Satellite image data for enhancing the production of plant seeds and seedlings

Abstract

The present invention relates to a computer-implemented method for selecting at least one agricultural field for the production of crop plant seedlings by utilizing past satellite image data of at least one agricultural field from at least one previous growing season, a system configured to execute the method, a method of using past image data of an agricultural field in the previous growing season in the method, and a computer program element and a computer-readable medium for executing the method.

Description

Technical Field

[0001] The present disclosure relates to a computer-implemented method for selecting at least one agricultural field for producing crop plant seeds by using past satellite image data of a field from at least one previous growing season. Further objects are the use of past satellite image data of agricultural fields of a previous growing season for the above method, a system configured to execute the above method, and a computer-readable medium and a computer program element configured to execute the above method.

Background Art

[0002] In modern agriculture, crops with high performance under various conditions are grown by farmers. Crops have a complex set of traits that enable them to withstand all kinds of external adverse conditions such as drought, nutrient deficiency, competition with weeds, fungal diseases, and pest infestations. Crops may also have traits that enable modern agricultural techniques such as herbicide tolerance that allows for the selective control of weeds by broad-spectrum herbicides. Further high-performance crops are being created by hybridization of parental lines that produce so-called hybrid vigor effects in the first generation of hybrids (F1). Farmers purchase seeds, roots, sprouts, and other seeds from suppliers. Suppliers produce the above seeds in agricultural fields by crossing appropriate parental lines or by propagation of the seeds.

[0003] In the production of crop seeds such as seeds, it is of utmost importance to produce genetically pure products that contain as few off-specification seeds as possible. Clearly, the presence of off-specification seeds poses a risk for farmers in terms of an observed undesirable impact on the proportion of off-specification seeds. For example, BASF produces various types of herbicide-tolerant crops under the trade name Clearfield. When a farmer purchases such Clearfield products, e.g., corn, the corn plants grown from the commercially available corn seeds are expected to have specific herbicide tolerance. Thus, the farmer expects the crops grown for Clearfield seeds to be suitable for herbicide applications such as herbicide burn-down treatment without harming the crops. However, if the seeds contain 5 - 10% impurities that are not herbicide-tolerant, the corn plants without herbicide tolerance are phytotoxic to the herbicide in exactly the same way as weed plants, so the farmer's corn harvest will be approximately 5 - 10% less than expected under normal conditions.

[0004] Similarly, suppliers of plant seeds also need to ensure that they produce only products that are free of impurities or have very few impurities. The main source of impurities is inappropriate production sites. If an agricultural field has been used to grow crops of the same species during a previous growing season, the plant seeds that were not harvested after the previous growing season remain on the field, such as in the soil. Thus, the plant seeds germinate during the current growing season and produce seeds again through vegetative propagation or hybridization events with the crops grown in the current season. This can result in a significant amount of off-specification seeds being produced in the final product.

[0005] Similarly harmful to the quality of plant propagation products are crops of the same species as the seeds growing in adjacent or nearby fields. The pollen of such crops is carried by wind or insects to the field where the seeds are produced, causing cross-pollination of the plants. Again, in this case, the seeds produced do not contain the desired traits to a significant extent. This causes economic damage to farmers and damages the reputation of suppliers and producers of crop seeds.

Summary of the Invention

Problems to be Solved by the Invention

[0006] Therefore, in order to reduce the above risks, it is desirable to provide a tool for selecting at least one field for the production of crop plant seedlings for producers of plant seedlings.

Means for Solving the Problems

[0007] The above object is a computer-implemented method (100) for selecting at least one agricultural field for the production of crop plant seedlings, comprising: a) providing, to a processing unit, past satellite image data of at least one agricultural field from at least one previous growing season (101); b) determining, by the processing unit based on the image data, whether a crop has grown in at least one field during at least one previous growing season (102); c) providing, by the processing unit, information on whether a crop has grown in at least one field during at least one previous growing season (103); d) selecting, based on the provided information, at least one suitable agricultural field for the production of plant seedlings, wherein the crop has not grown in at least one agricultural field during at least one previous growing season (104). This is achieved by a computer-implemented method (100) including the above steps.

[0008] Surprisingly, it has been found that past satellite image data can be used with very high accuracy to identify the types of crops that were grown in agricultural fields over previous growing seasons. Thus, the tool enables the selection of which agricultural fields are suitable for the production of plant seedlings of specific crops. These objectives, and other objectives that will become apparent upon reading the following description, are solved by the subject matter of the independent claims. The dependent claims refer to preferred embodiments of the invention.

[0009] In one aspect of the present disclosure, the invention is a computer-implemented method (100) for selecting at least one agricultural field for the production of plant seedlings of a crop, the method comprising: a) providing, to a processing unit, past satellite image data of at least one agricultural field from at least one previous growing season (101); b) determining, by the processing unit based on the image data, whether the crop was grown in at least one field during at least one previous growing season (102); c) providing, by the processing unit, information as to whether the crop was grown in at least one field during at least one previous growing season (103); d) selecting, based on the provided information, at least one suitable agricultural field for the production of plant seedlings, wherein the crop was not grown in at least one agricultural field during at least one previous growing season (104). The invention relates to a computer-implemented method (100) comprising the steps above.

[0010] Another aspect relates to the use of past satellite image data of agricultural fields of a previous growing season in the method according to any one of the preceding claims.

[0011] A third aspect is a system for selecting an agricultural field for the production of crop seeds, the system comprising: a) a receiving unit configured to receive (101) past image data of at least one agricultural field from at least one previous growing season b) a processing unit, - based on the image data, determining (102) whether a crop was grown in at least one field during at least one previous growing season, - providing (103) information on whether a crop was grown in at least one field during at least one previous growing season, - based on the provided information, selecting (104) at least one suitable agricultural field for the production of plant seedlings, if the crop was not grown in at least one agricultural field during at least one previous growing season a processing unit configured as such, and relates to a system comprising the same.

[0012] A fourth aspect relates to a computer program element having instructions configured to execute the steps of the above method in the above system when executed on a computing device in a computing environment. A fifth aspect relates to a computer-readable medium storing the above computer program element.

[0013] Any disclosure and embodiment described herein relates to the methods, systems, methods of use, computer program elements, and / or computer-readable media outlined above, and vice versa. Advantageously, the advantages provided by any of the plurality of embodiments and examples apply equally to all other embodiments and examples, and vice versa.

[0014] This disclosure is based on the discovery that past satellite image data of fields during previous growing seasons can be used to determine whether a particular type of crop, e.g., a particular plant species, was grown in an agricultural field during a previous growing season. This information can be used to select a suitable production site for creating plant seedlings, e.g., seeds.

[0015] Definition (Agriculture) The term "field" relates to a workable plot of land on which crops such as fruits and vegetables, or row crops such as corn or rapeseed are grown, and / or which has been grown on in at least one previous growing season. In relation to step b), i.e., in relation to determining whether a crop has been grown on at least one field, the term "field" includes both the entire field and parts of the field such as half or a third of the field. Agricultural fields are not related to covered facilities such as greenhouses. The location of at least one agricultural field can be provided by a human user or obtained from a GPS signal. For example, the location of a GPS data point representing at least one of the field boundary and the field center coordinates can be used to determine the geographical location of the field. Thus, the first agricultural field or several agricultural fields for the purpose for which plant seedlings are expected to be grown are provided either from human input or from a GPS signal. Then, satellite image data of the above fields is acquired. In addition, satellite image data of adjacent or remote agricultural fields can be acquired as described below.

[0016] The term "plant seedlings" relates to any kind of reproductive plant material such as seeds, roots, and sprouts. In a preferred embodiment, the plant seedlings relate to seeds, in particular to seeds of plants of the Brassicaceae family such as Brassica napus, Sinapis alba, in particular Brassica napus. The term "crop" relates to the plants obtained by growing plant seedlings. Thus, the term "crop" typically relates to plants of the Brassicaceae family such as Brassica napus, Sinapis alba, in particular Brassica napus.

[0017] The term "previous growing period" relates to the last or previous period in a growing season during which an agricultural field was tilled, i.e., a crop was grown, or grass or clover was grown in a rotation system. Depending on the location of the agricultural field, two or more growing periods may be included in one growing season. The term "at least one previous growing period" typically relates to the last growing period prior to the current or upcoming growing period. In one embodiment, the term "at least one previous growing period" relates to the previous two, three, four, five, or up to ten growing periods prior to the current or upcoming growing period. In one embodiment, the term "at least one previous growing period" relates to at least the last three, preferably at least the last five previous growing periods. In one embodiment, the term "previous growing period" relates to at least the last two previous growing periods.

[0018] The term "adjacent (agricultural) fields" relates to two fields that are separated by a road, a river, or any other artificial or natural object that does not exceed 50 meters in the horizontal dimension on the ground and that share at least one common boundary.

[0019] The term "machine learning algorithm" should be understood in a broad sense and preferably includes decision trees, naive Bayes classification, nearest neighbor methods, neural networks, convolutional neural networks, adversarial generative networks, support vector machines, linear regression, logistic regression, random forests, and / or gradient boosting algorithms. Preferably, the machine learning algorithm is configured to process an input having a high dimension into an output of a much lower dimension. Such a machine learning algorithm is called "intelligent" because it can be "trained". The algorithm may be trained using records of training data. The records of training data include training input data and corresponding training output data. The training output data of the records of training data is the result expected to be generated by the machine learning algorithm when the training input data of the same record of training data is given as input. The deviation between this expected result and the actual result generated by the algorithm is observed and evaluated by a "loss function". This loss function is used as feedback for adjusting the parameters of the internal processing chain of the machine learning algorithm. For example, the parameters can be adjusted according to an optimization goal of minimizing the value of the loss function, which is obtained when all the training input data is input into the machine learning algorithm and the result is compared with the corresponding training output data. As a result of this training, even if a relatively small number of records of training data are given as "ground truth", the machine learning algorithm can perform its work well for a number of records of input data that is many orders of magnitude larger.

[0020] As used herein, the term "crop" can refer to plants such as grains, fruits, or vegetables grown in large quantities. Preferred crops are onion (Allium cepa), pineapple (Ananas comosus), peanut (Arachis hypogaea), asparagus (Asparagus officinalis), oat (Avena sativa), sugar beet (Beta vulgaris spec.altissima), turnip (Beta vulgaris spec.rapa), rapeseed (Brassica napus var.napus), rutabaga (Brassica napus var.napobrassica), turnip rape (Brassica rapa var.silvestris), Brassica oleracea, Brassica nigra, Camellia sinensis, Carthamus tinctorius, Carya illinoinensis, Citrus limon, Citrus sinensis, Coffea arabica (Coffea canephora), Coffea liberica, Cucumis sativus, Cynodon dactylon, Daucus carota, Elaeis guineensis, Fragaria vesca, Glycine max, Gossypium hirsutum, (Gossypium arboreum, Gossypium herbaceum, Gossypium vitifolium, Helianthus annuus, Hevea brasiliensis, Hordeum vulgare, Humulus lupulus, Ipomoea batatas, Juglans regia, Lens culinaris, Linum usitatissimum, Lycopersicon lycopersicum, Malus spec., Manihot esculenta, Medicago sativa, Musa spec., Nicotiana tabacum (N. rustica), Olea europaea, Oryza sativa, Phaseolus lunatus, Phaseolus vulgaris, Picea abies, Pinus spec.) Pistacia vera, Pisum sativum, Prunus avium, Prunus persica, Pyrus communis, Prunus armeniaca, Prunus cerasus, Prunus dulcis and Prunus domestica, Ribes sylvestre, Ricinus communis, Saccharum officinarum, Secale cereale, Sinapis alba, Solanum tuberosum, Sorghum bicolor (s. vulgare), Theobroma cacao, Trifolium pratense, Triticum aestivum, Triticale, Triticum durum, Vicia faba, Vitis vinifera and Zea mays. The most preferred crops are the following: Arachis hypogaea, Beta vulgaris spec. altissima, Brassica napus var.rapeseed (Brassica napus), kale (Brassica oleracea), lemon (Citrus limon), orange (Citrus sinensis), coffee (Coffea arabica) (Coffea canephora, Coffea liberica), Bermuda grass (Cynodon dactylon), soybean (Glycine max), upland cotton (Gossypium hirsutum) (Gossypium arboreum, Gossypium herbaceum, Gossypium vitifolium), sunflower (Helianthus annuus), barley (Hordeum vulgare), walnut (Juglans regia), lentil (Lens culinaris), flax (Linum usitatissimum), tomato (Lycopersicon lycopersicum), apple (Malus spec.), alfalfa (Medicago sativa), tobacco (Nicotiana tabacum) (Nicotiana rustica), olive (Olea europaea), rice (Oryza sativa), lima bean (Phaseolus lunatus), common bean (Phaseolus vulgaris), pistachio (Pistacia vera), pea (Pisum sativum), almond (Prunus dulcis), sugarcane (Saccharum officinarum), rye (Secale cereale), potato (Solanum tuberosum), sorghum (Sorghum bicolor (s. vulgare)), triticale (Triticale), bread wheat (Triticum aestivum), durum wheat (Triticum durum), broad bean (Vicia faba), European grape (Vitis vinifera) and maize (Zea mays).

[0021] As used herein, "determining" includes "starting or causing to determine", "generating" includes "starting or causing to generate", and "providing" includes "starting or causing to determine, generate, select, transmit, or receive". "Starting or causing to execute an action" includes any processing signal that causes a computing device or processing unit to execute each action.

[0022] The "determining" used in "determining by a processing unit, [...]" relates to an automatic determination executed by the processing unit without human interaction.

[0023] The computer-implemented method (100) includes step a) of providing (101) to a processing unit past satellite image data of at least one agricultural field from at least one previous growing season. This step is typically realized by providing past satellite image data of at least two agricultural fields, preferably of a larger area such as an entire geographical region, country, or state.

[0024] The method may typically include step a1) of determining (101a) field boundaries in the past satellite image data. This step may typically be executed after step a) and before step b). However, it is also possible to execute step a1) after step b) and before step c).

[0025] Such determination can be realized by a field boundary detection model. The field boundary detection model is typically obtained by using supervised machine learning. Suitable machine learning techniques include deep learning techniques, in particular, the use of convolutional neural networks for segmentation, such as UNet or ResUNet-a (Diakogiannis, F.I., Waldner, F., Caccetta, P., Wu, C., 2019. Resunet-a: a deep learning framework for semantic segmentation of remotely sensed data. arXiv preprint arXiv:1904.00592; https: / / www.tensorflow.org / ).

[0026] Training images for these supervised machine learning techniques should be as uniform as possible by selecting the observations with the lowest cloud cover rate from a specified interval (e.g., within 3 months). Gaps due to clouds in the selected observations may be filled with observations in the absence of other clouds. This approach can generate both complete images and images with minimal artificial disturbances in the images generated by freely synthesizing images from different observation dates. Additionally, an additional layer can be created to encode the observation time of each pixel to show the artificial disturbances due to replacement to the model. Thereby, the model can learn to identify the disturbances. Further, Sobel filters applied to individual satellite bands are typically generated to enhance the visibility of optical edges in the images.

[0027] The selected training data in vector format is typically rasterized to match the satellite data. To train the model, three different targets can be derived, as described in Waldner et al. (Deep learning on edge: extracting field boundaries from satellite images with a convolutional neural network; 2020; arXiv preprint arXiv:1910.12023v2), namely, the binary mask of the field boundary, the binary mask of the field extent, and the normalized distance per field (distance to the nearest boundary).

[0028] To train the model, the satellite data and the training data can be sliced down into smaller parts (e.g., images of 128×128 pixels) to fit into the GPU memory. The model based on the tensorflow library (www.tensorflow.org) is typically optimized by minimizing the Tanimoto loss (see Waldner, F., Diakogiannis, F.I., 2020: Deep learning on edge: extracting field boundaries from satellite images with a convolutional neural network. arXiv preprint arXiv:1910.12023v2).

[0029] Next, the trained segmentation model can be used to derive field boundary predictions and field extent predictions for the targeted new area. They typically utilize the mechanisms described above for the training process. A satellite image is selected, preprocessed, and sliced down to generate the input for inference. The inference is performed independently for multiple time points over one or more seasons to mitigate the changes in the appearance of agricultural fields due to the growth process throughout the year. Then, their predictions for the slices are recombined into larger units (tiles).

[0030] Field boundary predictions can ultimately be combined to generate a field boundary in vector format. The individual predictions of the areas at different times can be merged (e.g., via an average or maximum operation), artificially enlarged to a higher resolution (e.g., from 10 m to 2 m), and smoothed. This enlargement allows the system to compensate for the effects of the coarse pixels of the satellite data and result in a smoother field boundary. The probability model predictions with continuous values from 0 to 1 are classified by a threshold to give binary values. These binary masks can then be combined into the final field boundary mask by subtracting the binary field extent prediction from the binary field boundary prediction. Finally, the raster format segments can be vectorized, and minor adjustments such as smoothing the boundary and filling in smaller gaps within the field can be performed. The obtained field boundary is stored in file storage for easy access and use and referenced in a database.

[0031] In one embodiment, the past satellite image data includes time-resolved image data over at least one previous growing season. The term time-resolved image data relates to a series of image data captured over a past growing season. The accuracy of the object of the invention described herein increases only by the number of measurements per growing season. Typically, a set of image data for one field per week is captured over the growing season.

[0032] In one embodiment, step b) includes determining time-resolved vegetation index data selected from normalized difference vegetation index (NDVI) data and / or leaf area index (LAI) data, normalized difference water index (NDWI), enhanced vegetation index (EVI) data, and / or any other vegetation-based index data from the image data. Such vegetation indices and their calculations from satellite images are known to those skilled in the art and are described, for example, at https: / / en.wikipedia.org / wiki / Vegetation_Index, or at https: / / earthobservatory.nasa.gov / features / MeasuringVegetation / measuring_vegetation_2.php.

[0033] In one embodiment, step b) includes classifying at least one field by crop type according to the time-resolved vegetation index data. The classification typically includes assigning the crops grown in the field to a specific plant species or at least a plant genus or family.

[0034] It has been found that it is possible to accurately classify plants growing in agricultural fields from satellite image data, particularly time-resolved image data such as time-resolved vegetation index data.

[0035] Therefore, it is possible for the processing unit to determine which type of plants are growing in the field. The processing unit executes the above task by using a classification model. The classification model can be obtained by various methods such as machine learning, especially supervised learning. Techniques useful for supervised learning to obtain the classification model are decision tree-based modeling such as logistic regression, perceptron algorithm, Bayes classification, naive Bayes classification, k-nearest neighbor algorithm, artificial neural network, and random forest algorithm. Training data for generating an appropriate model is typically obtained by annotating satellite image data with ground truth data, and the ground truth data can be collected by, for example, an agronomy advisor, a user, or a field-based machine such as a BASF Smart Sprayer.

[0036] The highest prediction accuracy is typically achieved by using time-resolved vegetation index data, but it is equally possible to determine the type of crops growing in the field using a single satellite image of an agricultural field.

[0037] When a vegetation index is used, a single index such as the NDVI or LAI index may be sufficient. However, it goes without saying that the accuracy of the determination can be significantly improved by combining various vegetation indexes.

[0038] Another method useful in step b) is a wide range of statistical modeling techniques. For example, satellite images indexed with ground truth data can be used to generate calibration curves for specific crops. These calibration curves can then be used in regression analysis to classify new satellite image data.

[0039] The accuracy achievable by such a method is typically at least 90%, usually at least 95%. When a large number of agricultural fields are evaluated, even a small error can cause a rather unstable situation, so it is particularly important that the accuracy of the determination in step b) is high.

[0040] Thus, the determination in step b) first results in a specific probability that the crop was grown on an agricultural field. This information can be provided directly in step c). Alternatively, the agricultural field can be classified as a field on which the crop was or was not grown during at least one previous growing period based on a predefined benchmark, and then this classified information can be provided in step c).

[0041] The method of the present invention is used to select an agricultural field for the production of plant seedlings of a crop based on the provided information, and step d) of selecting an agricultural field suitable for the production of plant seedlings, wherein the crop has not been grown on an agricultural field during at least one previous growing period, includes step (104).

[0042] The term "selecting at least one agricultural field" is typically executed by a processing unit, which may be the same as or different from the one used in steps a) to c), and preferably may be the same. When the selection is executed by the same processing unit, step c) of providing the determined information only relates to the use of the determined information in step b) as an input for the selection process in step d) by the processing unit. When the selection in step d) is executed by another processing unit, the processing unit provides the determined information in step c) to the other processing unit. In this case, the processing units are usually communicatively coupled to exchange the above information. In one embodiment, the selection in step d) is executed by a human user.

[0043] The term "selecting at least one agricultural field" includes, when the above information in step b) is determined for two or more fields, selecting a plurality of agricultural fields suitable for the production of crop plant seedlings. The selection is carried out based on information on whether the crop that can grow for the produced seedlings has been grown in at least one field in at least one previous growing period. Only satellite image data of one field is provided in step a), and when the determination in step b) relates to only one field, the selection of the appropriate field may simply result in the selection of the one field being analyzed, or, if it is determined in step b) that the crop has already been grown in the above field in at least one previous growing period, it may result in a selection of no field.

[0044] In one embodiment, the method includes step e) of outputting information regarding the selection in step d), or information when the crop in step c) has been grown in at least one field. The term "output" may refer to an electronic signal that can be output via an output interface, and the output interface is connected or otherwise coupled to the processing units in steps b) / c) and d). The electronic signal can be used to display information on a suitable user interface that receives the electronic signal, such as a personal computer, mobile phone, tablet, smartwatch, virtual reality device, etc. When the user directly receives the information in step c), i.e., information on whether the crop has been grown in at least one field in at least one previous growing period, the selection of the appropriate agricultural field in step d) can be carried out by the user based on the received information.

[0045] However, the output of the selection in step d) can be used in subsequent steps, preferably in an automatic crop seedling production system, for carrying out production in the selected agricultural field. For example, if the agricultural field is identified as useful for the creation of high-purity products, an automatic planting device can be used to plant the plant seedlings of the target crop in the above field. In this case, the automatic planting device can typically be an autonomous mobile vehicle such as a tractor.

[0046] In one embodiment, image data of at least two agricultural fields is provided, at least two of the agricultural fields being adjacent fields or the boundaries of at least two of the agricultural fields being at most 10 km apart, and the selection in step d) is also based on information regarding the presence of crops in at least one of the second at least one field during a previous growing season.

[0047] The main causes of impurities in the produced plant seedlings are: a) residual plant seedlings from a previous growing season in the soil of the agricultural field in which a new batch of plant seedlings is produced, which have grown into crops and are harvested together with the produced seedlings, and b) cross - breeding events with other plants that change the genotype of the produced plant seedlings of the crop that produces a new batch of plant seedlings in an undesirable way.

[0048] Cross - breeding typically occurs by the exchange of pollen via insects or wind. Therefore, it is desirable to find an agricultural field that not only has a history of not growing the same crop (same species, genus, or at least family, preferably species) in at least one previous growing season itself, but also has no agricultural fields where the crop (species, genus, or at least family, preferably species) was grown in a previous growing season nearby.

[0049] Therefore, the method may include imaging data of at least one adjacent field and / or further remote fields with respect to the target field. The distance between the remote field and the target field depends on the desired purity of the produced seedlings, but is typically at most 10 kilometers, preferably at most 5 kilometers, more preferably at most 3 kilometers, particularly at most 2 kilometers, for example at most 1 kilometer. For example, the typical foraging radius of a honeybee is at most 3 kilometers. Therefore, if the target crop can be pollinated by honeybees, agricultural fields up to 3 kilometers away from the target field should be considered.

[0050] In one embodiment, image data of one agricultural field and all adjacent fields is provided, and the selection in step d) is based on information regarding the presence of crops in the field and all adjacent fields during at least one previous growing season. The probability of cross-pollination decreases with the distance of the crops. Thus, the fields adjacent to the target field are the cause of most of the impurities in the produced plant seedlings, and it is advisable to monitor at least all directly adjacent agricultural fields.

[0051] In one embodiment, the image data is acquired by using synthetic aperture radar (SAR) via a satellite or light detection and ranging (LIDAR).

[0052] The method may include an additional step of outputting, by a processing unit, the provided information regarding whether the crop was grown in at least one field during at least one previous growing season or regarding the selection of at least one suitable agricultural field for the production of plant seedlings, where the crop was not grown in at least one agricultural field over at least one previous growing season.

[0053] Another embodiment relates to the use of past satellite image data of agricultural fields during the previous growing season in the above method.

[0054] In another embodiment, the present invention is a system for selecting an agricultural field for the production of crop seeds, the system comprising a) a receiving unit configured to receive (101) past image data of at least one agricultural field from at least one previous growing season, a) a processing unit, - determining (102) based on the image data whether the crop was grown in at least one field during at least one previous growing season, - providing (103) information on whether the crop was grown in at least one field during at least one previous growing season, - Based on the provided information, select at least one suitable agricultural field for the production of plant seedlings (104), where the crop has not been grown in at least one agricultural field during at least one previous growing period A processing unit configured as and a system comprising the same.

[0055] The receiving unit and the processing unit are communicatively coupled, and preferably, the receiving and processing units have a connection interface. The connection interface may be a direct local connection such as realized by an electrical connection, a local area network, a wireless connection, etc., or the connection may be a long-distance connection such as a connection through the Internet, a wide area network, a dial-in connection, a cable modem, etc.

[0056] The receiving unit is configured to receive past image data of at least one agricultural field from at least one previous growing period to the processing unit. The receiving unit typically relates to a receiving interface communicatively coupled to the processing unit.

[0057] The receiving unit can communicate with a remote device such as a data source computing device that stores the original satellite image data captured by a satellite, or a local or remote memory device that stores past satellite image data. The data source computing device can be accessed via a web service or an API (Application Programming Interface).

[0058] In one embodiment, the receiving unit may be the computing device itself having a storage memory and a processing unit, such as a personal computer or a server, or may be an electrical interface of a precession unit such as on a computing device or a personal computer for receiving image data via a short - distance or long - distance connection via the Internet, such as a LAN connection, a USB connection, a WLAN connection, etc. (e.g., from a cloud - based memory).

[0059] The processing unit is configured to determine (102) whether the crop has been grown in at least one field during at least one previous growth period based on the image data, and to provide (103) information as to whether the crop has been grown in at least one field during at least one previous growth period. The processing unit obtains the image data from a receiving unit to which the processing unit is communicatively coupled as described above. The term processing unit typically relates to general-purpose processing devices such as microprocessors, microcontrollers, central processing units, etc. More specifically, the processing means may be a CISC (Complex Instruction Set Computing) microprocessor, a RISC (Reduced Instruction Set Computing) microprocessor, a VLIW (Very Long Instruction Word) microprocessor, or a processor implementing another instruction set, or a processor implementing a combination of instruction sets. The processing means may also be one or more dedicated processing devices such as ASICs (Application Specific Integrated Circuits), FPGAs (Field Programmable Gate Arrays), CPLDs (Complex Programmable Logic Devices), DSPs (Digital Signal Processors), network processors, etc. The methods, systems, and devices described herein may be implemented as software within a DSP, microcontroller, or any other side processor, or as hardware circuitry within an ASIC, CPLD, or FPGA. The term "processing unit" or processor may also refer to one or more processing devices, such as a distributed system of processing devices (e.g., cloud computing) arranged across multiple computer systems, and is not limited to a single device unless otherwise specified.

[0060] The processing unit is also configured to select at least one suitable agricultural field for the production of plant seedlings based on information as to whether the crop has been grown in at least one field during at least one previous growing period, where the crop has not been grown in at least one field over at least one previous growing period (104). The processing unit may also be configured to output information (104) regarding the selection. The processing unit may also be configured to output information as to whether the crop has been grown in at least one field during at least one previous growing period.

[0061] Output is typically realized via an output interface communicatively coupled to the processing unit. The output interface may be a human-machine interface. Thus, the processing means can then output, via the human-machine interface, the information of step c) or the information regarding the selection of step d). Thus, the user can be informed about the suitability of at least one agricultural field or the selection of one or more suitable fields within a plurality of fields.

[0062] The human-machine interface may include a video display unit (e.g., a liquid crystal display), a cathode ray tube display, or a touch screen, and / or a signal generating device (e.g., a speaker). Thus, the human-machine interface may be, for example, a visual interface such as a screen and / or an audio interface such as a speaker. Thus, the output can be displayed to the user and / or announced via the speaker.

[0063] Thus, the process of informing the user about the selection of a suitable field can be made simple and intuitive for the user. For example, the device can be intuitively coded, for example, using color codes.

[0064] The human-machine interface may include, among other possibilities, a web browser and client applications. With the web browser, a user can typically display and interact with media and other information embedded in a web page or website. With the client application, a user can interact with a server application from a server system or a web server.

[0065] Hereinafter, the present disclosure will be described in detail with reference to the accompanying drawings.

Brief Description of the Drawings

[0066]

Figure 1

Figure 2

Figure 3

Figure 4

Modes for Carrying Out the Invention

[0067] FIG. 1 is a flowchart of an exemplary method (100) for determining whether a crop has been grown in at least one agricultural field during a previous growing season. The crop may be, for example, canola of BASF's Clearfield product series, which is highly resistant to the herbicide product imazamox. Thus, a farmer can use imazamox to control unwanted vegetation on the agricultural field without damaging the crop. Clearfield canola seeds should not be contaminated by seeds of canola produced in an agricultural field and not enhanced with imazamox resistance. This is because plants grown from these seeds are strongly affected by the farmer's imazamox treatment and the yield decreases accordingly.

[0068] In a first step (101), past satellite image data of at least one agricultural field from at least one previous growing season is provided to a processing unit. The satellite image data is typically obtained from a server via the Internet, for example via an API interface. The satellite image data relates to only one target field where it can be evaluated whether plant seedlings are produced. Of course, it can also relate to more agricultural fields within a predetermined geographical area, such as within a predetermined radius around the target field, or within a predetermined radius around a city, farm, etc. The target field or the target geographical area may be input by the user or automatically determined automatically by using a positioning signal such as a GPS signal or a mobile phone positioning signal. For example, image data of agricultural fields within a predetermined radius centered on the positioning signal can be obtained.

[0069] The image data is past image data of a previous growing season, such as last year or the past three years. Past satellite image data is available in many regions and is stored on a generally accessible server. The accuracy of the method (100) of the present invention increases when image data of two or more previous growing seasons is used, and the strongest influence is expected to be from the last growing season.

[0070] The image data may be a single satellite photo or may be time - series image data covering at least one growth period, such as at least two satellite photos per growth period. The accuracy of the method (100) of the present invention increases when more image data is provided per growth period.

[0071] In the next step b (102), the processing unit determines from the image data whether the crop has been grown in at least one field during at least one previous growth period. For this purpose, the processing unit may classify the satellite image data according to the type of crop. As outlined above, the classification can be achieved by various statistical methods, for example, by using a machine - learning - based model. Such a machine - learning - based model can typically be obtained by using annotated satellite image data as a training data set, and the annotation is based on ground - truth data collected by agricultural workers or agronomic advisors. Modeling and classification are discussed and illustrated in Figure 3. The model may use raw image data as input or may use data with reduced complexity, such as vegetation index data like the NDVI index.

[0072] In the next step c (103), the processing unit provides the information determined in step b). In one embodiment, the term "provide" refers to the generation of an output presented by a human-machine interface such as a personal computer, mobile phone, or tablet. The human-machine interface can display the information on a special application such as Xarvio or on web browser software. Thus, the term "provide" means the generation of an output signal to initiate the presentation of the information by the human-machine interface. In another embodiment, the term "provide" refers to a selection step d (104) of selecting at least one suitable agricultural field for the production of plant seedlings based on the provided information, where the crop has not been grown in this at least one agricultural field during at least one previous growing period. The selection is typically performed by the same processing unit as steps b) and c) (102, 103), but may also be realized by a different processing unit or even by the user. The selection can be performed based on the classification of at least one agricultural field in step c (103) and, if two or more fields are analyzed, based on the relative location of the fields to each other.

[0073] If one target field is analyzed, the classification of the field will be either that a crop that could cause contamination of the produced seedlings was grown in the field during at least one previous growing period ("red") or not grown ("green"). Thus, as a result of the selection, this field is selected ("green") or excluded as not being suitable ("red").

[0074] When two or more target fields are analyzed, each of the fields is first analyzed to determine whether a crop that can cause contamination of the produced seedlings has been grown in each of these fields during at least one previous growing season. Additionally, the selection step (104) can consider the results of adjacent or nearby fields for each field in order to further reduce the risk of contamination. For example, all adjacent fields and optionally fields within a radius of 3 kilometers can be considered. The radius covering other fields can be automatically adaptable according to an acceptable risk level. For example, the user can define a probability of contamination, which is converted into the radius of the fields considered in the analysis of the target agricultural field. The risk can typically be lower for fields that are further away or even adjacent fields compared to closer fields, such as projected by an exponential risk distance dependency. When two or more fields are analyzed, the selection process (104) is typically executed by a processing unit rather than by a human user because complex risk scoring calculations are required.

[0075] Next, the selected field can be presented by a human-machine interface such as a personal computer, a mobile phone, or a tablet. The human-machine interface can display information on a special application such as Xarvio, on web browser software, etc. Thus, the processing unit generates an output signal to initiate the presentation of information by the human-machine interface. Next, the user can determine, based on the information, which of the pre-selected fields based on the risk of contamination should be chosen to produce plant seedlings by the processing unit.

[0076] FIG. 2 shows a schematic view of fields (201, 204-206) from which past satellite image data can be collected. In each of the fields, different crops were grown during previous growing seasons such as last year. For example, rapeseed was grown in field (201), leguminous plants were grown in the first adjacent field (204), rye was grown in the second adjacent field (205), and corn was grown in the remote field (206). Satellite image data of the geographical area where the fields are located is collected and typically stored in a server (203) or cloud environment that is generally publicly available and can be used in the method of the present invention.

[0077] For example, a user may define a field, several fields, or a geographical area in which Clearfield corn is to be produced. The processing unit then initiates a process of receiving past satellite image data of the geographical area. Next, the processing unit determines for each of the fields, or for each of the fields within the geographical area, whether corn (Zea mays) was grown in the field during a previous growing season. The information is then either presented directly to the user or used in a further analysis step where a field suitable for the production of Clearfield corn is automatically selected. The selection will of course take into account the result of the determination step b) for the fields, but typically may also take into account the risk of contamination from other fields.

[0078] For example, if Clearfield maize is to be produced, in order to protect the herbicide tolerance genotype of the F1 generation, it is necessary to reduce the risk of contamination by pollen from other lines of maize (Zea mays) as much as possible. Maize (Zea mays) pollen is mainly carried by the wind and is carried only over a relatively short distance, which can be taken into account in selection step d). Pollen from other plants such as rapeseed is carried by insects such as bees and can thus be carried over longer distances (such as several kilometers). Therefore, when seeds of the Clearfield line of maize (Zea mays) are produced in a field (201), image data captured by a satellite (202) in at least one previous growing season for the said field (201) and optionally adjacent fields (204), (205) and remote fields (206) can also be used. The distance to the target field (201) can be defined by distance, and the distance can be the result of a risk scoring analysis. Typically, the risk of contamination is extremely low when the remote field (206) is at least 1 kilometer away from the field (201), and increases as the fields get closer.

[0079] Therefore, the selection process takes into account the information determined in step b) for fields within a predetermined radius centered on the field (201), and the radius depends on a predetermined risk level. However, since the risk of contamination is unacceptable, adjacent fields are usually considered regardless of the defined risk level.

[0080] Therefore, selection step d) typically takes into account the information determined in step b) for at least the field (201) to be evaluated and adjacent fields (204, 205). For example, if a Clearfield line of rapeseed is to be produced in a field (201) and the determination in step b) classifies the field as having grown rapeseed in the field during a previous growing season, then naturally, such a field (201) is not eligible to grow the Clearfield rapeseed line because the probability of contamination from seeds not harvested in the previous season is evaluated to be high.

[0081] Similarly, when analyzing whether the field (204) is suitable for Clearfield rapeseed production, the processing unit considers the classification result of the agricultural field (201). Since the field is directly adjacent, the field (204) is not selected as suitable for production. The probability that non-Clearfield rapeseed seeds remain in the soil of the field (201) is estimated to be very high, and rapeseed plants will germinate and grow in the coming season. Pollen from such non-Clearfield plants is carried to the field (204) by insects, likely causing harmful mating events with Clearfield plants and resulting in seed products with individual seeds having undesirable genotypes.

[0082] Figure 3 shows the change in the NDVI index over a season. The NDVI vegetation index takes into account and correlates the relationship between the intensities of different spectral regions within an image, particularly the near-infrared and visible red light portions of the spectrum. The NDVI index of a plant changes during the season as shown in the left window of Figure 3. In the case of green, healthy leaves (303), the near-infrared intensity is somewhat higher compared to the red light portion of the spectrum. This is different for young leaves (302) or brown leaves (301). On the right side of Figure 3, a time-resolved data series of NDVI values over one season is shown for different crops (304, 305, 306, 307). Depending on the growth cycle of the plant, the NDVI has a very specific pattern for each plant. For example, some crops germinate and start growing at different times during the season. As can be seen from curve (304), the rise of the curve is somewhat early and steep at the beginning of the winter season, while curve (306) rises slowly and steadily during the winter season. Finally, the NDVI curve (307) rises only in the summer in early June. Although only the characteristic patterns of various crops are shown for the NDVI index, they also exist for other vegetation indices such as the LAI index.

[0083] These patterns can be used to identify the types of crops grown in the field. It is not necessary to record the entire time-resolved pattern of the index, but the more available data points there are, the higher the accuracy of the information. Typically, not only one vegetation index but also a series of vegetation index factors are used as input factors. When a machine learning model is used, it is also possible to use the entire satellite image data as input data without reducing the complexity. As described above, satellite image data in an annotated format can be used as a training data set for generating a machine learning model. The annotation can be achieved, for example, by recording agricultural worker data using a customer front-end tool such as the BASF Xarvio suite. It can also be achieved by using the observation data obtained by sales representatives. The annotated training data can be used in supervised machine learning techniques. For example, the training data can include satellite images of a geographical area, and for a specific agricultural field, the type of crop growing in the field and the time when the image was captured during the season are annotated. Alternatively, the training data can include one or more vegetation indices annotated at the time when the image of the crop type and the season was captured. Preferably, the image data is time-resolved and includes image data at at least two time points during the season. Machine learning tools typically use a loss function to generate a model that describes the training data in the best possible way. Typically, validation data is used to avoid overfitting. Next, the newly captured satellite image data can be analyzed using the model thus obtained.

[0084] When machine learning techniques are not used, it may be advisable to generate a calibration curve of the vegetation index from the annotated training data, such as by determining the average of various curves captured for the same crop. The calibration curve can then be used in a regression technique to classify new data according to a predefined calibration curve. Typically, this is achieved by minimizing the deviation of the newly measured curve from the calibration curve and classifying the field according to the fit with the minimum deviation.

[0085] Therefore, the captured satellite image data can be analyzed to classify the crops grown in the agricultural fields during previous growing seasons.

[0086] Figure 4 shows a satellite image of a target geographic area with specific highlighted agricultural fields (401 - 412). The presentation of such information can be displayed to a user on a computer or smartphone to inform the user of the suitability of a particular piece of farmland for generating plant seedlings of a specific crop. The fields can be highlighted in monochrome tones or by colors indicating the suitability of the fields for producing plant seedlings of a specific crop.

[0087] For example, a user may input the type of crop to be used for producing specific plant seedlings in a geographic area. Next, the method of the present invention can determine whether the above - mentioned crop has already been grown in an agricultural field during a previous growing season and present that information to the user via a human - machine interface such as a computer screen. The information may also include information regarding the likelihood of crop identification. For example, if there is a high deviation in the regression, such as a low R2 value, this can be reflected by different tones of the color of the highlighted fields. As an example, if the system has not found that the target crop has been grown in the field over a previous growing season, the field can be colored green. However, if the regression is somewhat poor, the system of the present invention can determine that the accuracy of this information is not high, which can be reflected by a light - green tone. Similarly, if the system has found that the crop has been grown in the field during a previous growing season, the system can represent the estimated accuracy of this information in different tones of red.

[0088] If the method includes step d) of selecting a field suitable for the production of plant seedlings, the highlighting of the field may also reflect the risk of contamination from other nearby fields. Thus, the green or red gradation indicating the suitability of a field for the production of a particular type of plant seedlings can be affected by the classification of the fields within a given radius. As described above, the radius reflects the level of acceptable risk for the farmer and the type of crop, i.e., how pollen is carried by wind or insects. Information regarding the dependence of the contamination risk on a given type of crop and the distance between two fields is typically stored in a memory communicatively coupled to the processing unit. When the user inputs which type of crop should be used for production, the processing unit accesses the memory storage and selects a radius for the risk assessment in step d) according to a predefined acceptable risk score. Such a risk score may also, of course, be input by the user, which may depend on the need to produce a pure product.

[0089] Aspects of the present disclosure relate to computer program elements configured to perform the steps of the above method. Thus, the computer program elements may be stored on a computing unit of a computing device, which may also be part of one embodiment. This computing unit may also be configured to perform the steps of the above method or to induce its execution. Further, the computing unit may be configured to operate the components of the above system. The computing unit may be configured to operate automatically and / or to execute user instructions. The computing unit may include a data processor. The computer program may be loaded into the working memory of the data processor. Thus, the data processor may have the equipment for performing the method according to one of the above embodiments. This exemplary embodiment of the present disclosure includes both a computer program that uses the present disclosure from the beginning and a computer program that changes an existing program into a program that uses the present disclosure by means of an update means. Further, the computer program elements may be able to provide all the steps necessary to perform the procedure of the exemplary embodiment of the above method. According to a further exemplary embodiment of the present disclosure, a computer-readable medium such as a CD-ROM, a USB stick, etc., a downloadable executable file, etc. are presented, the computer-readable medium stores the computer program elements thereon, and the computer program elements are described in the previous section. The computer program may be stored on a suitable medium such as an optical storage medium or a solid-state medium supplied together with or as part of other hardware and / or distributed by such a suitable medium, but may also be distributed in other forms, such as via the Internet or other wired or wireless telecommunications systems. However, the computer program may also be presented through a network such as the World Wide Web and downloaded from such a network into the working memory of the data processor.According to a further exemplary embodiment of the present disclosure, a medium is provided that enables a computer program element to be downloaded, and the computer program element is arranged to execute a method according to one of the above-described embodiments of the present disclosure.

[0090] The present disclosure has been described by way of example in conjunction with the preferred embodiments. However, upon consideration of the drawings, the present disclosure, and the claims, other variations may be understood and practiced by those skilled in the art and by practicing the invention as set forth in the claims. In particular, any steps presented, especially, can be executed in any order, that is, the present invention is not limited to a particular order of these steps. Further, it is not essential that different steps be implemented at a particular location or at one node of a distributed system. That is, each of the steps can be implemented at different nodes using different devices / data processing units.

[0091] In the claims and in this specification, the word "comprising" does not exclude other elements or steps, and the indefinite article "a" or "an" does not exclude a plurality. A single element or other unit may perform the functions of a plurality of entities or items recited in the claims. The fact that certain means are recited in mutually different dependent claims does not indicate that a combination of these means cannot be used in an advantageous embodiment.

Claims

1. A computer-aided method (100) for selecting at least one agricultural field for the production of crop plant seedlings, a) Providing a processing unit with historical satellite image data of at least one agricultural field from at least one previous growing period (101), b) A step (102) in which the processing unit determines, based on the image data, whether or not the crop was grown in the at least one field during at least one previous growing period, c) The processing unit provides information on whether the crop was grown in the at least one field during at least one previous growing period (103), d) step (104) selecting at least one suitable agricultural field for the production of plant seedlings based on the information provided, wherein the crop has not been grown in the at least one agricultural field during at least one previous growing period. A computer implementation method (100), including the above.

2. The method according to claim 1, further comprising step a1) the processing unit determining a field boundary based on the image data using a field boundary detection model, wherein the field boundary detection model is machine learning.

3. The method according to claim 1, wherein the past satellite image data includes time-resolved image data over at least one previous growth period.

4. The method according to claim 3, wherein step b) determines from the image data time-resolved vegetation index data selected from normalized differential vegetation index (NDVI) data and / or leaf area index (LAI) data, normalized differential water index (NDWI), extended vegetation index (EVI) data and / or any other vegetation-based index data.

5. The method according to claim 4, wherein step b) includes classifying the at least one field by crop according to the time-resolved vegetation index data.

6. The method according to claim 5, wherein the classification is performed by using a classification model, and the classification model is obtainable by machine learning.

7. The method according to claim 6, wherein the machine learning is supervised machine learning and the training data is obtained by annotating satellite image data using ground truth data.

8. The method according to claim 1, wherein the image data is acquired via satellite using synthetic aperture radar (SAR) or light detection and ranging (LIDAR).

9. Image data of at least two agricultural fields is provided, and at least two of the agricultural fields are adjacent fields, or the boundaries of at least two of the agricultural fields are separated by a maximum of 10 km. The method according to claim 1, wherein the selection in step d) is also based on the information relating to the presence of the crop in a second of the at least one fields during a previous growing season.

10. The method according to claim 9, wherein image data of one agricultural field and all adjacent fields are provided, and the selection in step d) is based on the information relating to the presence of the crop in the field and all adjacent fields during a previous growing season.

11. The method according to claim 1, wherein the image data is provided for at least two previous growth periods.

12. A method for using past satellite image data of an agricultural field from a previous growing season as described in claim 1, in the method according to claim 1.

13. A system for selecting agricultural fields for producing plant seedlings of crops, wherein the system is a) A receiving unit configured to receive (101) historical image data of at least one agricultural field from at least one previous growing period, b) A processing unit, - Based on the image data, determine whether the crop was grown in the at least one field during at least one previous growing period (102), - Provide information on whether the crop was grown in the at least one field during at least one previous growing period (103), - Based on the information provided, select at least one suitable agricultural field for the production of plant seedlings (104), wherein the crop has not been grown in the at least one agricultural field during at least one previous growing period. A processing unit configured as follows A system that includes these features.

14. A computer program element comprising instructions configured to perform the steps of the method according to claim 1 in the system according to claim 13 when executed on a computing device of a computing environment.

15. A computer-readable medium storing the computer program elements described in claim 14.

Images