Seed detection method, device, electronic equipment and storage medium

Through the pre-trained type-type port carrying status prediction model, the missed-cast detection is performed using the sorting video data, which solves the problem of low missed-casting accuracy in the air-suction seed ejaculation device, and achieves higher detection and statistical accuracy.

CN118537781BActive Publication Date: 2024-11-26NANJING AGRICULTURAL UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411009441.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-26
Publication Date
2024-11-26
Estimated Expiration
2044-07-26

AI Technical Summary

Technical Problem

In soybean seed seed discharger, the air suction seed discharger frequently occurs due to the different sizes and shapes of soybeans and the complex operating environment, and the existing photoelectric sensor detection methods have low detection and statistical accuracy.

Method used

The pre-trained type-type hole carrying status prediction model is adopted to predict the type-type status by obtaining the sorting video data, tracking and counting missing type holes, and calculating the missed play rate to improve detection and statistical accuracy.

Benefits of technology

By accurately predicting the carrying status and tracking count, the detection and statistical accuracy of seeding missed seeds are significantly improved, and the occurrence of missed seeds is reduced.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118537781B_ABST
    Figure CN118537781B_ABST
Patent Text Reader

Abstract

The present invention discloses a seeding detection method, device, electronic device and storage medium. The method comprises: obtaining seeding video data; inputting the seeding video data into a pre-trained seeding hole seeding condition prediction model to obtain a seeding condition prediction result of the seeding hole; when the seeding condition prediction result of the seeding hole is a seed-missing state, tracking and counting the seeding hole in the seed-missing state to obtain a target seed-missing hole count value, and determining the seeding missed sowing rate based on the target seed-missing hole count value. The above technical scheme realizes the accurate prediction of the seed-missing state through the seeding hole seeding condition prediction model, thereby improving the detection and statistical accuracy of seeding missed sowing.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of target detection technology, and in particular to a seeding detection method, device, electronic equipment and storage medium. Background Art

[0002] The pneumatic seed meter is a precision seeding equipment currently commonly used in soybean plot sowing. It has the advantages of high efficiency, high reliability and high precision. However, when using the pneumatic seed meter for sowing operations, due to factors such as the different sizes and shapes of soybeans and the complex working environment, missed seeding still occurs frequently, which will eventually seriously affect the soybean yield.

[0003] In the prior art, the traditional seeding detection method mainly relies on photoelectric sensor detection, which has the problems of seeding missed detection and low statistical accuracy. Summary of the invention

[0004] The present invention provides a seeding detection method, device, electronic equipment and storage medium to improve the detection and statistical accuracy of seeding missed planting.

[0005] According to one aspect of the present invention, a seeding detection method is provided, comprising:

[0006] Obtaining seeding video data;

[0007] Inputting the seeding video data into a pre-trained seeding hole seed carrying condition prediction model to obtain a seeding hole seed carrying condition prediction result;

[0008] When the seed carrying status prediction result of the seed-discharging type hole is a seed-deficient state, the seed-discharging type holes in the seed-deficient state are tracked and counted to obtain target seed-deficient type hole count values, and the seed-discharging missed planting rate is determined based on the target seed-deficient type hole count values.

[0009] According to another aspect of the present invention, there is provided a seeding detection device, comprising:

[0010] A seeding video data acquisition module is used to acquire seeding video data;

[0011] A seed-carrying condition prediction module for the seed-planting holes is used to input the seed-planting video data into a pre-trained seed-carrying condition prediction model for the seed-planting holes to obtain a prediction result for the seed-carrying condition of the seed-planting holes;

[0012] The module for determining the seeding missed planting rate is used to track and count the seeding holes in the seed-deficient state when the seed-carrying status prediction result of the seeding hole is a seed-deficient state, obtain the target seed-deficient hole count value, and determine the seeding missed planting rate based on the target seed-deficient hole count value.

[0013] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0014] at least one processor;

[0015] and a memory communicatively coupled to the at least one processor;

[0016] The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the seeding detection method described in any embodiment of the present invention.

[0017] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the seeding detection method described in any embodiment of the present invention when executed.

[0018] The technical solution of the embodiment of the present invention obtains the prediction result of the seed carrying condition of the seed-forming type hole by inputting the seed-forming video data into the seed carrying condition prediction model of the seed-forming type hole that has been trained in advance; when the prediction result of the seed carrying condition of the seed-forming type hole is the seed-missing state, the seed-forming type hole in the seed-missing state is tracked and counted to obtain the target seed-missing type hole count value, and then the seeding missed sowing rate is determined based on the target seed-missing type hole count value. The above technical solution realizes the accurate prediction of the seed carrying condition through the seed-forming type hole prediction model, and then when the seed-forming type hole in the seed-missing state is tracked and counted and the seeding missed sowing rate is calculated, the detection and statistical accuracy of seeding missed sowing can be effectively improved.

[0019] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0021] Figure 1 is a flow chart of a seeding detection method provided according to Embodiment 1 of the present invention;

[0022] Figure 2 is a flow chart of a seeding detection method provided according to Embodiment 2 of the present invention;

[0023] Figure 3 A schematic diagram of the structure of a backbone network module provided by an embodiment of the present invention;

[0024] Figure 4 A schematic diagram of the structure of a C2f-DAR module provided in an embodiment of the present invention;

[0025] Figure 5 A schematic diagram of the structure of a Bottleneck-R module provided in an embodiment of the present invention;

[0026] Figure 6 A schematic diagram of the structure of a seed carrying condition prediction model for a seed row hole provided in an embodiment of the present invention;

[0027] Figure 7 is a flow chart of a seeding detection method provided according to Embodiment 3 of the present invention;

[0028] Figure 8 is a schematic diagram of the position of a rectangular area provided according to an embodiment of the present invention;

[0029] Fig. 9 is a structural schematic diagram of a seeding detection device provided according to a fourth embodiment of the present invention;

[0030] Fig.10 It is a structural schematic diagram of an electronic device for implementing the seeding detection method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0031] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0032] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices. The acquisition, storage, use, processing, etc. of data in the technical solution of this application comply with the relevant provisions of national laws and regulations.

[0033] Embodiment 1

[0034] Figure 1 This is a flow chart of a seeding detection method provided in the first embodiment of the present invention. This embodiment is applicable to the case of detecting the seed shortage status in real time. The method can be executed by a seeding detection device. The seeding detection device can be implemented in the form of hardware and / or software. The seeding detection device can be configured in a monitoring device arranged on a seeding device. Figure 1 As shown, the method includes:

[0035] S110, obtaining seeding video data.

[0036] In this embodiment, the seeding video data refers to the seeding video data stream acquired in real time, which may include multiple seeding video frames. Any seeding video frame may include one or more seeding hole-related features of the seeding device. The seeding hole may be a seeding hole that is normally filled with seeds, or a seeding hole that is not filled with seeds, which is not specifically limited here.

[0037] Specifically, the real-time seeding video data can be collected through a camera, and the seeding video data can also be retrieved from a preset storage location of a monitoring device, which is not specifically limited here.

[0038] S120, inputting the seeding video data into a pre-trained seed-carrying condition prediction model of the seeding hole pattern to obtain a seed-carrying condition prediction result of the seeding hole pattern.

[0039] In this embodiment, the seed-carrying state prediction model of the seed-rowing hole refers to a pre-trained neural network prediction model, which can be used to predict the seed-carrying state prediction results of the seed-rowing hole. For each seed-rowing video frame, the seed-carrying state prediction result of the seed-rowing hole can be a seed-missing state or a seed-not-missing state. In other words, the seed-carrying state prediction result of the seed-rowing hole can include the seed-carrying state prediction results corresponding to multiple seed-rowing video frames.

[0040] Specifically, multiple groups of seed carrying status prediction model training samples can be obtained, wherein the seed carrying status prediction model training samples include seeding video sample frames and label information corresponding to the seeding video sample frames, wherein the label information can be a normally filled seeding hole or an unfilled seeding hole; and then the seeding video sample frames are input into the initial neural network model in sequence, and the frame initial neural network model outputs the seed carrying status prediction result, and then the model loss is determined according to the seed carrying status prediction result and the label information corresponding to the seeding video sample frame, and the network parameters of the initial neural network model are adjusted according to the model loss until the model training is completed, and the seed carrying status prediction model of the seeding hole is obtained, wherein the loss function for determining the model loss can be the MPDIoU loss function.

[0041] S130. When the seed carrying status prediction result of the seed-discharging hole is a seed-deficient state, the seed-discharging hole in the seed-deficient state is tracked and counted to obtain a target seed-deficient hole count value, and the seed-discharging missed planting rate is determined based on the target seed-deficient hole count value.

[0042] In this embodiment, the target seed-missing type hole count value refers to the cumulative result of tracking and counting the seed-missing type holes in the seed-missing state in the seed-seeding video data within a preset time range.

[0043] Specifically, the seed-missing hole status in the seed-missing video data within a preset time range is tracked and counted to obtain a target seed-missing hole count value, and then the target seed-missing hole count value is divided by the theoretical sowing quantity within the preset time range to obtain the seed-missing rate.

[0044] In some optional embodiments, after determining the seeding missed rate based on the target seed-missing type hole count value, the method includes: generating a missed seeding alarm signal if the seeding missed rate is greater than a preset missed seeding rate threshold; and / or determining the current seeding speed, determining the seeding speed difference based on the current seeding speed and the preset seeding speed, and performing closed-loop control of the motor speed based on the seeding speed difference.

[0045] Exemplarily, the monitoring device may include but is not limited to components such as Raspberry Pi 4B and cameras, and the control device may include a single-chip microcomputer with an ARM processor as the core and components such as motors. Specifically, the monitoring device may send the seeding missed sowing rate and the current seeding speed to the control device through serial communication, and then the control device transmits the seeding missed sowing rate to the host computer through network communication. When the seeding missed sowing rate is greater than the preset missed sowing rate threshold, the host computer generates a missed sowing alarm signal, and according to the missed sowing alarm signal, drives the general-purpose input / output (GPIO) port electrically connected to the buzzer to make the buzzer sound an alarm. For the current seeding speed, the control device may compare the current seeding speed with the preset seeding speed to obtain the seeding speed difference, and then according to the seeding speed difference, the motor speed may be closed-loop controlled by the PID algorithm to make the difference between the current seeding speed and the preset seeding speed approach zero, so that the current seeding operation speed tends to be optimized and the occurrence of missed sowing is reduced. Specifically, the closed-loop control of the motor speed can be calculated by the following formula:

[0046] ;

[0047] in, It is the PWM signal that the control device outputs to the motor through the PID algorithm; is the proportional gain coefficient; is the integration time constant; is the differential time constant; It is the difference between the current seeding speed and the preset seeding speed.

[0048] The technical solution of the embodiment of the present invention obtains the prediction result of the seed carrying condition of the seed-forming type hole by inputting the seed-forming video data into the seed carrying condition prediction model of the seed-forming type hole that has been trained in advance; when the prediction result of the seed carrying condition of the seed-forming type hole is the seed-missing state, the seed-forming type hole in the seed-missing state is tracked and counted to obtain the target seed-missing type hole count value, and then the seeding missed sowing rate is determined based on the target seed-missing type hole count value. The above technical solution realizes the accurate prediction of the seed carrying condition through the seed-forming type hole prediction model, and then when the seed-forming type hole in the seed-missing state is tracked and counted and the seeding missed sowing rate is calculated, the detection and statistical accuracy of seeding missed sowing can be effectively improved.

[0049] Embodiment 2

[0050] Figure 2A flow chart of a seeding detection method provided in the second embodiment of the present invention, the method of this embodiment can be combined with the various optional schemes in the seeding detection method provided in the above embodiments. The seeding detection method provided in this embodiment is further optimized. Optionally, the seed-carrying condition prediction model of the seeding hole includes a backbone network module, a progressive feature pyramid network module and a detection head module; correspondingly, the seeding video data is input into the pre-trained seeding condition prediction model of the seeding hole to obtain the seed-carrying condition prediction result of the seeding hole, including: inputting the seeding video data into the backbone network module to obtain a seeding feature image; inputting the seeding feature image into the progressive feature pyramid network module to obtain a seeding fusion feature image; inputting the seeding fusion feature image into the detection head module to obtain the seeding condition prediction result of the seeding hole.

[0051] like Figure 2 As shown, the method includes:

[0052] S210, obtaining seeding video data.

[0053] S220, inputting the seeding video data into the backbone network module to obtain a seeding feature image.

[0054] S230, inputting the seeding feature image into the progressive feature pyramid network module to obtain a seeding fusion feature image.

[0055] S240, inputting the seeding fusion feature image into the detection head module to obtain the seed carrying status prediction result of the seeding hole.

[0056] S250. When the seed carrying status prediction result of the seed-discharging hole is a seed-deficient state, the seed-discharging hole in the seed-deficient state is tracked and counted to obtain a target seed-deficient hole count value, and the seed-discharging missed planting rate is determined based on the target seed-deficient hole count value.

[0057] In this embodiment, the backbone network module can be a module improved based on Backbone in YOLOv8. Specifically, a partial convolution module can be used to replace the Conv module in the Backbone structure in YOLOv8, thereby reducing the number of parameters and the amount of calculation of the detection network. The receptive field attention convolution module can also be used to replace the Conv module in the C2f module in the Backbone structure in YOLOv8, thereby increasing the network's detection capability for irregularly shaped targets. The ACmix module can also be inserted between the Conv module and the Bottleneck module in the C2f module to filter invalid feature information, thereby improving the detection capability for small target features. The receptive field attention convolution module can also be used to replace the Conv module in the Bottleneck module, which can achieve lightweight network while enhancing feature extraction capabilities and improving computational efficiency. It should be noted that the above improvements are only examples and are not specific limitations on the backbone network module.

[0058] The progressive feature pyramid network module can use an adaptive feature fusion method to assign different spatial weights to features at different levels in the seed feature image. It first fuses the low-level features, then the deep-level features, and finally the top-level features. Through progressive fusion, the semantic gap between features at adjacent levels is narrowed, thereby improving the feature expression ability of the network.

[0059] The detection head module is used to perform target detection operations on the seeding fusion feature image, so as to predict the position and category of the seeding hole in the seeding video data.

[0060] Optionally, the backbone network module includes a first partial convolution module, a second partial convolution module, a first C2f-DAR module, a third partial convolution module, a first spatial depth conversion convolution module, a second C2f-DAR module, a fourth partial convolution module, a second spatial depth conversion convolution module, a third C2f-DAR module, a fifth partial convolution module, a third spatial depth conversion convolution module, a fourth C2f-DAR module and a fast spatial pyramid pooling module; accordingly, the seeding video data is input into the backbone network module to obtain a seeding feature image, including The method comprises the following steps: inputting the seeding video data into a first partial convolution module to obtain a first partial convolution feature image; inputting the first partial convolution feature image into a second partial convolution module to obtain a second partial convolution feature image; inputting the second partial convolution feature image into the first C2f-DAR module to obtain a first C2f-DAR processed feature image; inputting the first C2f-DAR processed feature image into a third partial convolution module to obtain a third partial convolution feature image; inputting the third partial convolution feature image into a first spatial depth conversion convolution module to obtain a first spatial depth conversion convolution feature image; inputting the first spatial depth conversion convolution feature image into a second C2f-DAR module to obtain a second C2f-DAR processed feature image; inputting the second C2f-DAR processed feature image into a fourth partial convolution module to obtain a fourth partial convolution feature image; inputting the fourth partial convolution feature image into a second spatial depth conversion convolution module to obtain a second spatial depth conversion convolution feature image; inputting the second spatial depth conversion convolution feature image into a third C2f-DAR module to obtain a third C 2f-DAR processing feature image; inputting the third C2f-DAR processing feature image into the fifth partial convolution module to obtain a fifth partial convolution feature image; inputting the fifth partial convolution feature image into the third spatial depth conversion convolution module to obtain a third spatial depth conversion convolution feature image; inputting the third spatial depth conversion convolution feature image into the fourth C2f-DAR module to obtain a fourth C2f-DAR processing feature image; inputting the fourth C2f-DAR processing feature image into the fast spatial pyramid pooling module to obtain a seeding feature image.

[0061] Among them, the C2f-DAR module refers to the improved C2f module.

[0062] For example, Figure 3A schematic diagram of the structure of a backbone network module provided by an embodiment of the present invention, wherein PConv represents a partial convolution module, SPD-conv represents a spatial depth conversion convolution module, and SPPF represents a fast spatial pyramid pooling module. Specifically, real-time seeding video data can be input into the improved backbone network module, and feature extraction is performed in sequence through PConv, PConv, C2f-DAR, PConv, SPD-conv, C2f-DAR, PConv, SPD-conv, C2f-DAR, PConv, SPD-conv, C2f-DAR, and SPFF to obtain a seeding feature image.

[0063] It should be noted that the improved C2f-DAR module replaces the original C2f module in Backbone, which can improve the network feature extraction capability and detection rate. In addition, the SPD-conv module is inserted before the second C2f-DAR module, the third C2f-DAR module and the fourth C2f-DAR module. The SPD-conv module can retain the feature information of small targets as much as possible by dividing the feature map into channels and stacking adjacent pixels, laying the foundation for the subsequent detection of row-type holes and effectively improving the model's detection performance for small target features.

[0064] Optionally, the C2f-DAR module includes a first receptive field attention convolution module, an ACmix module, a Split module, a first Bottleneck-R module, a second Bottleneck-R module, a third Bottleneck-R module and a second receptive field attention convolution module, wherein the output of the first receptive field attention convolution module is used as the input of the ACmix module, the output of the ACmix module is used as the input of the Split module, the output of the Split module is used as the input of the first Bottleneck-R module, the output of the first Bottleneck-R module is used as the input of the second Bottleneck-R module, and the second Bottleneck-R The output of the module is used as the input of the third Bottleneck-R module; the output of the first Bottleneck-R module, the output of the second Bottleneck-R module, the output of the third Bottleneck-R module and the output of the Split module are added as the input of the second receptive field attention convolution module; the Bottleneck-R module includes a third receptive field attention convolution module and a fourth receptive field attention convolution module, wherein the output of the third receptive field attention convolution module is used as the input of the fourth receptive field attention convolution module, and the output of the fourth receptive field attention convolution module and the input of the third receptive field attention convolution module are added as the output of the Bottleneck-R module.

[0065] In this embodiment, the receptive field attention convolution module combines the spatial attention mechanism with the convolution operation, and can learn the attention feature map through the interactive receptive field spatial features, thereby improving the network feature extraction performance. The calculation formula of the receptive field attention convolution module can be:

[0066] ;

[0067] in, represents the attention feature map; Represents the spatial characteristics of the receptive field.

[0068] For example, Figure 4 A schematic diagram of the structure of a C2f-DAR module provided in an embodiment of the present invention; Figure 5 A schematic diagram of the structure of a Bottleneck-R module provided in an embodiment of the present invention. RFAConv represents a receptive field attention convolution module, Concat represents addition, and Add represents addition. In this embodiment, the ACmix module combines the self-attention mechanism and the standard convolution operation, which has a lower computational cost than the traditional self-attention mechanism.

[0069] Optionally, the progressive feature pyramid network module includes a fifth C2f-DAR module, a sixth C2f-DAR module, and a seventh C2f-DAR module; the seeding fusion feature image includes a first seeding fusion feature image, a second seeding fusion feature image, and a third seeding fusion feature image; accordingly, the seeding feature image is input into the progressive feature pyramid network module to obtain a seeding fusion feature image, including: adding the upsampled images corresponding to the first C2f-DAR processing feature image and the second C2f-DAR processing feature image and inputting them into the fifth C2f-DAR module to obtain the fifth C2f-DAR processing feature image; The down-sampled image corresponding to the first C2f-DAR processing feature image, the second C2f-DAR processing feature image, and the up-sampled image corresponding to the third C2f-DAR processing feature image are added and input to the sixth C2f-DAR module to obtain the sixth C2f-DAR processing feature image; the down-sampled image corresponding to the second C2f-DAR processing feature image and the third C2f-DAR processing feature image are added and input to the seventh C2f-DAR module to obtain the seventh C2f-DAR processing feature image; the fifth C2f-DAR processing feature image and the up-sampled image corresponding to the sixth C2f-DAR processing feature image are added and input to the seventh C2f-DAR module to obtain the seventh C2f-DAR processing feature image. The up-sampled image and the up-sampled image corresponding to the seeding feature image are added to obtain a first seeding fusion feature image; the down-sampled image corresponding to the fifth C2f-DAR processing feature image, the sixth C2f-DAR processing feature image, and the up-sampled image corresponding to the seventh C2f-DAR processing feature image are added to obtain a second seeding fusion feature image; the down-sampled image corresponding to the sixth C2f-DAR processing feature image, the seventh C2f-DAR processing feature image, and the up-sampled image corresponding to the seeding feature image are added to obtain a third seeding fusion feature image; accordingly, the detection head module includes a first detection head module, a second A detection head module and a third detection head module; accordingly, the seeding fusion feature image is input into the detection head module to obtain a prediction result of the seed carrying condition of the seeding type hole, including: inputting the first seeding fusion feature image into the first detection head module to obtain a first seed carrying condition prediction result; inputting the second seeding fusion feature image into the second detection head module to obtain a second seed carrying condition prediction result; inputting the third seeding fusion feature image into the third detection head module to obtain a third seed carrying condition prediction result; determining the seed carrying condition prediction result of the seeding type hole based on the first seed carrying condition prediction result, the second seed carrying condition prediction result and the third seed carrying condition prediction result.

[0070] Specifically, any one of the first seed-carrying status prediction result, the second seed-carrying status prediction result and the third seed-carrying status prediction result can be selected as the seed-carrying status prediction result of the seed-rowing type hole, or the same result among the first seed-carrying status prediction result, the second seed-carrying status prediction result and the third seed-carrying status prediction result can be determined as the seed-carrying status prediction result of the seed-rowing type hole, without specific limitation here.

[0071] For example, Figure 6 A structural diagram of a seed-carrying status prediction model for a seed row hole provided in an embodiment of the present invention, wherein Backbone represents a backbone network module, Neck represents a progressive feature pyramid network module, Head represents a detection head module, Downsample represents a downsampling operation on a feature map, Upsample represents an upsampling operation on a feature map, Detect (160×160) represents a first detection head module, Detect (80×80) represents a second detection head module, and Detect (40×40) represents a third detection head module. It should be noted that Detect (160×160) can be an added small target detection head, and the original deep-level detection head is deleted, which can make the network lightweight and improve the real-time detection capability for small targets.

[0072] The technical solution of the embodiment of the present invention effectively improves the feature extraction capability of the network by improving the backbone network module, and introduces a progressive feature pyramid network module to help reduce the semantic gap between features at different levels, enhance the robustness and multi-scale capability of detection, and thus achieve accurate prediction of seed carrying status.

[0073] Embodiment 3

[0074] Figure 7 A flow chart of a seeding detection method provided in Embodiment 3 of the present invention, the method of this embodiment can be combined with the various optional schemes in the seeding detection method provided in the above embodiments. The seeding detection method provided in this embodiment is further optimized. Optionally, the tracking and counting of seed-missing type holes to obtain a target seed-missing type hole count value includes: for a ray extending in any direction from the center of the detection frame of the seed-missing type hole, if the number of intersections of the ray and the preset rectangular area is an odd number, then the seed-missing type hole count is increased by one, until the traversal of each seeding video frame in the seeding video data is completed, and the target seed-missing type hole count value is obtained.

[0075] like Figure 7 As shown, the method includes:

[0076] S310, obtaining seeding video data.

[0077] S320, inputting the seeding video data into a pre-trained seed-carrying condition prediction model of the seed-carrying hole to obtain a seed-carrying condition prediction result of the seed-carrying hole.

[0078] S330. When the prediction result of the seed-carrying status of the seed-planting type hole is a seed-missing state, for rays drawn in any direction from the center of the detection frame of the seed-planting type hole in the seed-missing state, if the number of intersections of the rays with the preset rectangular area is an odd number, the seed-missing type hole count is increased by one until the traversal of each seed-planting video frame in the seed-planting video data is completed to obtain the target seed-missing type hole count value.

[0079] S340, determining the seed missing rate based on the target seed-missing type hole count value.

[0080] The preset rectangular area is a rectangular area containing one or more row-shaped holes. Figure 8 It is a schematic diagram of the position of a rectangular area provided according to an embodiment of the present invention. It can be understood that if the number of intersections between the ray and the preset rectangular area is an odd number, it indicates that the seed-missing row hole is located in the preset rectangular area.

[0081] Exemplarily, the seed-carrying status prediction model of the seed-seeding type hole can be used as a detector of the Deepsort algorithm to detect the missing seed-type hole in the current seed-seeding video frame, and the trajectory of the seed-seeding type hole in the missing seed state (i.e., the missing seed-type hole) in the next seed-seeding video frame can be predicted by the Kalman filter, and then the detection frame and prediction frame of the previous seed-seeding video frame and the next seed-seeding video frame can be matched using the Hungarian algorithm, and each missing seed-type hole is assigned a unique identification ID to track the missing seed-type hole. Furthermore, the process of counting the missing seed-type holes includes: setting a virtual rectangular area at a specific position of the seed-seeding device; dividing the scene into two areas, one of which is outside the rectangular area, and the other is inside the rectangular area. Furthermore, a ray is drawn from the center of the seed-missing hole detection frame in any direction. If the ray has no intersection with the rectangular area or has an even number of intersections, it is judged that the seed-missing hole has not entered the rectangular area. If the ray has only an odd number of intersections with the rectangular area, it is judged that the seed-missing hole has entered the rectangular area, and the seed-missing hole count is increased by one, and the ID of the seed-missing hole remains unchanged to prevent multiple judgments or misjudgments. This process is repeated until the traversal of each seed-missing video frame in the seed-missing video data is completed, and the target seed-missing hole count value is obtained.

[0082] The calculation formula for the current seeding speed can be:

[0083] ;

[0084] in, Indicates the current seeding speed; Indicates the ratio of real distance to pixel distance; Indicates the pixel distance that the seed-shaped hole moves through the rectangular area; Indicates the actual time it takes for the seed-seeding hole to pass through the rectangular area. Specifically, the ratio of the real distance to the pixel distance can be calculated from the real size of the seed-seeding hole set in advance and the pixel size of the seed-seeding hole detected; the pixel distance moved by the seed-seeding hole through the rectangular area can be calculated using the center distance of the detection frame at the two moments when each seed-seeding hole enters the rectangular area and passes through the rectangular area; the frame difference experienced by the seed-seeding hole through the rectangular area can be calculated from the start and end frames of the detection frame of the seed-seeding hole from entering the rectangular area to passing through the rectangular area, and the frame difference experienced by the seed-seeding hole through the rectangular area can be divided by the monitoring frame rate to obtain the actual time it takes for the seed-seeding hole to pass through the rectangular area.

[0085] In some optional embodiments, before training the seed-carrying status prediction model of the seed-carrying holes, the method also includes: using a high-definition camera to shoot seed-carrying operation videos of the seed-carrying device under different visual conditions; using OpenCV graphics processing tools to extract 1 frame of image every 5 frames from the collected seed-carrying operation video; preprocessing all the extracted frame images to obtain a processed image set, and the preprocessing methods include but are not limited to flipping, scaling, color gamut conversion and / or random cropping; using an image annotation tool to annotate the preprocessed image set with detection frames one by one, and the annotated objects are normally filled seed-carrying holes and unfilled seed-carrying holes, and then every four pictures are spliced ​​into a picture with label information to form an initial data set; the initial data set is divided into a training set, a test set and a verification set in a ratio of 7:2:1 for subsequent model training and verification.

[0086] The technical solution of the embodiment of the present invention is to draw rays in any direction from the center of the detection frame of the seed-missing type hole in the seed-missing state. If the number of intersections of the rays with the preset rectangular area is an odd number, the count of the seed-missing type hole will be increased by one until the traversal of each seed-missing video frame in the seed-missing video data is completed, and the target seed-missing type hole count value is obtained, thereby effectively improving the statistical accuracy of seed missing.

[0087] Embodiment 4

[0088] Fig. 9 This is a schematic diagram of the structure of a seeding detection device provided in the fourth embodiment of the present invention. Fig. 9 As shown, the device comprises:

[0089] A seeding video data acquisition module 410 is used to acquire seeding video data;

[0090] The seed-carrying condition prediction module 420 is used to input the seed-planting video data into the seed-carrying condition prediction model of the seed-planting hole that has been trained in advance to obtain the seed-carrying condition prediction result of the seed-planting hole;

[0091] The seeding missed planting rate determination module 430 is used to track and count the seeding type holes in the seed-deficient state when the seed carrying status prediction result of the seeding type holes is a seed-deficient state, obtain the target seed-deficient type hole count value, and determine the seeding missed planting rate based on the target seed-deficient type hole count value.

[0092] The technical solution of the embodiment of the present invention obtains the prediction result of the seed carrying condition of the seed-forming type hole by inputting the seed-forming video data into the seed carrying condition prediction model of the seed-forming type hole that has been trained in advance; when the prediction result of the seed carrying condition of the seed-forming type hole is the seed-missing state, the seed-forming type hole in the seed-missing state is tracked and counted to obtain the target seed-missing type hole count value, and then the seeding missed sowing rate is determined based on the target seed-missing type hole count value. The above technical solution realizes the accurate prediction of the seed carrying condition through the seed-forming type hole prediction model, and then when the seed-forming type hole in the seed-missing state is tracked and counted and the seeding missed sowing rate is calculated, the detection and statistical accuracy of seeding missed sowing can be effectively improved.

[0093] In some optional embodiments, the seed-carrying status prediction model of the seed-row hole includes a backbone network module, a progressive feature pyramid network module and a detection head module;

[0094] Correspondingly, the hole seed carrying condition prediction module 420 includes:

[0095] A backbone network processing unit, used for inputting the seeding video data into the backbone network module to obtain a seeding feature image;

[0096] A progressive feature pyramid network processing unit, used for inputting the seeding feature image into the progressive feature pyramid network module to obtain a seeding fusion feature image;

[0097] The detection head processing unit is used to input the seeding fusion feature image into the detection head module to obtain the seed carrying status prediction result of the seeding hole.

[0098] In some optional embodiments, the backbone network module includes a first partial convolution module, a second partial convolution module, a first C2f-DAR module, a third partial convolution module, a first spatial depth conversion convolution module, a second C2f-DAR module, a fourth partial convolution module, a second spatial depth conversion convolution module, a third C2f-DAR module, a fifth partial convolution module, a third spatial depth conversion convolution module, a fourth C2f-DAR module and a fast spatial pyramid pooling module;

[0099] Accordingly, the backbone network processing unit is also specifically used for:

[0100] Inputting the seeding video data into a first partial convolution module to obtain a first partial convolution feature image;

[0101] Inputting the first partial convolution feature image into the second partial convolution module to obtain a second partial convolution feature image;

[0102] Inputting the second part of the convolution feature image into the first C2f-DAR module to obtain a first C2f-DAR processed feature image;

[0103] Inputting the first C2f-DAR processed feature image into a third partial convolution module to obtain a third partial convolution feature image;

[0104] Inputting the third part of the convolution feature image into the first space depth conversion convolution module to obtain a first space depth conversion convolution feature image;

[0105] Inputting the first spatial depth conversion convolution feature image into a second C2f-DAR module to obtain a second C2f-DAR processed feature image;

[0106] Inputting the second C2f-DAR processed feature image into a fourth partial convolution module to obtain a fourth partial convolution feature image;

[0107] Inputting the fourth part of the convolution feature image into the second space depth conversion convolution module to obtain a second space depth conversion convolution feature image;

[0108] Inputting the second spatial depth conversion convolution feature image into a third C2f-DAR module to obtain a third C2f-DAR processed feature image;

[0109] Inputting the third C2f-DAR processed feature image into the fifth partial convolution module to obtain a fifth partial convolution feature image;

[0110] Inputting the fifth part of the convolution feature image into a third space depth conversion convolution module to obtain a third space depth conversion convolution feature image;

[0111] Inputting the third spatial depth conversion convolution feature image into a fourth C2f-DAR module to obtain a fourth C2f-DAR processed feature image;

[0112] The fourth C2f-DAR processed feature image is input into a fast spatial pyramid pooling module to obtain a seeding feature image.

[0113] In some optional embodiments, the C2f-DAR module includes a first receptive field attention convolution module, an ACmix module, a Split module, a first Bottleneck-R module, a second Bottleneck-R module, a third Bottleneck-R module and a second receptive field attention convolution module, wherein the output of the first receptive field attention convolution module is used as the input of the ACmix module, the output of the ACmix module is used as the input of the Split module, the output of the Split module is used as the input of the first Bottleneck-R module, the output of the first Bottleneck-R module is used as the input of the second Bottleneck-R module, and the output of the second Bottleneck-R module is used as the input of the third Bottleneck-R module; the output of the first Bottleneck-R module, the output of the second Bottleneck-R module, the output of the third Bottleneck-R module and the output of the Split module are added as the input of the second receptive field attention convolution module;

[0114] The Bottleneck-R module includes a third receptive field attention convolution module and a fourth receptive field attention convolution module, wherein the output of the third receptive field attention convolution module is used as the input of the fourth receptive field attention convolution module, and the output of the fourth receptive field attention convolution module and the input of the third receptive field attention convolution module are added as the output of the Bottleneck-R module.

[0115] In some optional embodiments, the progressive feature pyramid network module includes a fifth C2f-DAR module, a sixth C2f-DAR module, and a seventh C2f-DAR module;

[0116] The seeding fusion characteristic image includes a first seeding fusion characteristic image, a second seeding fusion characteristic image and a third seeding fusion characteristic image;

[0117] Correspondingly, the progressive feature pyramid network processing unit is also specifically used for:

[0118] Adding the first C2f-DAR processed feature image and the upsampled image corresponding to the second C2f-DAR processed feature image, and inputting the resultant image into a fifth C2f-DAR module to obtain a fifth C2f-DAR processed feature image;

[0119] Adding the down-sampled image corresponding to the first C2f-DAR processing feature image, the second C2f-DAR processing feature image, and the up-sampled image corresponding to the third C2f-DAR processing feature image, and inputting the sum into a sixth C2f-DAR module to obtain a sixth C2f-DAR processing feature image;

[0120] Adding the down-sampled image corresponding to the second C2f-DAR processing feature image and the third C2f-DAR processing feature image, and inputting the resultant image into a seventh C2f-DAR module to obtain a seventh C2f-DAR processing feature image;

[0121] Adding the fifth C2f-DAR processed feature image, the upsampled image corresponding to the sixth C2f-DAR processed feature image, and the upsampled image corresponding to the seeding feature image to obtain a first seeding fusion feature image;

[0122] Adding the down-sampled image corresponding to the fifth C2f-DAR processing feature image, the sixth C2f-DAR processing feature image, and the up-sampled image corresponding to the seventh C2f-DAR processing feature image to obtain a second seed fusion feature image;

[0123] Adding the down-sampled image corresponding to the sixth C2f-DAR processing feature image, the seventh C2f-DAR processing feature image, and the up-sampled image corresponding to the seeding feature image to obtain a third seeding fusion feature image;

[0124] Correspondingly, the detection head module includes a first detection head module, a second detection head module and a third detection head module;

[0125] Correspondingly, the detection head processing unit is also specifically used for:

[0126] Inputting the first seed row fusion feature image into a first detection head module to obtain a first seed carrying status prediction result;

[0127] Inputting the second seed row fusion feature image into the second detection head module to obtain a second seed carrying status prediction result;

[0128] Inputting the third seed row fusion feature image into a third detection head module to obtain a third seed carrying status prediction result;

[0129] The seed carrying condition prediction result of the seed row hole is determined based on the first seed carrying condition prediction result, the second seed carrying condition prediction result and the third seed carrying condition prediction result.

[0130] In some optional implementations, the seeding missed rate determination module 430 includes:

[0131] The missing-seed type hole counting unit is used to draw rays in any direction from the center of the detection frame of the seeding type hole in the missing-seed state. If the number of intersections between the rays and the preset rectangular area is an odd number, the missing-seed type hole count is increased by one until the traversal of each seeding video frame in the seeding video data is completed to obtain the target missing-seed type hole count value.

[0132] In some optional embodiments, the seeding detection device further includes:

[0133] The abnormal seeding processing module is used to generate a missed seeding alarm signal if the missed seeding rate is greater than a preset missed seeding rate threshold; and / or determine the current seeding speed, determine the seeding speed difference based on the current seeding speed and the preset seeding speed, and perform closed-loop control of the motor speed based on the seeding speed difference.

[0134] The seeding detection device provided in the embodiment of the present invention can execute the seeding detection method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0135] Embodiment 5

[0136] Fig.10 A schematic diagram of the structure of an electronic device 10 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0137] like Fig.10 As shown, the electronic device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. The I / O interface 15 is also connected to the bus 14.

[0138] A number of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0139] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as a seeding detection method, which includes:

[0140] Obtaining seeding video data;

[0141] Inputting the seeding video data into a pre-trained seeding hole seed carrying condition prediction model to obtain a seeding hole seed carrying condition prediction result;

[0142] When the seed carrying status prediction result of the seed-discharging type hole is a seed-deficient state, the seed-discharging type holes in the seed-deficient state are tracked and counted to obtain target seed-deficient type hole count values, and the seed-discharging missed planting rate is determined based on the target seed-deficient type hole count values.

[0143] In some embodiments, the seeding detection method may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the seeding detection method described above may be performed. Alternatively, in other embodiments, the processor 11 may be configured to perform the seeding detection method in any other appropriate manner (e.g., by means of firmware).

[0144] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0145] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0146] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, device, or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0147] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0148] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0149] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0150] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0151] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A seeding detection method, characterized in that: include: Obtaining seeding video data; Inputting the seeding video data into a pre-trained seeding hole seed carrying condition prediction model to obtain a seeding hole seed carrying condition prediction result; When the seed carrying status prediction result of the seed-discharging type hole is a seed-deficient state, tracking and counting the seed-discharging type holes in the seed-deficient state to obtain a target seed-deficient type hole count value, and determining the seed-discharging missed planting rate based on the target seed-deficient type hole count value; The seed-carrying status prediction model of the seed-row hole comprises a backbone network module, a progressive feature pyramid network module and a detection head module; Correspondingly, the seeding video data is input into a pre-trained seeding hole seed carrying condition prediction model to obtain a seeding hole seed carrying condition prediction result, including: Inputting the seeding video data into the backbone network module to obtain a seeding feature image; Inputting the seeding feature image into the progressive feature pyramid network module to obtain a seeding fusion feature image; The seeding fusion feature image is input into the detection head module to obtain the seed carrying status prediction result of the seeding hole; The backbone network module includes a first partial convolution module, a second partial convolution module, a first C2f-DAR module, a third partial convolution module, a first spatial depth conversion convolution module, a second C2f-DAR module, a fourth partial convolution module, a second spatial depth conversion convolution module, a third C2f-DAR module, a fifth partial convolution module, a third spatial depth conversion convolution module, a fourth C2f-DAR module and a fast spatial pyramid pooling module; Accordingly, the step of inputting the seeding video data into the backbone network module to obtain a seeding feature image includes: Inputting the seeding video data into a first partial convolution module to obtain a first partial convolution feature image; Inputting the first partial convolution feature image into the second partial convolution module to obtain a second partial convolution feature image; Inputting the second part of the convolution feature image into the first C2f-DAR module to obtain a first C2f-DAR processed feature image; Inputting the first C2f-DAR processed feature image into a third partial convolution module to obtain a third partial convolution feature image; Inputting the third part of the convolution feature image into the first space depth conversion convolution module to obtain a first space depth conversion convolution feature image; Inputting the first spatial depth conversion convolution feature image into a second C2f-DAR module to obtain a second C2f-DAR processed feature image; Inputting the second C2f-DAR processed feature image into a fourth partial convolution module to obtain a fourth partial convolution feature image; Inputting the fourth part of the convolution feature image into the second space depth conversion convolution module to obtain a second space depth conversion convolution feature image; Inputting the second spatial depth conversion convolution feature image into a third C2f-DAR module to obtain a third C2f-DAR processed feature image; Inputting the third C2f-DAR processed feature image into the fifth partial convolution module to obtain a fifth partial convolution feature image; Inputting the fifth part of the convolution feature image into a third space depth conversion convolution module to obtain a third space depth conversion convolution feature image; Inputting the third spatial depth conversion convolution feature image into a fourth C2f-DAR module to obtain a fourth C2f-DAR processed feature image; The fourth C2f-DAR processed feature image is input into a fast spatial pyramid pooling module to obtain a seeding feature image.

2. The method according to claim 1, characterized in that The C2f-DAR module includes a first receptive field attention convolution module, an ACmix module, a Split module, a first Bottleneck-R module, a second Bottleneck-R module, a third Bottleneck-R module and a second receptive field attention convolution module, wherein the output of the first receptive field attention convolution module is used as the input of the ACmix module, the output of the ACmix module is used as the input of the Split module, the output of the Split module is used as the input of the first Bottleneck-R module, the output of the first Bottleneck-R module is used as the input of the second Bottleneck-R module, and the output of the second Bottleneck-R module is used as the input of the third Bottleneck-R module; the output of the first Bottleneck-R module, the output of the second Bottleneck-R module, the output of the third Bottleneck-R module and the output of the Split module are added together as the input of the second receptive field attention convolution module; The Bottleneck-R module includes a third receptive field attention convolution module and a fourth receptive field attention convolution module, wherein the output of the third receptive field attention convolution module is used as the input of the fourth receptive field attention convolution module, and the output of the fourth receptive field attention convolution module and the input of the third receptive field attention convolution module are added as the output of the Bottleneck-R module.

3. The method according to claim 1, characterized in that: The progressive feature pyramid network module includes a fifth C2f-DAR module, a sixth C2f-DAR module, and a seventh C2f-DAR module; The seeding fusion characteristic image includes a first seeding fusion characteristic image, a second seeding fusion characteristic image and a third seeding fusion characteristic image; Accordingly, the seeding feature image is input into the progressive feature pyramid network module to obtain a seeding fusion feature image, including: Adding the first C2f-DAR processed feature image and the upsampled image corresponding to the second C2f-DAR processed feature image, and inputting the resultant image into a fifth C2f-DAR module to obtain a fifth C2f-DAR processed feature image; Adding the down-sampled image corresponding to the first C2f-DAR processing feature image, the second C2f-DAR processing feature image, and the up-sampled image corresponding to the third C2f-DAR processing feature image, and inputting the sum into a sixth C2f-DAR module to obtain a sixth C2f-DAR processing feature image; Adding the down-sampled image corresponding to the second C2f-DAR processing feature image and the third C2f-DAR processing feature image, and inputting the resultant image into a seventh C2f-DAR module to obtain a seventh C2f-DAR processing feature image; Adding the fifth C2f-DAR processed feature image, the upsampled image corresponding to the sixth C2f-DAR processed feature image, and the upsampled image corresponding to the seeding feature image to obtain a first seeding fusion feature image; Adding the down-sampled image corresponding to the fifth C2f-DAR processing feature image, the sixth C2f-DAR processing feature image, and the up-sampled image corresponding to the seventh C2f-DAR processing feature image to obtain a second seed fusion feature image; Adding the down-sampled image corresponding to the sixth C2f-DAR processing feature image, the seventh C2f-DAR processing feature image, and the up-sampled image corresponding to the seeding feature image to obtain a third seeding fusion feature image; Correspondingly, the detection head module includes a first detection head module, a second detection head module and a third detection head module; Accordingly, the seeding fusion feature image is input into the detection head module to obtain the seed carrying status prediction result of the seeding hole, including: Inputting the first seed row fusion feature image into a first detection head module to obtain a first seed carrying status prediction result; Inputting the second seed row fusion feature image into the second detection head module to obtain a second seed carrying status prediction result; Inputting the third seed row fusion feature image into a third detection head module to obtain a third seed carrying status prediction result; The seed carrying condition prediction result of the seed row hole is determined based on the first seed carrying condition prediction result, the second seed carrying condition prediction result and the third seed carrying condition prediction result.

4. The method according to claim 1, characterized in that: The tracking and counting of the seed-missing type holes to obtain the target seed-missing type hole count value includes: For rays extending in any direction from the center of the detection frame of the seed-missing type hole, if the number of intersections between the rays and the preset rectangular area is an odd number, the count of the seed-missing type hole is increased by one until the traversal of each seed-missing video frame in the seed-missing video data is completed to obtain the target seed-missing type hole count value.

5. The method according to any one of claims 1 to 4, characterized in that: After determining the seeding missed rate based on the target seed-missing hole count value, the method includes: If the seeding missed rate is greater than a preset missed rate threshold, a missed alarm signal is generated; And / or, determining a current seeding speed, determining a seeding speed difference based on the current seeding speed and a preset seeding speed, and performing closed-loop control on the motor speed based on the seeding speed difference.

6. A seeding detection device, characterized in that: include: A seeding video data acquisition module is used to acquire seeding video data; A seed-carrying condition prediction module for the seed-planting holes is used to input the seed-planting video data into a pre-trained seed-carrying condition prediction model for the seed-planting holes to obtain a prediction result for the seed-carrying condition of the seed-planting holes; A seeding missed planting rate determination module is used to track and count the seeding type holes in the seed-missing state when the seed carrying status prediction result of the seeding type hole is a seed-missing state, obtain a target seed-missing type hole count value, and determine the seeding missed planting rate based on the target seed-missing type hole count value; The seed-carrying status prediction model of the seed-row hole comprises a backbone network module, a progressive feature pyramid network module and a detection head module; Correspondingly, the hole seed carrying condition prediction module includes: A backbone network processing unit, used for inputting the seeding video data into the backbone network module to obtain a seeding feature image; A progressive feature pyramid network processing unit, used for inputting the seeding feature image into the progressive feature pyramid network module to obtain a seeding fusion feature image; A detection head processing unit is used to input the seeding fusion feature image into the detection head module to obtain a prediction result of the seeding status of the seeding hole; The backbone network module includes a first partial convolution module, a second partial convolution module, a first C2f-DAR module, a third partial convolution module, a first spatial depth conversion convolution module, a second C2f-DAR module, a fourth partial convolution module, a second spatial depth conversion convolution module, a third C2f-DAR module, a fifth partial convolution module, a third spatial depth conversion convolution module, a fourth C2f-DAR module and a fast spatial pyramid pooling module; Accordingly, the backbone network processing unit is also specifically used for: Inputting the seeding video data into a first partial convolution module to obtain a first partial convolution feature image; Inputting the first partial convolution feature image into the second partial convolution module to obtain a second partial convolution feature image; Inputting the second part of the convolution feature image into the first C2f-DAR module to obtain a first C2f-DAR processed feature image; Inputting the first C2f-DAR processed feature image into a third partial convolution module to obtain a third partial convolution feature image; Inputting the third part of the convolution feature image into the first space depth conversion convolution module to obtain a first space depth conversion convolution feature image; Inputting the first spatial depth conversion convolution feature image into a second C2f-DAR module to obtain a second C2f-DAR processed feature image; Inputting the second C2f-DAR processed feature image into a fourth partial convolution module to obtain a fourth partial convolution feature image; Inputting the fourth part of the convolution feature image into the second space depth conversion convolution module to obtain a second space depth conversion convolution feature image; Inputting the second spatial depth conversion convolution feature image into a third C2f-DAR module to obtain a third C2f-DAR processed feature image; Inputting the third C2f-DAR processed feature image into the fifth partial convolution module to obtain a fifth partial convolution feature image; Inputting the fifth part of the convolution feature image into a third space depth conversion convolution module to obtain a third space depth conversion convolution feature image; Inputting the third spatial depth conversion convolution feature image into a fourth C2f-DAR module to obtain a fourth C2f-DAR processed feature image; The fourth C2f-DAR processed feature image is input into a fast spatial pyramid pooling module to obtain a seeding feature image.

7. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively coupled to the at least one processor; The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the seeding detection method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the seeding detection method according to any one of claims 1 to 5 when executed.