预设显示比例相符，且所述参考图中的与所述病症信息对应的图像特征位于所述参考图的中部区域内。
  8. 根据权利要求1所述的方法，其特征在于，根据所确定的识别信息，在内容管理系统中提取诊断信息，并输出所述诊断信息包括：
    在内容管理系统中，根据所确定的识别信息，按照预设输出字段提取相应的诊断数据；
    在提取到完整的诊断数据时，根据所述诊断数据生成诊断信息，并输出诊断信息。
  9. 根据权利要求8所述的方法，其特征在于，根据所确定的识别信息，在内容管理系统中提取诊断信息，并输出所述诊断信息还包括：
    在未提取到完整的诊断数据时，在所述内容管理系统中，根据所确定的识别信息，检索相应的诊断文献；以及
    根据所述诊断文献生成诊断信息，并输出诊断信息。
  10. 根据权利要求8所述的方法，其特征在于，根据所确定的识别信息，在内容管理系统中提取诊断信息，并输出所述诊断信息包括：
    在所述内容管理系统中，按照预设输出格式生成诊断信息，并输出诊断信息。
  11. 根据权利要求8所述的方法，其特征在于，所述诊断数据包括诊断概要数据，所述诊断概要数据包括与所述预设输出字段中的病症名称字段相应的病症名称以及与所述预设输出字段中的诊断摘要字段相应的诊断摘要中的至少一者。
  12. 根据权利要求8所述的方法，其特征在于，所述诊断数据包括诊断详细数据，所述诊断详细数据包括与所述预设输出字段中的症状分析字段相应的症状分析、与所述预设输出字段中的解决方案字段相应的解决方案以及与所述预设输出字段中的预防措施字段相应的预防措施中的至少一者。
  13. 根据权利要求1所述的方法，其特征在于，根据所获取的植物图像确定识别信息包括：
    在当前诊断模式为被动诊断模式时，根据所述植物图像确定候选物种和与至少部分所述候选物种对应的候选病症；
    针对具有对应的候选病症的候选物种，根据第一预设条件对所述候选物种的候选病症执行筛除；以及
    在执行筛除后，如果存在剩余病症，则将剩余病症作为病症信息，将与所述剩余病症对应的候选物种作为物种信息。
  14. 根据权利要求13所述的方法，其特征在于，根据所述植物图像确定候选物种和与至少部分所述候选物种对应的候选病症包括：
    判断候选物种是否在预设物种白名单中；
    当所述候选物种在所述预设物种白名单中时，根据所述植物图像和所述候选物种确定与所述候选物种对应的候选病症；
    当所述候选物种不在所述预设物种白名单中时，不针对所述候选物种确定与所述候选物种对应的候选病症。
  15. 根据权利要求13所述的方法，其特征在于，根据第一预设条件对所述候选物种的候选病症执行筛除包括：
    当存在与同一个候选物种对应的至少两种候选病症时，按照所述候选病症的诊断置信度由高到低的顺序，根据第一预设条件对所述候选物种的候选病症执行筛除，直至筛选出剩余病症或者筛除了与所述候选物种对应的所有候选病症。
  16. 根据权利要求13所述的方法，其特征在于，根据第一预设条件对所述候选物种的候选病症执行筛除包括：
    判断所述候选物种是否在预设物种白名单中；
    当所述候选物种不在所述预设物种白名单中时，筛除与所述候选物种对应的候选病症。
  17. 根据权利要求13所述的方法，其特征在于，根据第一预设条件对所述候选物种的候选病症执行筛除包括：
    比较所述候选病症的诊断置信度与第一预设置信度；
    当所述候选病症的诊断置信度小于所述第一预设置信度时，筛除所述候选病症。
  18. 根据权利要求13所述的方法，其特征在于，根据第一预设条件对所述候选物种的候选病症执行筛除包括：
    比较候选病症的诊断准确度与预设准确度；
    当所述候选病症的诊断准确度小于所述预设准确度时，筛除所述候选病症。
  19. 根据权利要求13所述的方法，其特征在于，根据第一预设条件对所 述候选物种的候选病症执行筛除包括：
    判断所述候选物种是否在与候选病症对应的第一候选物种黑名单中；
    当所述候选物种在所述第一候选物种黑名单中时，筛除所述候选病症。
  20. 根据权利要求1所述的方法，其特征在于，根据所获取的植物图像确定识别信息还包括：
    在当前诊断模式为主动诊断模式时，根据所述植物图像确定候选物种和与所述候选物种对应的候选病症信息，其中，所述候选病症信息包括候选病症或未检测到候选病症；
    根据第二预设条件从所述候选病症信息中筛选出病症信息。
  21. 根据权利要求20所述的方法，其特征在于，根据第二预设条件从所述候选病症信息中筛选出病症信息包括：
    比较所述候选物种的物种置信度和第二预设置信度，以及比较与所述候选物种对应的候选病症信息的诊断置信度和第三预设置信度；
    在所述候选物种的物种置信度大于或等于所述第二预设置信度，且所述候选病症信息的诊断置信度大于或等于所述第三预设置信度时，将所述候选物种筛选为第一待定物种，以及将所述候选病症信息筛选为第一待定病症信息。
  22. 根据权利要求21所述的方法，其特征在于，根据第二预设条件从所述候选病症信息中筛选出病症信息还包括：
    判断所述第一待定物种是否在与所述第一待定病症信息对应的第二候选物种黑名单中；
    当所述第一待定物种在所述第二候选物种黑名单中时，筛除所述第一待定病症信息。
  23. 根据权利要求22所述的方法，其特征在于，所述植物图像至少有两幅，且根据每幅植物图像确定有候选物种和与所述候选物种对应的候选病症信息；
    根据第二预设条件从所述候选病症信息中筛选出病症信息还包括：
    分别针对每幅所述植物图像，将与所述植物图像对应的、具有最大诊断置信度的第一待定病症信息筛选为所述植物图像的第二待定病症信息；
    根据第三预设条件从所有植物图像的第二待定病症信息中筛选出病症信息。
  24. 根据权利要求1所述的方法，其特征在于，根据所获取的植物图像确定识别信息还包括：
    在当前诊断模式为主动诊断模式时，根据多幅所述植物图像分别确定与每幅植物图像对应的候选物种；
    分别针对每幅植物图像，将具有最大物种置信度的候选物种筛选为与该植物图像对应的第二待定物种；
    在与各植物图像对应的第二待定物种中，将具有最大数目的第二待定物种筛选为物种信息，或者将具有最大数目的且具有最大物种置信度的第二待定物种筛选为物种信息；
    确定与所述物种信息对应的病症信息。
  25. 根据权利要求1所述的方法，其特征在于，根据所获取的植物图像确定识别信息包括：
    利用预先训练好的物种识别模型，根据所述植物图像确定物种信息；
    其中，所述物种识别模型为神经网络模型。
  26. 根据权利要求1所述的方法，其特征在于，根据所获取的植物图像确定识别信息包括：
    利用预先训练好的病症诊断模型，根据所述植物图像确定病症信息；
    其中，所述病症诊断模型为神经网络模型。
  27. 一种植物病症诊断系统，其特征在于，所述植物病症诊断系统包括处理器和存储器，所述存储器上存储有指令，当所述指令被所述处理器执行时，实现如权利要求1至26中任一项所述的用于植物病症诊断的方法的步骤。
  28. 一种计算机可读存储介质，其特征在于，所述计算机可读存储介质上存储有指令，当所述指令被执行时，实现如权利要求1至26中任一项所述的用于植物病症诊断的方法的步骤。
  29. 一种计算机程序产品，其特征在于，所述计算机程序产品包括指令，当所述指令被所述处理器执行时，实现如权利要求1至26中任一项所述的用于植物病症诊断的方法的步骤。
PCT/CN2022/085172 2021-04-15 2022-04-02 用于植物病症诊断的方法和植物病症诊断系统 WO2022218185A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US18/286,754 US20240193778A1 (en) 2021-04-15 2022-04-02 Method for diagnosing plant diseases and plant disease diagnosis system

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202110403195.4 2021-04-15
CN202110403195.4A CN113096100B (zh) 2021-04-15 2021-04-15 用于植物病症诊断的方法和植物病症诊断系统

Publications (1)

Publication Number Publication Date
WO2022218185A1 true WO2022218185A1 (zh) 2022-10-20

Family

ID=76677508

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/085172 WO2022218185A1 (zh) 2021-04-15 2022-04-02 用于植物病症诊断的方法和植物病症诊断系统

Country Status (3)

Country Link
US (1) US20240193778A1 (zh)
CN (1) CN113096100B (zh)
WO (1) WO2022218185A1 (zh)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113096100B (zh) * 2021-04-15 2023-08-22 杭州睿胜软件有限公司 用于植物病症诊断的方法和植物病症诊断系统
CN113947697A (zh) * 2021-10-21 2022-01-18 杭州睿胜软件有限公司 植物病虫害诊断方法、诊断系统及可读存储介质
CN114120117A (zh) * 2021-11-19 2022-03-01 杭州睿胜软件有限公司 植物病症诊断信息的显示方法、显示系统及可读存储介质
CN115656167A (zh) * 2022-09-27 2023-01-31 杭州睿胜软件有限公司 植物诊断方法、装置及计算机可读存储介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105787446A (zh) * 2016-02-24 2016-07-20 上海劲牛信息技术有限公司 一种智慧农业病虫害远程自动诊断系统
CN109446250A (zh) * 2018-11-19 2019-03-08 天津市植物保护研究所 基于瓜类蔬菜病害典型特征的互联网蔬菜病害自助诊断方法与系统
CN110223383A (zh) * 2019-06-17 2019-09-10 重庆大学 一种基于深度图修补的植物三维重建方法及系统
CN111340070A (zh) * 2020-02-11 2020-06-26 杭州睿琪软件有限公司 植物病虫害诊断方法和系统
CN113096100A (zh) * 2021-04-15 2021-07-09 杭州睿胜软件有限公司 用于植物病症诊断的方法和植物病症诊断系统

Family Cites Families (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8301389B2 (en) * 2003-12-16 2012-10-30 Dunlap Susan C System and method for plant selection
CN102054114A (zh) * 2009-10-30 2011-05-11 上海市农业科学院 一种蔬菜病虫害诊断专家系统构建和服务方法
CN104484408A (zh) * 2014-12-16 2015-04-01 百度在线网络技术（北京）有限公司 植物信息的搜索处理方法和系统
JP6539901B2 (ja) * 2015-03-09 2019-07-10 学校法人法政大学 植物病診断システム、植物病診断方法、及びプログラム
US10028452B2 (en) * 2016-04-04 2018-07-24 Beesprout, Llc Horticultural monitoring system
CN106980754A (zh) * 2017-03-02 2017-07-25 西安慧云医疗科技有限公司 基于云心理诊断的量表生成系统
US10088816B1 (en) * 2017-06-26 2018-10-02 International Business Machines Corporation Cognitive plant clinic
GB2565211A (en) * 2017-06-26 2019-02-06 Ibm Cognitive plant clinic
CN108764183A (zh) * 2018-05-31 2018-11-06 寿光得峰生态农业有限公司 一种植物病害诊断方法、装置及存储介质
CN109166120A (zh) * 2018-09-11 2019-01-08 百度在线网络技术（北京）有限公司 用于获取信息的方法及装置
CN109271429A (zh) * 2018-11-19 2019-01-25 天津市植物保护研究所 基于豆类蔬菜病害典型特征的互联网蔬菜病害自助诊断方法与系统
JP2020149307A (ja) * 2019-03-13 2020-09-17 株式会社Ｉｎｇｅｎ 植物病名診断システム、サーバ装置及び植物の病気治療方法の決定方法
CN110992198A (zh) * 2019-10-31 2020-04-10 郑州西亚斯学院 作物病害防治方案推荐方法及装置、系统、设备和介质
CN111179216B (zh) * 2019-12-03 2023-03-28 中国地质大学（武汉） 一种基于图像处理与卷积神经网络的作物病害识别方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105787446A (zh) * 2016-02-24 2016-07-20 上海劲牛信息技术有限公司 一种智慧农业病虫害远程自动诊断系统
CN109446250A (zh) * 2018-11-19 2019-03-08 天津市植物保护研究所 基于瓜类蔬菜病害典型特征的互联网蔬菜病害自助诊断方法与系统
CN110223383A (zh) * 2019-06-17 2019-09-10 重庆大学 一种基于深度图修补的植物三维重建方法及系统
CN111340070A (zh) * 2020-02-11 2020-06-26 杭州睿琪软件有限公司 植物病虫害诊断方法和系统
CN113096100A (zh) * 2021-04-15 2021-07-09 杭州睿胜软件有限公司 用于植物病症诊断的方法和植物病症诊断系统

Also Published As

Publication number Publication date
CN113096100B (zh) 2023-08-22
CN113096100A (zh) 2021-07-09
US20240193778A1 (en) 2024-06-13

Similar Documents

Publication Publication Date Title
WO2022218185A1 (zh) 用于植物病症诊断的方法和植物病症诊断系统
RU2701995C2 (ru) Автоматическое определение набора категорий для классификации документа
US10303975B2 (en) Landmarks from digital photo collections
US11605019B2 (en) Visually guided machine-learning language model
DE102020001790A1 (de) Auf maschinellem Lernen basierende Text-in-Bild-Einbettungstechniken
US9710695B2 (en) Characterizing pathology images with statistical analysis of local neural network responses
CN111340070B (zh) 植物病虫害诊断方法和系统
Park et al. Image-based disease diagnosing and predicting of the crops through the deep learning mechanism
US11379717B2 (en) Systems and methods for classification of multi-dimensional time series of parameters
DE102019005851A1 (de) Objektdetektion in Bildern
WO2023065989A1 (zh) 植物病虫害诊断方法、诊断系统及可读存储介质
US20240037948A1 (en) Method for video moment retrieval, computer system, non-transitory computer-readable medium
US20220084679A1 (en) Deep neural network pre-training method for classifying electrocardiogram (ecg) data
WO2021004319A1 (zh) 对象识别及显示的方法及系统
CN112989990B (zh) 医疗票据识别方法、装置、设备及存储介质
US20200293528A1 (en) Systems and methods for automatically generating structured output documents based on structural rules
CN109543718B (zh) 修改病种描述的方法和装置
CN114170509A (zh) 植物识别方法、植物识别设备和植物识别系统
US20240153267A1 (en) Display method and display system for plant disease diagnosis information, and readable storage medium
CN115331071A (zh) 一种基于多尺度特征图的结核性脑膜脑炎预测方法及系统
CN111199801A (zh) 一种用于识别病历的疾病类型的模型的构建方法及应用
DE102020215928A1 (de) System und verfahren zum verbessern neuronaler satzklassifizierung
CN116469505A (zh) 数据处理方法、装置、计算机设备及可读存储介质
TWI684950B (zh) 物種數據解析方法、系統及電腦程式產品
Berg et al. Do you see what I see? Measuring the semantic differences in image‐recognition services' outputs

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22787412

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 18286754

Country of ref document: US

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 22787412

Country of ref document: EP

Kind code of ref document: A1