Control method, control device, agricultural machinery and processor for agricultural machinery

Abstract

The embodiment of the present invention provides a control method, a control device, an agricultural machine, a storage medium and a processor for agricultural machinery. The method includes: obtaining a first actual position of a first target object in a current area of ​​a farmland during the operation of the agricultural machine; determining a first working time interval of the sprinkler according to the first actual position; obtaining a second actual position of a second target object in a next area of ​​the farmland; determining a second working time interval of the sprinkler according to the second actual position; determining an actual working time interval of the sprinkler according to the first working time interval and the second working time interval, so as to quickly and accurately identify the actual position of the target object in the farmland and determine the actual working time interval of the sprinkler.

Description

Technical Field

[0001] The present invention relates to the field of agricultural technology, and in particular to a control method, a control device, an agricultural machine, a storage medium and a processor for agricultural machinery. Background Art

[0002] Survey statistics from the Crop Production Department of the Ministry of Agriculture and Rural Affairs show that the output of vegetables and grain crops in my country has grown steadily, but there are also some problems. Among them, weeds and insect pests are a major factor threatening crop yields and food safety.

[0003] The main method for large-scale control of agricultural weeds and pests is to use pesticides, which not only damages the environment, but also reduces the safety of agricultural products and increases production costs. For pesticide application, the commonly used methods mainly include manual judgment and treatment methods. The manual judgment method mainly uses the operator's expert experience to judge the type and location of weeds and pests in the field, and determines the appropriate time to open the nozzle for pesticide application. This method greatly increases the labor intensity of the operator. The operator needs to identify weeds and pests while driving agricultural machinery. The operation that requires driving and identification at the same time is very dependent on the operator's personal experience. As a result, the level of identification of weeds and pests is uneven, and there will be missed or over-spraying. The treatment method combines field images taken by drones or satellites to generate field operation prescription maps. During operation, agricultural machinery autonomously controls the amount of spraying according to the current position and prescription map. Although the treatment method realizes fully autonomous fertilization and pesticide application, it requires the generation of high-precision prescription maps through drones or satellite images and the construction of high-precision positioning systems for agricultural machinery. The economic cost is high and it is difficult to promote on a large scale. Summary of the invention

[0004] The purpose of the embodiments of the present invention is to provide a control method, a control device, an agricultural machine, a storage medium and a processor for agricultural machinery, which can quickly and accurately identify the actual position of a target object in a farmland and determine the actual working time interval of a sprinkler.

[0005] In order to achieve the above object, the first aspect of the present invention provides a control method for agricultural machinery, the agricultural machinery includes a sprinkler, and the control method includes:

[0006] Acquire a first actual position of a first target object in a current area of ​​a farmland during operation of the agricultural machine;

[0007] Determine a first working time interval of the nozzle according to the first actual position;

[0008] obtaining a second actual position of a second target object in a next area of ​​the farmland;

[0009] The second working time interval of the nozzle is determined according to the second actual position; and the actual working time interval of the nozzle is determined according to the first working time interval and the second working time interval.

[0010] In an embodiment of the present invention, determining the actual working time interval of the nozzle according to the first working time interval and the second working time interval includes: when the first working time interval and the second working time interval meet any one of the following conditions, determining the actual working time interval as the start time of the first working time interval to the end time of the second working time interval:

[0011] The first working time interval is continuous with the second working time interval;

[0012] The first working time interval overlaps with the second working time interval;

[0013] The time interval between the first working time interval and the second working time interval is smaller than a preset threshold.

[0014] In the embodiment of the present invention, the first working time interval is determined according to the first target object in the current area and the target object in the previous area.

[0015] In the embodiment of the present invention, there are multiple nozzles; determining the first working time interval of the nozzle according to the first actual position includes:

[0016] Determine a corresponding target nozzle among the plurality of nozzles according to the first actual position;

[0017] Determine an actual distance between the target object and the target nozzle according to the first actual position;

[0018] Obtain the moving speed and control delay of agricultural machinery;

[0019] The first working time interval of the target nozzle is determined according to the actual distance, the moving speed and the control delay.

[0020] In an embodiment of the present invention, determining a corresponding target nozzle among a plurality of nozzles according to the first actual position includes:

[0021] Divide the current area into a plurality of sub-areas corresponding to the plurality of sprinkler heads;

[0022] The nozzle corresponding to the sub-region where the first actual position is located is determined as the target nozzle.

[0023] In the embodiment of the present invention, determining the nozzle corresponding to the sub-region where the first actual position is located as the target nozzle includes:

[0024] Determine a plurality of sub-areas where the first actual position is located;

[0025] Determine the target sub-region according to the ratio of the area occupied by the first actual position in each sub-region to the area of ​​the first actual position;

[0026] The nozzle corresponding to the target sub-area is determined as the target nozzle.

[0027] In the embodiment of the present invention, determining the target sub-region according to the ratio of the area occupied by the first actual position in each sub-region to the area of ​​the first actual position includes:

[0028] Obtaining a first ratio and a second ratio of an area occupied by the first actual position in the adjacent first sub-region and the second sub-region respectively to an area of ​​the first actual position;

[0029] When the first ratio is higher than a first threshold and the second ratio is lower than a second threshold, the first sub-region is determined as a target sub-region.

[0030] In an embodiment of the present invention, the spraying range of the nozzle includes a first spraying range and a second spraying range, and the second spraying range is larger than the first spraying range; the method further includes:

[0031] The spraying range of the target nozzle corresponding to the target sub-area is switched from the first spraying range to the second spraying range.

[0032] In the embodiment of the present invention, determining the target sub-region according to the ratio of the area occupied by the first actual position in each sub-region to the area of ​​the first actual position further includes:

[0033] When the ratio of the first ratio to the second ratio is within a preset ratio interval, both the first sub-region and the second sub-region are determined as target sub-regions.

[0034] In an embodiment of the present invention, obtaining a first actual position of a first target object in a current area of ​​a farmland during the operation of the agricultural machinery includes: obtaining a regional image corresponding to the current area of ​​the farmland during the operation of the agricultural machinery; inputting the regional image into a perception machine model to determine an image position of the target object in the regional image in the regional image; obtaining camera calibration information of an image acquisition device used to capture the regional image; and determining a first actual position of the first target object based on the image position and the camera calibration information.

[0035] In an embodiment of the present invention, obtaining the camera calibration information of the image acquisition device used to capture the area image includes: calibrating the camera of the image acquisition device to obtain a plurality of coordinate data; and calibrating the camera of the image acquisition device through a preset algorithm and the plurality of coordinate data to determine the camera calibration information of the image acquisition device.

[0036] In an embodiment of the present invention, the camera of the image acquisition device is calibrated to obtain a plurality of coordinate data, including:

[0037] Placing the calibration plate at a first angle of the image acquisition device, and recording the first camera coordinates of each feature point of the calibration plate at the image acquisition device and the corresponding first world coordinates;

[0038] Placing the calibration plate at a second angle of the image acquisition device, and recording the second camera coordinates of each feature point of the calibration plate at the image acquisition device and the corresponding second world coordinates;

[0039] The calibration plate is placed at a third angle to the image acquisition device, and the third camera coordinates of each feature point of the calibration plate at the image acquisition device and the corresponding third world coordinates are recorded.

[0040] In an embodiment of the present invention, the method further includes one or both of the following: when the second target object is not detected in the next area of ​​the farmland and the end time corresponding to the first working time interval has been reached, turning off the sprinkler;

[0041] The earlier one between the start time of the first working time interval and the start time of the second working time interval is used as the start time of the nozzle.

[0042] A second aspect of the present invention provides a processor configured to execute the control method for agricultural machinery as described above.

[0043] A third aspect of the present invention provides a control device for agricultural machinery, the agricultural machinery includes a sprinkler, and the control device includes: an image acquisition device configured to acquire regional images of a farmland; and a processor according to the above.

[0044] A fourth aspect of the present invention provides an agricultural machine, comprising: a sprinkler; and a control device for the agricultural machine as described above.

[0045] A fifth aspect of the present invention provides a machine-readable storage medium having instructions stored thereon, the instructions being used to enable a machine to execute the above-mentioned control method for agricultural machinery.

[0046] The above-mentioned control method for agricultural machinery continuously determines the first target object in the current area and the second target object in the next area, as well as the first actual position of the first target object and the second actual position of the second target object during the operation of the agricultural machinery. The actual working time interval of the nozzle can be determined according to the first working time interval and the second working time interval, so that the nozzle can be opened and closed according to the actual working time interval, effectively avoiding the problem of continuous opening and closing of the nozzle due to continuous detection of target objects.

[0047] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present invention, but do not constitute a limitation on the embodiments of the present invention. In the accompanying drawings:

[0049] Figure 1 A diagram schematically shows an application environment of a method for determining a position of a target object according to an embodiment of the present invention;

[0050] Figure 2 A schematic diagram of a flow chart of a method for determining a position of a target object according to an embodiment of the present invention is shown;

[0051] Figure 3 A schematic diagram of a flow chart of a method for training a perceptual machine model according to an embodiment of the present invention is shown;

[0052] Figure 4 A schematic diagram of a flow chart of a control method for agricultural machinery according to an embodiment of the present invention is shown;

[0053] Figure 5 A schematic diagram of a flow chart of a control method for agricultural machinery according to another embodiment of the present invention is shown;

[0054] Fig. 6A A schematic diagram schematically shows the area division according to an embodiment of the present invention;

[0055] Figure 6B Schematically shows a schematic diagram of area distribution of a first actual position according to an embodiment of the present invention;

[0056] Figure 6C A schematic diagram schematically shows multiple spraying ranges of a spray head according to an embodiment of the present invention;

[0057] Figure 7 A schematic diagram of a flow chart of steps for determining a corresponding target nozzle among a plurality of nozzles according to a first actual position according to an embodiment of the present invention is schematically shown;

[0058] Figure 8 A schematic diagram of a flow chart of a method for controlling a sprinkler head of an agricultural machine according to an embodiment of the present invention is shown;

[0059] Fig. 9 A structural block diagram of an apparatus for determining a position of a target object according to an embodiment of the present invention is schematically shown;

[0060] Fig.10A schematic diagram of a structural block diagram of a control device for agricultural machinery according to an embodiment of the present invention is shown;

[0061] Fig.11 A structural block diagram of a control device for agricultural machinery according to another embodiment of the present invention is schematically shown;

[0062] Fig.12 A structural block diagram of an agricultural machine according to an embodiment of the present invention is schematically shown;

[0063] Fig.13 The internal structure of a computer device according to an embodiment of the present invention is schematically shown. DETAILED DESCRIPTION

[0064] The specific implementation of the embodiment of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the embodiment of the present invention, and is not used to limit the embodiment of the present invention.

[0065] The method for determining the position of a target object provided in this application can be applied to Figure 1 In the application environment shown. Figure 1 In the application environment shown, the agricultural machinery 100 may include an image acquisition device 101 and a perception machine model 102 (for example, a perception machine model included in or running by a processor). The image acquisition device 101 is used to acquire regional images of farmland and send the acquired regional images to the perception machine model 102 through network communication. The perception machine model 102 is used to determine the image position of the target object in the input regional image in the regional image, so that the processor of the agricultural machinery 100 can determine the actual position of the target object according to the image position and the camera calibration information of the image acquisition device 101. Examples of agricultural machinery 100 may include but are not limited to various human (manual) agricultural machinery, animal-powered agricultural machinery, small-powered agricultural machinery, tractor-supported agricultural machinery, self-propelled agricultural machinery, aviation agricultural machinery, etc. Examples of image acquisition devices 101 may include but are not limited to video cameras, cameras, video cameras, or various electronic devices with photo-taking functions.

[0066] Figure 2 The flowchart of the method for determining the position of a target object according to an embodiment of the present invention is schematically shown. Figure 2 As shown, in one embodiment of the present invention, a method for determining the position of a target object is provided, comprising the following steps:

[0067] Step 201, obtaining a regional image of a farmland.

[0068] Step 202, input the region image into a perceptual machine model to determine the image position of the target object in the region image, wherein the perceptual machine model is trained by image training samples generated by a generative adversarial network.

[0069] Step 203: Obtain camera calibration information of an image acquisition device used to capture the regional image.

[0070] Step 204: determine the actual position of the target object according to the image position and the camera calibration information.

[0071] The regional image of the farmland may refer to the image obtained by collecting images of the farmland through the image acquisition device on the agricultural machinery. Since the image collected by the image acquisition device is limited to a certain area range, the farmland image collected by the image acquisition device can be called a regional image. The target object may refer to the target of interest that needs to be identified. For example, there are many plants in the farmland, including rice, corn, wheat, weeds, etc. When it is necessary to perform farmland operations on rice, such as spraying fertilizers or pesticides, the target object in this scenario may be rice; when it is necessary to perform farmland operations on corn, such as spraying fertilizers or pesticides, the target object in this scenario may be corn; when it is necessary to remove weeds in the farmland and spray pesticides on the weeds, the target object may be weeds.

[0072] The captured regional image is input into the perception machine model, and the corresponding image position of the target object in the regional image is determined by the perception machine model. The image position refers to the corresponding position of the target object in the regional image when the target object is contained in the regional image. After determining the image position of the target object contained in the regional image, the camera calibration information of the image acquisition device can also be obtained, and the actual position of the target object in the farmland can be determined based on the image position of the target object and the camera calibration information. The camera calibration information refers to the camera parameters after the camera of the image acquisition device is calibrated, that is, the optimal calibration matrices R and T calculated after the camera calibration.

[0073] For example, if the target object is weeds, the position of the weeds in the regional image can be determined as (x1, y1), the position of the weeds relative to the image acquisition device can be determined as (X1, Y1, Z1), and the camera calibration information is the camera calibration matrix R and T. The position of the weeds relative to the image acquisition device, that is, the actual position of the weeds in the farmland, can be calculated by the following formula:

[0074]

[0075] In an embodiment of the present invention, the perceptual machine model can be trained by image training samples generated by a generative adversarial network. A generative adversarial network (GAN) is a deep learning model, referred to as a GAN network. Through a generative adversarial network, multiple image training samples can be randomly generated, so that the training sample data is more, the scenes involved are more and more complex, and the perceptual machine model trained in this way has a higher recognition accuracy when recognizing the input regional image.

[0076] The above-mentioned method for determining the position of the target object trains the perception machine model through image training samples generated by a generative adversarial network. After obtaining the regional image of the farmland, the regional image is input into the trained perception machine model to determine the image position of the target object in the regional image, and the actual position of the target object is finally determined based on the camera calibration information of the image acquisition device. In this method for determining the position of the target object, the perception machine model is trained through image training samples generated by a generative adversarial network, which effectively improves the recognition efficiency and accuracy of the perception machine model for the target object. Therefore, when the acquired regional image is input into the perception machine model, the target object contained in the regional image and the actual position of the target object in the farmland can be quickly and accurately determined.

[0077] In one embodiment, the perceptual machine model may be trained. Specifically, the perceptual machine model is trained by the following steps: obtaining a plurality of regional image samples; performing data enhancement on the regional image samples through a generative adversarial network to obtain a plurality of image training samples; and inputting the image training samples into the perceptual machine model to train the perceptual machine model.

[0078] Before the perceptual machine model is put into practical application, the perceptual machine model can be trained first. The trained perceptual machine model will have a higher recognition accuracy rate for the target object. Multiple regional image samples can be obtained first, and the regional image samples can be data enhanced by using a generative adversarial network to generate multiple random image training samples based on the input regional image samples through the generative adversarial network to increase the number of training samples of the perceptual machine model and the complexity of the training samples. Therefore, the image training samples can be input into the perceptual machine model to train the perceptual machine model.

[0079] In one embodiment, the generative adversarial network includes a generative network and a discriminative network. Performing data enhancement on the regional image samples through the generative adversarial network to obtain multiple image training samples includes: inputting the regional image samples into the generative network to generate multiple random regional images; inputting the random regional images into the discriminative network to generate a determination result on whether the random regional images contain a target object; the generative network adjusts the generation strategy of the random regional images according to the determination result to output multiple random image training samples according to the input regional image samples.

[0080] The generative adversarial network contains two networks, namely the generative network and the discriminative network. When the target object is weeds in farmland, the generative network and the discriminative network in the GAN network can be trained at the same time according to the characteristics of large target color difference and shape difference in the crop weed data set. Specifically, when the regional image sample is input into the GAN network, it actually means that the regional image sample is input into the generative network, and a plurality of random regional images are generated according to the regional image sample by the generative network. The random regional image output by the generative network is used as the input of the discriminative network, and the discriminative network judges whether the random regional image contains the target object. The generative network then adjusts the generation strategy of the random regional image according to the judgment result, and randomly generates a plurality of random regional images, and then inputs the generated random regional images into the discriminative network for recognition. The generative network and the discriminative network are constantly engaged in a game, and the recognition ability of the discriminative network is improved while the image generation ability of the generative network can also be improved. In this way, the generative adversarial network can perform data enhancement on the regional image samples and generate more image training samples with more complex scenes.

[0081] In one embodiment, inputting an image training sample into a perceptual machine model to determine an image position of a target object in the image training sample includes: inputting a regional image into a perceptual machine model to extract a plurality of regional image features in the regional image; determining whether the regional image contains the target object based on the plurality of regional image features; determining a feature position of the regional image feature corresponding to the target object when it is determined that the regional image contains the target object; and determining the image position of the target object in the regional image based on the feature position.

[0082] After the captured regional image is input into the perceptual machine model, the perceptual machine model can be used to extract features of the regional image to obtain multiple regional image features, and the perceptual machine model can determine whether the regional image contains the target object based on the regional image features. When the perceptual machine model determines that a certain regional image contains the target object, the feature position of the regional image feature belonging to the target object can be obtained, so that the image position of the target object in the regional image can be determined based on the feature position.

[0083] In one embodiment, determining whether a target object is contained in a regional image based on multiple regional image features includes: obtaining a labeling box and corresponding labeling box parameters of the regional image; determining whether the target object is contained in the labeling box; adjusting the labeling box parameters based on the determination result to determine a detection box for the target object; determining a probability that the regional image feature belongs to a feature of the target object based on the detection box; and determining whether the target object is contained in the regional image based on the probability.

[0084] The perceptual machine model can extract features from the regional image to obtain multiple regional image features. The perceptual machine model can also obtain the annotation box of the regional image and the annotation box parameters corresponding to each annotation box. The annotation box can be pre-annotated according to whether each regional image contains a target object and the position of the target object in the regional image when the target object exists. This annotation box can be manually annotated for each image training sample in advance, and the annotation box annotates one or more target objects contained in each image training sample and the position of the target object in the image training sample. The annotation box parameters refer to parameters such as the size and aspect ratio of the annotation box. After obtaining the annotation box of the regional image, the perceptual machine model can judge whether the annotation box contains the target object. Specifically, the perceptual machine model can judge whether the annotation box contains features belonging to the target object, and adjust the annotation box parameters according to the judgment result to determine the detection box for the target object. This determined detection box is used for the perceptual machine model to determine the probability that each regional image sample feature belongs to the feature of the target object, so that it can be determined whether the image training sample contains the target object according to the probability.

[0085] In one embodiment, determining the probability that the regional image sample features belong to the features of the target object based on the detection frame includes: performing border regression on the detection frame indicating that the features belonging to the target object are detected to correct the position of the detection frame that determines the presence of the features belonging to the target object; filtering the corrected detection frame according to preset conditions; and determining the probability that the regional image sample features belong to the features of the target object based on the filtered detection frame.

[0086] After determining the detection frame for the target object, the perception machine model can obtain the detection results of each detection frame and perform bounding box regression on the detection frame indicating that the feature belonging to the target object is detected. The purpose of bounding box regression is to correct the position of the detection frame. Furthermore, the corrected detection frame can be screened according to preset conditions, and the probability that the regional image sample feature belongs to the feature of the target object is determined based on the screened detection frame.

[0087] like Figure 3 As shown, in one embodiment of the present invention, a method for training a perceptual machine model is provided, comprising the following steps:

[0088] Step 301: Acquire multiple regional image samples.

[0089] Step 302: input the region image samples into a generative network in a generative adversarial network to generate a plurality of random region images.

[0090] Step 303: input the random region image into the discriminant network in the generative adversarial network to generate a determination result on whether the random region image contains the target object.

[0091] Step 304: The generation network adjusts the generation strategy of the random region image according to the determination result, so as to output a plurality of random image training samples according to the input region image samples.

[0092] When agricultural machinery is performing agricultural spraying operations, it is necessary to promptly identify the target object in the farmland and the specific location of the target object in the farmland in order to achieve accurate spraying operations. When the target object is weeds, since the crop weed data has the characteristics of large color difference and shape difference of the concentrated target, in order to improve the recognition accuracy and efficiency of the perception machine model for weeds, the acquired regional image samples can be input into the generative adversarial network, so that multiple image training samples can be randomly generated through the generative adversarial network. When the generative adversarial network generates image training samples, the generated images are more randomized, multi-scenario and more complex. For example, input a daytime farmland area image into the generative adversarial network, and the generative adversarial network can generate farmland area images on rainy days, cloudy days, farmland area images during heavy rain, farmland area images at night, etc. In this way, the data volume of the training samples of the perception machine model can be more diversified and more complex, thereby making the recognition accuracy and recognition efficiency of the perception machine model higher.

[0093] There are two networks in the generative adversarial network, namely the generative network and the discriminative network. Specifically, when a plurality of random region images are randomly generated by the generative adversarial network, the region image samples are input into the generative network in the generative adversarial network, and the generative network can generate a plurality of random region images according to the input region image samples, and the random region images are input into the discriminative network, and the discriminative network can judge whether the random region images contain the target object, and the generative network adjusts the generation strategy of the random region images according to the judgment result of the discriminative network. In other words, the goal of the generative network is to generate real pictures as much as possible to "cheat" the discriminative network, while the goal of the discriminative network is to distinguish the generated pictures from the real pictures as much as possible. Furthermore, the discriminative network also needs to identify the target objects contained in the pictures generated by the generative network, that is, the generative network and the discriminative network constitute a dynamic "game process", so that the generative network and the discriminative network are in a continuous optimization process, so that the generative network can continuously adjust its generation strategy according to the judgment structure of the discriminative network, and finally output a plurality of random image training samples according to the input region image samples.

[0094] In an embodiment of the present invention, after obtaining multiple regional image samples, the regional image samples can also be preprocessed, such as scaling, normalizing, and other processing operations on the regional image samples, and then the processed regional image samples are input into the generative adversarial network.

[0095] Step 305: input the image training sample into the perceptual machine model to extract multiple regional image sample features in the image training sample.

[0096] Step 306 , obtaining the annotation box of the image training sample and the corresponding annotation box parameters to determine whether the target object is included in the annotation box.

[0097] Step 307: Adjust the annotation box parameters according to the judgment result to determine the detection box for the target object.

[0098] After the generative adversarial network outputs multiple random image training samples according to the input regional image samples, the perceptual machine model can be trained by the random image training samples. Specifically, after the randomly generated image training samples are input into the perceptual machine model, the perceptual machine can extract the features of the image training samples to obtain multiple corresponding regional image sample features. During the training process, in order to improve the prediction accuracy of the perceptual machine model, the image training samples can be manually annotated in advance, and the target objects contained in each image training sample and the position of each target object in the image can be marked with an annotation box. The perceptual machine model can obtain the annotation box of each image training sample and the corresponding annotation box parameters, and judge whether each annotation box contains the target object. The process of feature extraction of the perceptual machine model and the process of obtaining the annotation box can be performed simultaneously, or the feature extraction can be performed first and then the annotation box can be obtained. The order of these two steps can be flexibly changed.

[0099] After the perception machine determines whether the labeled box contains the target object, it can adjust the parameters of the labeled box according to its judgment result to determine the detection box for the target object. The detection box refers to the detection box used by the perception machine model when identifying the target object. If the detection box is too large or too small, it may lead to inaccurate detection results of the target object. Therefore, the perception machine model can determine the size of the detection box it recognizes based on the result that the pre-labeled labeled box contains the target object.

[0100] Step 308 , performing bounding box regression on the detection box indicating the detected features belonging to the target object, so as to correct the position of the detection box.

[0101] Step 309: filter the corrected detection frame according to preset conditions.

[0102] Step 310: Determine the probability that the regional image sample features belong to the features of the target object based on the filtered detection frame.

[0103] After determining the size of the detection frame, the perceptual machine model can determine whether there are features belonging to the target object through the detection frame. Therefore, it is possible to obtain detection frames indicating that features belonging to the target object have been detected, and perform border regression on these detection frames to correct the position of the detection frames, so that the positioning of the detection frames is more accurate. Furthermore, after correcting the position of the detection frame, the corrected detection frame can be screened according to preset conditions. When the perceptual machine model determines whether there are features belonging to the target object through the detection frame, it can also obtain the probability of each detection frame determining that a feature belonging to the target object has been detected. After screening the detection frames, the probability that the regional image sample features belong to the features of the target object can be determined based on the screened detection frames and the probability that each detection frame believes that a feature belonging to the target object has been detected.

[0104] In one embodiment, screening the corrected detection frames according to preset conditions includes: retaining the corrected detection frames that meet the preset conditions, wherein the preset conditions include any one of the following: the probability value of the corrected detection frame is greater than a preset probability threshold; the size of the corrected detection frame is greater than a preset size; the area overlap ratio of any two detection frames in the corrected detection frame is lower than a preset overlap threshold.

[0105] When the corrected detection frame is screened, the screening can be performed in the following manner. In an example screening method, the probability value of each detection frame determining that the feature belonging to the target object is detected can be obtained, and a preset probability threshold is set to remove the detection frames with a probability value less than or equal to the preset probability threshold, and retain the detection frames with a probability value greater than the preset probability threshold. That is, when each detection frame detects whether it contains the feature belonging to the target object, it can not only confirm whether it contains the feature belonging to the target object or not, but also determine the probability of each detection frame determining that it contains the feature belonging to the target object when the detection frame determines that it contains the feature belonging to the target object. For example, the preset probability threshold is set to 55%, and there are 100 detection frames in total, of which 90 detection frames believe that the probability value of containing the feature belonging to the target object is higher than 55%, and the remaining 10 detection frames believe that the probability value of containing the feature belonging to the target object is less than or equal to 55%. Then, when screening the detection frames, the 10 detection frames with probability values ​​less than or equal to 55% can be removed, and only the 90 detection frames with probability values ​​higher than 55% are retained.

[0106] The second example filtering method can be filtering based on the size of the detection frame. Specifically, the size of each detection frame can be obtained and compared with a preset size. Since a detection frame that is too small may result in failure to detect relevant features, the detection frames whose size is less than or equal to the preset size can be removed, and the detection frames whose size is larger than the preset size can be retained.

[0107] The third example screening method can select two detection frames at random to determine whether their area overlap ratio is lower than a preset overlap threshold. When the area overlap ratio of the two detection frames is too high, one of the detection frames can be removed and only one of the two can be retained. Specifically, when selecting one of the two, the one with a higher probability value can be selected.

[0108] Step 311, determining whether the image training sample contains the target object according to probability.

[0109] Step 312: When it is determined that the target object is included in the image training sample, the feature position of the regional image sample feature corresponding to the target object is determined.

[0110] Step 313: determine the image position of the target object in the image training sample according to the feature position.

[0111] After determining the probability that the regional image sample features belong to the features of the target object according to the filtered detection frame, it is possible to determine whether the image training sample contains the target object according to the probability finally calculated. Specifically, a threshold can be set. When the probability finally calculated is greater than the threshold, the perceptual machine model can determine that the image training sample contains the target object, otherwise it is determined that the image training sample does not contain the target object. For example, the threshold is set to 65%. If the probability finally calculated is greater than 65%, the perceptual machine model can output a prediction result of 1, indicating that the input image training sample contains the target object; if the probability finally calculated is less than or equal to 65%, the perceptual machine model can output a prediction result of 0, indicating that the input image training sample does not contain the target object. In other words, the probability value of the detection frame is an intermediate detection process of the perceptual machine model. The final output of the perceptual machine model is 0 or 1, that is, yes or no. According to the structure of the output of the perceptual machine model, it can be immediately known whether the input image training sample contains the target object.

[0112] After determining which image training samples contain the target object and which do not contain the target object through the perceptual machine model, the perceptual machine model can be further used to obtain the regional image sample features that are detected to belong to the target object in the image training sample, as well as the feature positions of these regional image sample features in the image training sample, so that the image position of the target object in the image training sample can be determined according to the feature position. In other words, the perceptual machine model can not only output 0 or 1, but also output the image position of the target object in the image training sample when the output is 1, that is, when the input image training sample is detected to contain the target object. Therefore, by inputting the image training sample into the perceptual machine model, it is possible to know whether the image training sample contains the target object and the image position of the target object contained in the image training sample. Then, in the actual application of the perceptual machine model, after the acquired regional image of the farmland is input into the perceptual machine model, the perceptual machine model can also be used to confirm whether the regional image contains the target object, and determine the image position of the target object contained in the regional image in the regional image.

[0113] Figure 4 The flowchart of the control method for agricultural machinery according to an embodiment of the present invention is schematically shown. Figure 4 As shown, in one embodiment of the present invention, a control method for agricultural machinery is provided, comprising the following steps:

[0114] Step 401 : obtaining a first actual position of a first target object in a current area of ​​a farmland during an agricultural machinery operation.

[0115] Step 402: Determine a first working time interval of the nozzle according to the first actual position.

[0116] Step 403: Acquire a second actual position of a second target object in a next area of ​​the farmland.

[0117] Step 404: determine a second working time interval of the nozzle according to the second actual position.

[0118] Step 405: determining the actual working time interval of the nozzle according to the first working time interval and the second working time interval.

[0119] During the operation of agricultural machinery, it is necessary to identify the target object in real time and determine the actual position of the target object in the farmland to facilitate precise operation. In this embodiment, the first target object and the second target object can be represented as the same target object. For the sake of convenience, the target object contained in the current area can be referred to as the first target object, and the target object contained in the next area can be referred to as the second target object. Furthermore, if there are two or more target objects, such as the target object can be a weed or a pest, then the first target object and the second target object can be simultaneously represented as a weed or a pest. Correspondingly, the actual position of the first target object in the farmland is referred to as the first actual position, and the actual position of the second target object in the farmland is referred to as the second actual position.

[0120] The division of the current area and the next area of ​​the farmland is determined by the image acquisition area of ​​the image acquisition device included in the agricultural machinery. In other words, the current area refers to the area range that the image acquisition device can currently capture when the agricultural machinery is in operation. When the agricultural machinery continues to move, the area range captured by the image acquisition device when it performs image acquisition again is the next area.

[0121] Specifically, the image acquisition device in the agricultural machinery continuously acquires images of the crops in the farmland, and performs image recognition and positioning to determine whether the acquired farmland area image contains the target object. If it is determined that the target object is contained, the actual position of the target object contained in the farmland is determined. The target object contained in the regional image acquired at the current location of the agricultural machinery is called the first target object, so the first actual position of the first target object in the current area can be determined, and the first working time interval of the nozzle can be determined according to the first actual position. That is to say, after determining the actual position of the first target object, the distance between the first target object and the agricultural machinery can be calculated, and according to the distance and the travel speed of the agricultural machinery, it can be calculated how long it will take for the agricultural machinery to travel to the position of the first target object, so at what time the nozzle is allowed to start the spraying operation and at what time the spraying operation is ended, so that the area where the first target object is located can be sprayed accurately. Then, according to the first actual position of the first target object, the first working time interval of the nozzle, that is, the opening time and the ending time of the nozzle can be determined.

[0122] In an embodiment of the present invention, the first target object actually occupies a certain area. Determining the actual position of the first target object can actually determine the actual position of the first target object occupying a certain area. In one example, the distances of the near point and far point of the first target object relative to the agricultural machinery (e.g., sprinkler) along the travel direction of the agricultural machinery can be calculated respectively, and the time it will take for the agricultural machinery to travel to the near point and far point of the first target object can be calculated respectively based on these two distances and the travel speed of the agricultural machinery, so as to determine the start time and end time of the sprinkler respectively.

[0123] In one embodiment of the present invention, in order to facilitate processing, especially for the case where the shape of the first target object in the farmland is an irregular shape, a regular shape can be used to fit, approximate or replace the shape of the first target object. Examples of regular shapes may include, but are not limited to, rectangles, circles, and ellipses. In this embodiment, the distances of the near point and far point of the regular shape relative to the agricultural machinery (e.g., the sprinkler) along the direction of travel of the agricultural machinery can be calculated respectively, and the time it will take for the agricultural machinery to travel to the near point and far point of the regular shape can be calculated respectively based on these two distances and the travel speed of the agricultural machinery, so as to determine the start time and end time of the sprinkler respectively.

[0124] In the process of continuous operation of agricultural machinery, the agricultural machinery is constantly moving forward, and the image acquisition device will also continuously acquire images. If the regional image acquired by the image acquisition device contains a target object, it can be considered that the next region contains a second target object, and the second actual position of the second target object in the farmland can be determined in the same manner as for the first target object as described above, and at the same time, the second working time interval for the sprinkler to spray the second target object is determined, that is, the start time and end time when the sprinkler sprays the second target object. Specifically, the above-mentioned method for determining the position of the target object can be used when identifying and locating the target object through the image acquired by the image acquisition device to determine the first target object included in the current region and the second target object included in the next region, as well as determining the first actual position of the first target object and the second actual position of the second target object.

[0125] After determining the first target object and the second target object contained in the two areas, as well as their actual positions in the farmland, the actual working time interval of the nozzle can be determined according to the position of the target object. In the actual process, if the area corresponding to the same nozzle continuously detects the target object, the nozzle will continuously receive the control command to open, which will cause the nozzle to be repeatedly opened and closed, thereby wasting a lot of pesticides and greatly reducing the life of the nozzle. Therefore, in order to solve the problem of repeated opening and closing of the nozzle, the control instructions of the same nozzle can be fused. In other words, for the same nozzle, if the current area and the next area corresponding to its spraying range both detect the target object, the first working time interval of the first target object and the second working time interval of the second target object can be fused to determine the actual working time interval of the nozzle, so that the nozzle is opened and closed according to the actual working time interval.

[0126] In the embodiment of the present invention, when any one of the following conditions is met, the actual working time interval can be determined as the time from the start time of the first working time interval to the end time of the second working time interval.

[0127] In one condition, the first working time interval is continuous with the second working time interval. For example, assuming that the first working time interval is [t 1 ,t 2 ], where t 1 is the nozzle opening time, t 2 is the nozzle closing time, and the second working time interval is [t 3 ,t 4 ], where t 3 is the nozzle opening time, t 4 is the nozzle closing time. If t 2 Equal to t 3 (t 2 =t 3 ), indicating that the first working time interval is continuous with the second working time interval. In this case, the first working time interval and the second working time interval can be merged to determine the actual working time interval as [t 1 ,t 4 ]. That is to say, cancel the original 2 Send a closing command to the nozzle at all times (of course, the nozzle remains open, thus canceling the original 3 Send an opening command to the nozzle at all times) to keep the nozzle open until t 4 time.

[0128] In another condition, the first working time interval overlaps with the second working time interval. For example, assuming that the first working time interval is [t 1 ,t 2], where t 1 is the nozzle opening time, t 2 is the nozzle closing time, and the second working time interval is [t 3 ,t 4 ], where t 3 is the nozzle opening time, t 4 is the nozzle closing time. If t 2 Greater than t 3 (t 2 >t 3 ), indicating that the first working time interval overlaps with the second working time interval. In this case, the first working time interval and the second working time interval can be merged to determine the actual working time interval as [t 1 ,t 4 ]. That is to say, cancel the original 2 Send a closing command to the nozzle at all times (of course, the nozzle remains open, thus canceling the original 3 Send an opening command to the nozzle at all times) to keep the nozzle open until t 4 time.

[0129] In the third condition, the time interval between the first working time interval and the second working time interval is less than a preset threshold. 1 ,t 2 ], where t 1 is the nozzle opening time, t 2 is the nozzle closing time, and the second working time interval is [t 3 ,t 4 ], where t 3 is the nozzle opening time, t 4 is the nozzle closing time. If t 2 Less than t 3 (t 2 <t 3 ), indicating that there is a time interval between the first working time interval and the second working time interval, that is, t 2 With t 3 The time interval between 3 –t 2 ). In this case, usually at t 2 Send a closing command to the nozzle at time t to close the nozzle. 3Send an opening instruction to the sprinkler head at all times to turn on the sprinkler head. However, if the time interval is relatively small, for example, less than a preset threshold value (for example, the preset threshold value may be 1s, 0.8s, 0.6s, 0.5s, etc.), even if the sprinkler head is kept in the open state during this time interval, it will not have much impact on the spraying operation (for example, spraying on non-target objects will cause waste), and the first working time interval may also be merged with the second working time interval.

[0130] In the above-mentioned control method for agricultural machinery, during the operation of the agricultural machinery, the first target object in the current area and the second target object in the next area, as well as the first actual position of the first target object and the second actual position of the second target object are continuously determined, and the actual working time interval of the nozzle can be determined according to the first working time interval and the second working time interval, so that the nozzle can be opened and closed according to the actual working time interval, effectively avoiding the problem of continuous opening and closing of the nozzle due to continuous detection of target objects, thereby wasting a large amount of pesticides and greatly reducing the life of the nozzle.

[0131] In one embodiment, obtaining a first actual position of a first target object in a current area of ​​a farmland during an agricultural machinery operation includes: obtaining a regional image corresponding to the current area of ​​the farmland during the agricultural machinery operation; inputting the regional image into a perception machine model to determine an image position of the target object in the regional image in the regional image; obtaining camera calibration information of an image acquisition device used to capture the regional image; and determining a first actual position of the first target object based on the image position and the camera calibration information.

[0132] That is to say, in this embodiment, the above method for determining the position of a target object may be used to determine the first actual position of the first target object. The specific process will not be described in detail here.

[0133] In one embodiment, obtaining camera calibration information of an image acquisition device used to capture area images includes: calibrating the camera of the image acquisition device to obtain multiple coordinate data; and calibrating the camera of the image acquisition device through a preset algorithm and multiple coordinate data to determine the camera calibration information of the image acquisition device.

[0134] Before using an image acquisition device to capture an image, the camera of the image acquisition device may be calibrated first, so as to avoid barrel or pincushion distortion of the captured image, and the calibrated camera parameters may be referred to as camera calibration information. Specifically, the camera of the image acquisition device may be calibrated first, for example, a checkerboard calibration plate may be used to calibrate the camera.

[0135] In one embodiment, the camera of an image acquisition device is calibrated to obtain multiple coordinate data, including: placing a calibration plate at a first angle of the image acquisition device, recording the first camera coordinates of each feature point of the calibration plate in the image acquisition device and the corresponding first world coordinates; placing the calibration plate at a second angle of the image acquisition device, recording the second camera coordinates of each feature point of the calibration plate in the image acquisition device and the corresponding second world coordinates; placing the calibration plate at a third angle of the image acquisition device, recording the third camera coordinates of each feature point of the calibration plate in the image acquisition device and the corresponding third world coordinates.

[0136] When calibrating the camera, the angle between the calibration plate and the camera of the image acquisition device can be adjusted to record multiple sets of data, so as to determine the calibration of the camera of the image acquisition device. Specifically, the calibration plate can be placed at a first angle of the image acquisition device to record the first camera coordinates of each feature point of the calibration plate in the image acquisition device and the corresponding first world coordinates; the calibration plate can be placed at a second angle of the image acquisition device to record the second camera coordinates of each feature point of the calibration plate in the image acquisition device and the corresponding second world coordinates; the calibration plate can be placed at a third angle of the image acquisition device to record the third camera coordinates of each feature point of the calibration plate in the image acquisition device and the corresponding third world coordinates. Among them, the first angle can refer to the angle directly in front of the image acquisition device, that is, it is located on the same vertical horizontal line as the shooting direction of the lens of the image acquisition device. Correspondingly, the coordinates (x1, y1) of each feature point of the calibration plate under this angle in the camera image coordinates, that is, the first camera coordinates, and the world coordinates (X1, Y1, Z1) relative to the camera, that is, the first world coordinates. The second angle may be 45 degrees to the left front of the image acquisition device, and the coordinates of each feature point of the calibration plate in the camera image coordinates at this angle, i.e., the second camera coordinates, and the world coordinates relative to the camera, i.e., the second world coordinates, are recorded. The third angle may be 45 degrees to the right front of the image acquisition device, and the coordinates of each feature point of the calibration plate in the camera image coordinates at this angle, i.e., the third camera coordinates, and the world coordinates relative to the camera, i.e., the third world coordinates, are recorded. Furthermore, more angles may be set for calibration and recording.

[0137] After calibrating the camera at multiple angles to obtain multiple coordinate data, the multiple coordinate data can be calculated using a preset algorithm to obtain an optimal camera calibration matrix, that is, to determine the camera calibration information of the image acquisition device. Specifically, the preset algorithm can be a least squares method.

[0138] In one embodiment, the first working time interval is determined according to the first target object in the current area and the target object in the previous area.

[0139] The agricultural machinery is constantly moving forward during the operation process. Therefore, the first working time interval of the sprinkler determined according to the first target object contained in the current area is actually obtained by integrating the working time intervals of the target objects in the current area and the previous area. In other words, the first working time interval of the sprinkler for spraying the first target object in the current area is actually determined according to the actual position of the target object contained in the previous area and the actual position of the target object contained in the current area.

[0140] In one embodiment, the number of nozzles may be multiple; determining the first working time interval of the nozzle according to the first actual position includes: determining the corresponding target nozzle among the multiple nozzles according to the first actual position; determining the actual distance between the first target object and the target nozzle according to the first actual position; obtaining the moving speed of the agricultural machinery and the control delay of the agricultural machinery; determining the first working time interval of the target nozzle according to the actual distance, moving speed and control delay.

[0141] There may be multiple nozzles included in the agricultural machinery, and different nozzles can perform spraying operations for different spraying areas. When determining the first working time interval of the nozzle according to the first actual position, the corresponding target nozzle can be first determined from the multiple nozzles according to the first actual position, and the actual distance between the first target object and the target nozzle can be further determined. At the same time, the moving speed of the agricultural machinery and the control delay of the agricultural machinery can also be obtained, so that the first working time interval of the target nozzle can be determined according to the actual distance, moving speed and control delay. Since the nozzle is turned on or off for spraying according to the received instructions, the control delay needs to be taken into account when considering the actual opening or closing time of the nozzle. If the control delay is not considered, it may result in that when the opening instruction is sent to the nozzle in the 10th second, the nozzle will actually start the spraying operation at the 12th second. In this way, some of the target objects that need to be sprayed may be missed. Therefore, when determining the first working time interval of the target nozzle, the control delay can be taken into account.

[0142] In one embodiment, determining a target nozzle corresponding to a plurality of nozzles according to the first actual position includes: dividing a current area into a plurality of sub-areas corresponding to the plurality of nozzles; and determining a nozzle corresponding to the sub-area where the first actual position is located as a target nozzle.

[0143] In agricultural machinery, multiple image acquisition devices may be included, and each image acquisition device may correspond to multiple nozzles. For each image acquisition device, the area corresponding to the image currently captured by the agricultural machinery during operation can be called the current area. The current area can be divided into multiple sub-areas according to the spraying range of each nozzle, wherein each sub-area corresponds to a nozzle. According to actual needs, it can also be set that each sub-area corresponds to multiple nozzles. Then, according to the first actual position of the first target object contained in the current area, the sub-area where it is located is determined, so that the corresponding nozzle, that is, the target nozzle, can be determined.

[0144] In one embodiment, the control method further includes: when the second target object is not detected in the next area of ​​the farmland and the end time corresponding to the first working time interval has been reached, turning off the sprinkler.

[0145] If no second target object is detected in the area image corresponding to the next area of ​​the farmland, it can be detected whether the sprinkler has reached the end time corresponding to the first working time interval. If so, the processor of the agricultural machinery can send a closing command to the sprinkler to close the sprinkler and end the current spraying operation.

[0146] In one embodiment, the control method further comprises: using the earlier one of the start time of the first working time interval and the start time of the second working time interval as the start time of the nozzle.

[0147] After the first working time interval and the second working time interval are integrated to determine the final actual working time interval, the start time of the nozzle can be further determined. Specifically, the earlier of the start time of the first working time interval and the start time of the second working time interval can be used as the start time of the nozzle. For example, the first target object detected in the current area is located at a position far away from the image acquisition device, and the second target object detected in the regional image of the next area is located at a position close to the image acquisition device. This situation may occur due to missed detection or shooting blind spots of the image acquisition device. Then, using the earlier of the starting times as the start time of the nozzle can effectively avoid missing the target object in the current area, and the second target object detected in the next area can be sprayed in time, thereby avoiding missed spraying and effectively improving the accuracy of spraying.

[0148] like Figure 5 As shown, in one embodiment of the present invention, a control method for agricultural machinery is provided, comprising the following steps:

[0149] Step 501 : obtaining a first actual position of a first target object in a current area of ​​a farmland during an agricultural machinery operation.

[0150] Step 502: divide the current area into a plurality of sub-areas corresponding to a plurality of nozzles.

[0151] Step 503: determine the nozzle corresponding to the sub-region where the first actual position is located as the target nozzle.

[0152] Step 504: determine the actual distance between the first target object and the target nozzle according to the first actual position.

[0153] Step 505: Obtain the moving speed of the agricultural machinery and the control delay of the agricultural machinery.

[0154] Step 506: determine the first working time interval of the target nozzle according to the actual distance, the moving speed and the control delay.

[0155] Step 507: Acquire a second actual position of a second target object in a next area of ​​the farmland.

[0156] Step 508: Determine a second working time interval of the nozzle according to the second actual position.

[0157] Step 509: determining the actual working time interval of the nozzle according to the first working time interval and the second working time interval.

[0158] During the operation of agricultural machinery, as the agricultural machinery continues to move forward, the image acquisition device will also continuously acquire images of the farmland. For the current moment, the image acquired by the image acquisition device is called the image corresponding to the current area, and the next acquired area image at the current moment is the image corresponding to the next area. Specifically, after the image acquisition device acquires the area image corresponding to the current area of ​​the farmland, the area image can be input into the perception machine model to determine the first target object contained in the area image of the current area and the first actual position of the first target object in the farmland. The area image corresponding to the next area can also be input into the perception machine model to determine the second target object contained in the area image of the next area and the second actual position of the second target object in the farmland.

[0159] For a certain image acquisition device, after determining the first actual position of the first target object contained in the regional image captured by the image acquisition device, the shooting region of the image acquisition device can be divided, that is, the current region can be divided into multiple sub-regions. Fig. 6AAs shown, the current area is divided into 4 sub-areas, namely A1, A2, A3 and A4, and the corresponding nozzles are N1, N2, N3 and N4. Among them, the first target object is detected in the A3 area, so the nozzle N3 corresponding to the A3 area can be determined as the target nozzle. Furthermore, the actual distance d between the first target object and the target nozzle N3 can be determined according to the actual position of the first target object. The actual distance d can refer to the longitudinal distance between the first target object and the target nozzle N3, At the same time, the moving speed V and control delay T of the agricultural machinery can also be obtained. delay , and calculate the opening time of the target nozzle N3 Since the nozzle has a fixed opening and closing time, for example, when the opening and closing time of the nozzle is 100ms, the closing time t of the target nozzle N3 is close =t open +0.1, then the first working time interval of the target nozzle N3 is As the agricultural machinery continues to move forward during the operation process, image acquisition and identification can be performed on the next area of ​​the current area. When the next area also contains a second target object, the second actual position of the second target object can be determined by the method of determining the actual position of the first target object, and the second working time interval of the target nozzle can be determined according to the second actual position by using the above-mentioned method for determining the first working time interval.

[0160] Assume that the first working time interval is [t1, t2] and the second working time interval is [t3, t4]. At this time, the actual working time interval of the nozzle can be determined based on the first working time interval and the second working time interval. Unlike the traditional control algorithm that directly sends the nozzle opening command, the control of the nozzle in this embodiment will be different. In this embodiment, the first working time interval and the second working time interval can be fused, that is, the actual working time interval finally determined is T = [t1, t2] ∪ [t3, t4]. Assume that [t1, t2] is [1s, 3s] and [t3, t4] is [2s, 4s]. After merging these two time intervals, the final actual working time interval can be obtained as [1s, 4s]. That is to say, after the target nozzle completes the first working time interval, it will not be closed, but will directly perform the spraying work on the second target object until the 4th second when the nozzle is closed to end the spraying operation.

[0161] If, when spraying the second target object, the image acquisition device detects a target object in the area image corresponding to the next area, the actual position of the target object is determined, and the working time interval corresponding to the detected target object is determined, and further fused with the last determined working time interval [1s, 4s] to re-determine the final actual working time interval of the target nozzle. In other words, the first working time interval of the target nozzle is actually determined based on the actual position of the target objects in the current area and the previous area, and when the target nozzle is performing the spraying operation on the first target object, the actual working time interval determined is a fusion of the working time interval of the target nozzle in the previous area and the first working time interval of the current area, as well as the second working time interval of the next area. In other words, assuming that the current area is the fourth area of ​​agricultural machinery operation, then when operating in the fourth area, the fourth area can be regarded as the current area, then the first working time interval of the target nozzle corresponding to the current area is actually determined by combining the target objects detected by the first three areas and the fourth area and the actual position of the target objects, and fusing multiple spraying control instructions of the target nozzle. That is, in this embodiment, the nozzle is closed according to the end time of the actual working time interval. If the end time of the actual working time interval is not reached, the closing instruction for the target nozzle will not be triggered, that is, the problem of repeated opening and closing of the nozzle will not occur.

[0162] It may happen that the determined first target object is located in multiple sub-areas. In this case, in an embodiment of the present invention, determining the nozzle corresponding to the sub-area where the first actual position is located as the target nozzle includes: determining multiple sub-areas where the first actual position is located; determining the target sub-area according to the ratio of the area occupied by the first actual position in each sub-area to the area of ​​the first actual position; and determining the nozzle corresponding to the target sub-area as the target nozzle.

[0163] For the first actual position where the first target object is located, the sub-region where the first actual position is located can be determined. When there are multiple sub-regions where the first actual position is located, that is, when there are multiple regions where the actual area of ​​the first actual position is located, the target sub-region can be determined according to the ratio of the area occupied by the first actual position in each sub-region to the area of ​​the first actual position, and the corresponding nozzle in the target sub-region is determined as the target nozzle.

[0164] In one embodiment, determining the target sub-region based on the ratio of the area occupied by the first actual position in each sub-region to the area of ​​the first actual position includes: obtaining a first ratio and a second ratio of the area occupied by the first actual position in adjacent first and second sub-regions, respectively, to the area of ​​the first actual position; and determining the first sub-region as the target sub-region when the first ratio is higher than a first threshold and the second ratio is lower than a second threshold. Figure 6B As shown, the first actual position is the shaded area in the figure, and the actual area of ​​the first actual position is distributed in the sub-areas corresponding to the nozzles N1 and N2. Assume that the range area of ​​the first actual position is S, S = 1 square meter. Among them, the area occupied by the first actual position in the first sub-area is S1, S1 = 0.8 square meters, and the area occupied by the first actual position in the second sub-area is S2, S2 = 0.2 square meters, then the first ratio is the ratio of the area occupied by the first actual position in the first sub-area to the area of ​​the first actual position, that is, the first ratio = S1 / S = 0.8. The second ratio is the ratio of the area occupied by the first actual position in the second sub-area to the area of ​​the first actual position, that is, the second ratio = S2 / S = 0.2. Among them, the first threshold is set to 75%, and the second threshold is set to 25%. Therefore, the above-mentioned first ratio 0.8 is higher than the first threshold 75%, and the second ratio 0.2 is lower than the second threshold 25%. In this case, the first sub-area corresponding to the nozzle N1 can be determined as the target sub-area.

[0165] In one embodiment, determining the target sub-region based on the ratio of the area occupied by the first actual position in each sub-region to the area of ​​the first actual position also includes: when the ratio of the first ratio to the second ratio is within a preset ratio interval, determining both the first sub-region and the second sub-region as target sub-regions.

[0166] Assume that there are multiple areas where the actual area of ​​the first actual position is located. For example, there are two areas where the actual area of ​​the first actual position is located, namely the first sub-area and the second sub-area, and the first sub-area is adjacent to the second sub-area. The processor of the agricultural machinery can obtain the first ratio and the second ratio of the area occupied by the first actual position in the first sub-area and the second sub-area respectively to the area of ​​the first actual position. When the first ratio is higher than the first threshold and the second ratio is lower than the second threshold, the first sub-area can be determined as the target sub-area. When the ratio of the first ratio to the second ratio is within a preset ratio interval, the first sub-area and the second sub-area are both determined as target sub-areas. Among them, the first threshold and the second threshold are both custom settings. Usually, the first threshold + the second threshold = 1, and it can also be set according to actual needs. At this time, the first threshold + the second threshold = 1 is not required.

[0167] In one embodiment, the spraying range of the nozzle includes a first spraying range and a second spraying range, and the second spraying range is larger than the first spraying range. The above method also includes: switching the spraying range of the target nozzle corresponding to the target sub-area from the first spraying range to the second spraying range.

[0168] Each nozzle can also be provided with a plurality of spraying gears, each spraying gear corresponding to a different spraying range. The nozzle includes at least two spraying gears, corresponding to a first spraying range and a second spraying range, respectively, and the second spraying range is larger than the first spraying range. Figure 6C A schematic diagram of the spraying range of the nozzle is shown. Specifically, as shown in the figure, the nozzle N1 includes two spraying gears, which correspond to the first spraying range and the second spraying range respectively. The spraying range corresponding to the gray dotted line in the figure is the first spraying range, and the spraying range corresponding to the black dotted line is the second spraying range. When the first ratio is higher than the first threshold value and the second ratio is lower than the second threshold value, the first sub-area can be determined as the target sub-area. In this case, only the nozzles in the first sub-area will be turned on, and the nozzles in the second sub-area will not be turned on. Therefore, the spraying range of the nozzles in the first sub-area can be switched from the first spraying range to the second spraying range to expand the spraying range of the nozzles in the first sub-area, which can effectively save related spraying agents, such as pesticides or fertilizers, etc. At the same time, the nozzles can be flexibly controlled according to the actual area of ​​the first target object to achieve more precise control of the nozzles.

[0169] In one embodiment, Figure 7 As shown, the step of determining a corresponding target nozzle among a plurality of nozzles according to the first actual position includes:

[0170] Step 701: divide the current area into multiple sub-areas corresponding to multiple nozzles.

[0171] Step 702: determine a plurality of sub-areas where the first actual location is located.

[0172] Step 703: Obtain a first ratio and a second ratio of the area occupied by the first actual position in the adjacent first sub-region and second sub-region respectively to the area of ​​the first actual position.

[0173] Step 704 , determining whether the first ratio is higher than a first threshold and the second ratio is lower than a second threshold, if so, executing step 705 ; if not, executing step 708 .

[0174] Step 705: determine the first sub-region as the target sub-region.

[0175] Step 706: determine the nozzle corresponding to the first sub-area as the target nozzle.

[0176] Step 707: Switch the spraying range of the target nozzle corresponding to the target sub-area from the first spraying range to the second spraying range.

[0177] Step 708 , determining whether the ratio of the first ratio to the second ratio is within a preset ratio interval, if so, executing step 709 ; if not, executing step 711 .

[0178] Step 709: determine both the first sub-region and the second sub-region as target sub-regions.

[0179] Step 710: determine the nozzles corresponding to the first sub-area and the second sub-area as target nozzles.

[0180] Step 711: determine the second sub-region as the target sub-region.

[0181] Step 712: determine the nozzle corresponding to the second sub-area as the target nozzle.

[0182] After determining the first actual position of the first target object, the control area corresponding to the image acquisition device, that is, the current area, can be divided into sub-areas corresponding to multiple nozzles, and the sub-area where the first actual position is located can be determined. When there are multiple sub-areas where the first actual position is located, the first sub-area and the second sub-area where the first actual position is specifically located can be obtained. Assume that the range area of ​​the first actual position is S. Among them, the area occupied by the first actual position in the first sub-area is S1, and the area occupied by the first actual position in the second sub-area is S2, then the first ratio is the ratio of the area occupied by the first actual position in the first sub-area to the area of ​​the first actual position, that is, the first ratio = S1 / S. The second ratio is the ratio of the area occupied by the first actual position in the second sub-area to the area of ​​the first actual position, that is, the second ratio = S2 / S. It can be understood that the first ratio + the second ratio = 1.

[0183] Furthermore, the first ratio and the second ratio can be compared. When the first ratio is higher than the first threshold and the second ratio is lower than the second threshold, it can be considered that the area occupied by the first actual position in the first sub-region is larger, and the first sub-region can be considered as the sub-region corresponding to the first actual position, and the second sub-region with a smaller area can be ignored. Therefore, the first sub-region can be determined as the target sub-region, and the nozzle corresponding to the first sub-region can be determined as the target nozzle. For example, the first ratio is 95%, the second ratio is 5%, the first threshold is set to 94%, and the second threshold is set to 6%. At this time, the first ratio of 95% is higher than the first threshold of 94% and the second ratio of 5% is lower than the second threshold of 6%, so the first sub-region can be determined as the target sub-region, and the nozzle corresponding to the first sub-region can be determined as the target nozzle.

[0184] The spraying range of the target nozzle corresponding to the target sub-area can also be switched from the first spraying range to the second spraying range. That is to say, at this time, the area range corresponding to the first actual position actually occupies two sub-areas, but in the end, based on the area ratio, the first sub-area is used as the sub-area where the first actual position is located, and the second sub-area is ignored. In this case, the processor will initiate a nozzle opening command to the target nozzle corresponding to the first sub-area, while the nozzle corresponding to the second sub-area will not receive the opening control command, that is, the nozzle corresponding to the second sub-area will not start the spraying operation. In order to allow the target object actually located in the area where the second sub-area is located to be covered by spraying, the spraying range of the target nozzle in the first sub-area can be increased, and the spraying range can be switched from the first spraying range to the second spraying range, so that when the second sub-area also includes the range occupied by the first actual position, it is only necessary to open the target nozzle in the first sub-area, and all target objects contained in all the first actual positions can be sprayed.

[0185] If the first ratio and the second ratio do not satisfy the above conditions, that is, the first ratio and the second ratio cannot simultaneously satisfy the conditions that the first ratio is higher than the first threshold and the second ratio is lower than the second threshold, the ratio of the first ratio to the second ratio can be judged. If the ratio of the first ratio to the second ratio is within the preset ratio interval, it can be considered that the area occupied by the first actual position in the first sub-region and the second sub-region is approximately the same. In this case, the first sub-region and the second sub-region can be determined as target sub-regions, and the nozzles corresponding to the first sub-region and the second sub-region can be determined as target nozzles. If the first ratio and the second ratio cannot simultaneously satisfy the conditions that the first ratio is higher than the first threshold and the second ratio is lower than the second threshold, and the ratio of the first ratio to the second ratio is not within the preset ratio interval, the second sub-region can be determined as the target sub-region, and the nozzle corresponding to the second sub-region can be determined as the target nozzle.

[0186] like Figure 8As shown in the flowchart, first, the image collected by the image acquisition device can be input, and the image can be input into the perception machine model, and the target object contained in the image can be determined by the perception algorithm, and the image position of the target object in the image can be determined. Then, the actual position of the target object in the farmland is determined according to the image position and the camera calibration information to determine the corresponding target sub-area, and the corresponding target nozzle is determined according to the target sub-area. At the same time, the working time interval of the target nozzle can be calculated according to the actual distance between the target object and the target nozzle, the moving speed of the agricultural machinery and the control delay of the agricultural machinery, that is, the calculated spraying area information in the figure. The processor can obtain the number of the target nozzle, and send an opening instruction to the target nozzle according to the starting time corresponding to the working time interval, so that the target nozzle can start the spraying operation according to the opening instruction. At the same time, the target object in the next area is continuously identified and located, and multiple control instructions for the same nozzle are merged to determine the actual working time interval of the nozzle. The vehicle controller refers to the processor of the agricultural machinery, which can control the nozzle accordingly according to the current state of the nozzle.

[0187] In one embodiment, Fig. 9 As shown, a device for determining the position of a target object is provided, comprising an image acquisition module, an image position confirmation module and an actual position determination module, wherein:

[0188] The image acquisition module 901 is configured to acquire regional images of farmland.

[0189] The image position confirmation module 902 is configured to input the region image into the perception machine model to determine the image position of the target object in the region image, and the perception machine model is trained by image training samples generated by a generative adversarial network.

[0190] The actual position determination module 903 is configured to obtain camera calibration information of an image acquisition device used to capture regional images; and determine the actual position of the target object according to the image position and the camera calibration information.

[0191] In one embodiment, Fig. 9 As shown, the above-mentioned device for determining the position of the target object also includes a model training module 904, which is configured to: obtain multiple regional image samples; perform data enhancement on the regional image samples through a generative adversarial network to obtain multiple image training samples; and input the image training samples into the perception machine model to train the perception machine model.

[0192] In one embodiment, the generative adversarial network includes a generative network and a discriminative network; the model training module 904 is also configured to: input the region image samples into the generative network to generate a plurality of random region images; input the random region images into the discriminative network to generate a determination result as to whether the target object is contained in the random region image; the generative network adjusts the generation strategy of the random region image according to the determination result to output a plurality of random image training samples according to the input region image samples.

[0193] In one embodiment, the model training module 904 is further configured to: input the image training sample into the perceptual machine model to extract multiple regional image sample features in the image training sample; determine whether the image training sample contains the target object based on the multiple regional image sample features; when it is determined that the image training sample contains the target object, determine the feature position of the regional image sample feature corresponding to the target object; and determine the image position of the target object in the image training sample based on the feature position.

[0194] In one embodiment, the model training module 904 is further configured to: obtain the annotation box and corresponding annotation box parameters of the image training sample; determine whether the target object is contained in the annotation box; adjust the annotation box parameters according to the judgment result to determine the detection box for the target object; determine the probability that the regional image sample features belong to the features of the target object according to the detection box; and determine whether the target object is contained in the image training sample according to the probability.

[0195] In one embodiment, the model training module 904 is further configured to: perform bounding box regression on the detection box indicating that the features belonging to the target object are detected to correct the position of the detection box; filter the corrected detection box according to preset conditions; and determine the probability that the regional image sample features belong to the features of the target object based on the filtered detection box.

[0196] In one embodiment, the model training module 904 is further configured to: retain the revised detection frame that meets the preset conditions, wherein the preset conditions include any one of the following: the probability value of the revised detection frame is greater than the preset probability threshold; the size of the revised detection frame is greater than the preset size; the area overlap ratio of any two detection frames in the revised detection frame is lower than the preset overlap threshold.

[0197] In one embodiment, the actual position determination module 903 is also configured to calibrate the camera of the image acquisition device to obtain multiple coordinate data; through a preset algorithm and multiple coordinate data, the camera of the image acquisition device is calibrated to determine the camera calibration information of the image acquisition device.

[0198] In one embodiment, the actual position determination module 903 is also configured to place the calibration plate at a first angle of the image acquisition device, record the first camera coordinates of each feature point of the calibration plate in the image acquisition device and the corresponding first world coordinates; place the calibration plate at a second angle of the image acquisition device, record the second camera coordinates of each feature point of the calibration plate in the image acquisition device and the corresponding second world coordinates; place the calibration plate at a third angle of the image acquisition device, record the third camera coordinates of each feature point of the calibration plate in the image acquisition device and the corresponding third world coordinates.

[0199] The device for determining the position of the target object includes a processor and a memory. The above-mentioned image acquisition module, image position confirmation module and actual position determination module are all stored in the memory as program units, and the processor executes the above-mentioned program modules stored in the memory to implement corresponding functions.

[0200] In one embodiment, Fig.10 As shown, a control device for agricultural machinery is provided, comprising:

[0201] The target object determination module 1001 is configured to obtain a first actual position of a first target object in a current area of ​​a farmland during an agricultural machinery operation, wherein the target object determination module 1001 includes: Fig. 9 The image acquisition module 901, the image position confirmation module 902 and the actual position determination module 903 shown in FIG.

[0202] The working time determination module 1002 is configured to determine a first working time interval of the nozzle according to the first actual position.

[0203] The target object determination module 1001 is further configured to obtain a second actual position of a second target object in a next area of ​​the farmland.

[0204] The working time determination module 1002 is further configured to determine a second working time interval of the nozzle according to the second actual position; and determine an actual working time interval of the nozzle according to the first working time interval and the second working time interval.

[0205] In one embodiment, the working time determination module 1002 is also configured to: determine the actual working time interval as the start time of the first working time interval to the end time of the second working time interval when the first working time interval and the second working time interval meet any one of the following conditions: the first working time interval and the second working time interval are continuous; the first working time interval and the second working time interval overlap; the time interval between the first working time interval and the second working time interval is less than a preset threshold.

[0206] In one embodiment, the first working time interval is determined according to the first target object in the current area and the target object in the previous area.

[0207] In one embodiment, the control device further includes a nozzle confirmation module 1003, which is configured to determine a corresponding target nozzle among multiple nozzles according to the first actual position; wherein the number of nozzles is multiple. The working time determination module 1002 is further configured to determine an actual distance between the first target object and the target nozzle according to the first actual position; obtain the moving speed of the agricultural machinery and the control delay of the agricultural machinery; and determine a first working time interval of the target nozzle according to the actual distance, the moving speed and the control delay.

[0208] In one embodiment, the nozzle confirmation module 1003 is further configured to divide the current area into a plurality of sub-areas corresponding to the plurality of nozzles; and determine the nozzle corresponding to the sub-area where the first actual position is located as the target nozzle.

[0209] In one embodiment, the nozzle confirmation module 1003 is also configured to determine multiple sub-areas where the first actual position is located; determine the target sub-area based on the ratio of the area occupied by the first actual position in each sub-area to the area of ​​the first actual position; and determine the nozzle corresponding to the target sub-area as the target nozzle.

[0210] In one embodiment, the nozzle confirmation module 1003 is also configured to obtain a first ratio and a second ratio of the area occupied by the first actual position in the adjacent first sub-region and the second sub-region respectively to the area of ​​the first actual position; when the first ratio is higher than the first threshold and the second ratio is lower than the second threshold, the first sub-region is determined as the target sub-region.

[0211] In one embodiment, the nozzle confirmation module 1003 is further configured to determine both the first sub-region and the second sub-region as target sub-regions when the ratio of the first ratio to the second ratio is within a preset ratio interval.

[0212] In one embodiment, the control device further includes a nozzle control module 1004, which is configured to control the nozzle to be turned on and off according to the determined actual working time interval after the actual working time interval of the nozzle is determined by the working time determination module 1002.

[0213] In one embodiment, the spraying range of the nozzle includes a first spraying range and a second spraying range, and the second spraying range is larger than the first spraying range; the nozzle control module 1004 is also configured to switch the spraying range of the target nozzle corresponding to the target sub-area from the first spraying range to the second spraying range.

[0214] In one embodiment, the target object determination module 1001 is further configured to obtain a regional image of the current area of ​​the farmland; and determine the first actual position using the above-mentioned method for determining the position of the target object. In one embodiment, the sprinkler control module 1004 is further configured to turn off the sprinkler if the second target object is not detected in the next area of ​​the farmland and the end time corresponding to the first working time interval has been reached.

[0215] In one embodiment, the working time determination module 1002 is further configured to use the earlier one of the start time of the first working time interval and the start time of the second working time interval as the start time of the nozzle.

[0216] The control device for agricultural machinery includes a processor and a memory. The above-mentioned target object determination module and working time determination module are stored in the memory as program units, and the processor executes the above-mentioned program modules stored in the memory to implement corresponding functions.

[0217] In one embodiment, the processor includes a kernel, and the kernel retrieves the corresponding program unit from the memory. One or more kernels may be provided, and the kernel parameters may be adjusted to determine the location of the target object.

[0218] In one embodiment, the processor includes:

[0219] The image acquisition module 901 is configured to acquire regional images of farmland.

[0220] The image position confirmation module 902 is configured to input the region image into the perception machine model to determine the image position of the target object in the region image, and the perception machine model is trained by image training samples generated by a generative adversarial network.

[0221] The actual position determination module 903 is configured to obtain camera calibration information of an image acquisition device used to capture regional images; and determine the actual position of the target object according to the image position and the camera calibration information.

[0222] In one embodiment, the processor also includes a model training module 904, which is configured to: obtain multiple regional image samples; perform data enhancement on the regional image samples through a generative adversarial network to obtain multiple image training samples; and input the image training samples into the perceptual machine model to train the perceptual machine model.

[0223] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0224] An embodiment of the present invention provides a storage medium on which a program is stored. When the program is executed by a processor, the method for determining the position of a target object is implemented.

[0225] An embodiment of the present invention provides a processor, which is used to run a program, wherein the program executes the above method for determining the position of a target object when it is run.

[0226] In one embodiment, the processor includes a kernel, and the kernel retrieves the corresponding program unit from the memory. One or more kernels may be provided, and the kernel parameters may be adjusted to determine the location of the target object.

[0227] In one embodiment, the processor includes:

[0228] The target object determination module 1001 is configured to obtain a first actual position of a first target object in a current area of ​​a farmland during an agricultural machinery operation.

[0229] The working time determination module 1002 is configured to determine a first working time interval of the nozzle according to the first actual position.

[0230] The target object determination module 1001 is further configured to obtain a second actual position of a second target object in a next area of ​​the farmland.

[0231] The working time determination module 1002 is further configured to determine a second working time interval of the nozzle according to the second actual position; and determine an actual working time interval of the nozzle according to the first working time interval and the second working time interval.

[0232] The nozzle confirmation module 1003 is configured to determine a corresponding target nozzle among a plurality of nozzles according to the first actual position; wherein the number of the nozzles is multiple.

[0233] The nozzle control module 1004 is configured to control the nozzle to be turned on and off according to the determined actual working time interval after the actual working time interval of the nozzle is determined by the working time determination module 1002.

[0234] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0235] An embodiment of the present invention provides a storage medium on which a program is stored. When the program is executed by a processor, the above-mentioned control method for agricultural machinery is implemented.

[0236] An embodiment of the present invention provides a processor, which is used to run a program, wherein the above-mentioned control method for agricultural machinery is executed when the program is run.

[0237] In one embodiment, Fig.11 As shown, a control device 1100 for agricultural machinery is provided. The agricultural machinery includes a sprinkler. The control device 1100 includes: an image acquisition device 1101 configured to acquire regional images of farmland; and a processor 1102 configured to execute the above-mentioned control method for agricultural machinery.

[0238] In one embodiment, Fig.12 As shown, an agricultural machine 1200 is provided, comprising: a spray head 1201; and the control device 1100 for the agricultural machine as described above.

[0239] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Fig.13 As shown. The computer device includes a processor A01, a network interface A02, a memory (not shown in the figure) and a database (not shown in the figure) connected through a system bus. Among them, the processor A01 of the computer device is used to provide computing and control capabilities. The memory of the computer device includes an internal memory A03 and a non-volatile storage medium A04. The non-volatile storage medium A04 stores an operating system B01, a computer program B02 and a database (not shown in the figure). The internal memory A03 provides an environment for the operation of the operating system B01 and the computer program B02 in the non-volatile storage medium A04. The database of the computer device is used to store data related to the execution of a method for determining the position of a target object and the execution of a control method for agricultural machinery. The network interface A02 of the computer device is used to communicate with an external terminal through a network connection. When the computer program B02 is executed by the processor A01, a method for determining the position of a target object is implemented.

[0240] Those skilled in the art will understand that Fig.13 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0241] An embodiment of the present invention provides a device, which includes a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps of the method for determining the position of a target object as described above are implemented.

[0242] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing the steps of the method for determining the position of a target object as described above.

[0243] An embodiment of the present invention provides a device, which includes a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps of the control method for agricultural machinery as described above are implemented.

[0244] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing the steps of the control method for agricultural machinery as described above.

[0245] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0246] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0247] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0248] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0249] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0250] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0251] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0252] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0253] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A control method for agricultural machinery, the agricultural machinery comprising a sprinkler, characterized in that: The control method comprises: obtaining a first actual position of a first target object in a current area of ​​the farmland during the operation of the agricultural machinery; Determining a first working time interval of the nozzle according to the first actual position; Acquire a second actual position of a second target object in a next area of ​​the farmland; determining a second working time interval of the nozzle according to the second actual position; Determine the actual working time interval of the nozzle according to the first working time interval and the second working time interval; Wherein, determining the actual working time interval of the nozzle according to the first working time interval and the second working time interval includes: when the first working time interval and the second working time interval meet any one of the following conditions, determining the actual working time interval as the start time of the first working time interval to the end time of the second working time interval: The first working time interval is continuous with the second working time interval; The first working time interval overlaps with the second working time interval; The time interval between the first working time interval and the second working time interval is smaller than a preset threshold.

2. The control method according to claim 1, characterized in that: The first working time interval is determined according to the first target object in the current area and the target object in the previous area.

3. The control method according to claim 1, characterized in that: The number of the nozzles is multiple; and determining the first working time interval of the nozzle according to the first actual position includes: Determine a corresponding target nozzle among the plurality of nozzles according to the first actual position; Determine an actual distance between the first target object and the target nozzle according to the first actual position; Acquiring the moving speed of the agricultural machinery and the control delay of the agricultural machinery; A first working time interval of the target nozzle is determined according to the actual distance, the moving speed and the control delay.

4. The control method according to claim 3, characterized in that: The step of determining a corresponding target nozzle among a plurality of nozzles according to the first actual position comprises: Dividing the current area into a plurality of sub-areas corresponding to the plurality of nozzles; The nozzle corresponding to the sub-region where the first actual position is located is determined as the target nozzle.

5. The control method according to claim 4, characterized in that: The step of determining the nozzle corresponding to the sub-area where the first actual position is located as the target nozzle comprises: Determine a plurality of sub-areas where the first actual location is located; determining a target sub-region according to a ratio of an area occupied by the first actual position in each sub-region to an area of ​​the first actual position; The nozzle corresponding to the target sub-area is determined as the target nozzle.

6. The control method according to claim 5, characterized in that: The determining the target sub-region according to the ratio of the area occupied by the first actual position in each sub-region to the area of ​​the first actual position comprises: Acquire a first ratio and a second ratio of areas respectively occupied by the first actual position in adjacent first sub-regions and second sub-regions to the area of ​​the first actual position; When the first ratio is higher than a first threshold and the second ratio is lower than a second threshold, the first sub-region is determined as a target sub-region.

7. The control method according to claim 6, characterized in that: The spraying range of the nozzle includes a first spraying range and a second spraying range, and the second spraying range is larger than the first spraying range; the method further includes: The spraying range of the target nozzle corresponding to the target sub-area is switched from the first spraying range to the second spraying range.

8. The control method according to claim 6, characterized in that: The determining the target sub-region according to the ratio of the area occupied by the first actual position in each sub-region to the area of ​​the first actual position further includes: When the ratio of the first ratio to the second ratio is within a preset ratio interval, both the first sub-region and the second sub-region are determined as target sub-regions.

9. The control method according to claim 1, characterized in that: The obtaining of the first actual position of the first target object in the current area of ​​the farmland during the operation of the agricultural machinery comprises: Acquiring a regional image corresponding to a current region of the farmland during the operation of the agricultural machinery; Inputting the region image into a perceptual machine model to determine an image position of a target object in the region image; Obtaining camera calibration information of an image acquisition device used to capture images of the area; A first actual position of the first target object is determined according to the image position and the camera calibration information.

10. The control method according to claim 9, characterized in that: The obtaining of camera calibration information of an image acquisition device used to capture the regional image comprises: Calibrate the camera of the image acquisition device to obtain a plurality of coordinate data; The camera of the image acquisition device is calibrated by using a preset algorithm and the plurality of coordinate data to determine the camera calibration information of the image acquisition device.

11. The control method according to claim 10, characterized in that: The camera of the image acquisition device is calibrated to obtain a plurality of coordinate data, including: Placing the calibration plate at a first angle of the image acquisition device, and recording the first camera coordinates of each feature point of the calibration plate at the image acquisition device and the corresponding first world coordinates; Placing the calibration plate at a second angle of the image acquisition device, and recording the second camera coordinates of each feature point of the calibration plate at the image acquisition device and the corresponding second world coordinates; The calibration plate is placed at a third angle to the image acquisition device, and the third camera coordinates of each feature point of the calibration plate at the image acquisition device and the corresponding third world coordinates are recorded.

12. The control method according to claim 1, characterized in that: The control method also includes one or both of the following: If the second target object is not detected in the next area of ​​the farmland and the end time corresponding to the first working time interval has been reached, turning off the sprinkler; The earlier one of the start time of the first working time interval and the start time of the second working time interval is used as the start time of the nozzle.

13. A processor, characterized in that: The method is configured to execute the control method for agricultural machinery according to any one of claims 1 to 12.

14. A control device for agricultural machinery, the agricultural machinery comprising a sprinkler, characterized in that: The control device comprises: An image acquisition device configured to acquire regional images of farmland; and A processor according to claim 13.

15. An agricultural machine, characterized in that: include: Sprinklers; as well as A control device for an agricultural machine according to claim 14.

16. A machine-readable storage medium having instructions stored thereon, characterized in that: When the instruction is executed by a processor, the processor is configured to execute the control method for an agricultural machine according to any one of claims 1 to 12.

Images