Straw detection method, system and equipment based on land parcel straw detection model
By performing clustering and detection on images of target plots, the problems of low accuracy and low efficiency in traditional straw detection methods are solved, achieving efficient and accurate straw detection.
Patent Information
- Application Number
- CN202610036274.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-13
- Publication Date
- 2026-02-10
AI Technical Summary
Traditional straw detection methods are based on single-image analysis, which suffers from low detection accuracy and low execution efficiency. In particular, they face high runtime pressure when the image data volume is large, and the model has poor robustness and generalization ability.
A straw detection model based on land plots is used to cluster several target images of the target land plots into clusters with single straw features. The straw detection model is then used to detect the clusters to obtain the straw detection results of the land plots and images.
It improves the accuracy and robustness of straw detection, can process batch images simultaneously, improves detection efficiency, and reduces the need for manual judgment.
Smart Images
Figure CN121505310A_ABST
Abstract
Description
Technical Field
[0001] This disclosure relates to the field of image processing technology, and in particular to a straw detection method, system and device based on a plot straw detection model. Background Technology
[0002] With the development of agricultural automation, the demand for detection of straw returning to the field and straw burning has placed higher requirements on straw identification and monitoring.
[0003] Traditional straw detection methods are mostly based on single-image analysis. In practical applications, batches of images are collected. Different images have different characteristics, and images collected at different times, locations, and with different devices are diverse. Batch images also suffer from problems such as large differences in scenes, significant changes in lighting, and different shooting angles. Single-image analysis models cannot account for all situations. The models have low robustness and poor generalization ability, resulting in low detection accuracy and causing agricultural losses. At the same time, when the amount of image data is very large, single-image detection puts a great burden on the running time and has low detection efficiency. Summary of the Invention
[0004] The technical problem to be solved by this disclosure is to overcome the shortcomings of existing technologies that use single image analysis for straw detection, such as low detection accuracy and low execution efficiency, and to provide a straw detection method, system and equipment based on a plot straw detection model.
[0005] This disclosure solves the above-mentioned technical problems through the following technical solution:
[0006] This disclosure provides a straw detection method based on a plot straw detection model, the straw detection method comprising:
[0007] Acquire several target images of the target plot;
[0008] The target images are clustered using a pre-trained plot straw detection model to obtain several clusters, each cluster comprising several target images with a single straw feature.
[0009] The straw detection model is used to detect several clusters to obtain straw detection results for the target plot and straw detection results for each target image. The straw detection results for the plot are used to characterize the straw features that match the target plot, and the straw detection results for the image are used to characterize the straw features of the corresponding target image.
[0010] Optionally, the step of clustering several target images using a pre-trained plot straw detection model to obtain several clusters includes:
[0011] Extract the feature vector corresponding to each of the target images;
[0012] Based on the feature vectors, the similarity between different target images is obtained;
[0013] Based on the similarity, several target images are clustered to obtain several clusters.
[0014] Optionally, the step of detecting several clusters using the plot straw detection model to obtain the plot straw detection result corresponding to the target plot and the image straw detection result corresponding to each target image includes:
[0015] Based on several clusters, global features of the land parcel corresponding to the target land parcel and image features corresponding to each target image are obtained. The global features of the land parcel are used to characterize the overall features of all target images corresponding to the target land parcel.
[0016] Based on the global features of the land parcel and the image features, the straw detection results of the target land parcel and the straw detection results of each target image are obtained.
[0017] Optionally, the step of detecting several clusters using the plot straw detection model to obtain the plot straw detection result corresponding to the target plot and the image straw detection result corresponding to each target image includes:
[0018] Obtain the global clustering feature corresponding to each cluster, and the global clustering feature is used to characterize the overall features of all target images contained in the corresponding cluster;
[0019] Based on the cluster and the global cluster features, the global features of the target plot and the image features of each target image are obtained. The global plot features are used to characterize the overall features of all the target images corresponding to the target plot.
[0020] Based on the global features of the land parcel and the image features, the straw detection results of the target land parcel and the straw detection results of each target image are obtained.
[0021] Optionally, the plot straw detection results and image straw detection results include at least one of the following: straw type, straw quantity, soil condition, and straw burning status.
[0022] Optionally, the loss function of the straw detection model for the plot includes at least one of the cross-entropy loss function and the mean squared error loss function.
[0023] This disclosure also provides a straw detection system based on a plot straw detection model, the straw detection system comprising:
[0024] The image acquisition module is used to acquire several target images of the target plot;
[0025] The clustering module is used to perform clustering processing on several target images using a pre-trained plot straw detection model to obtain several clusters, each of which includes several target images with a single straw feature;
[0026] The detection module is used to detect several clusters using the plot straw detection model to obtain plot straw detection results corresponding to the target plot and image straw detection results corresponding to each target image. The plot straw detection results are used to characterize the straw features that match the target plot, and the image straw detection results are used to characterize the straw features of the corresponding target image.
[0027] Optionally, the clustering module includes:
[0028] The feature vector extraction unit is used to extract the feature vector corresponding to each target image;
[0029] A similarity acquisition unit is used to obtain the similarity between different target images based on the feature vector;
[0030] A clustering unit is used to perform clustering processing on several target images based on the similarity to obtain several clusters.
[0031] Optionally, the detection module includes:
[0032] The first feature acquisition unit is used to obtain, based on several clusters, the global features of the land parcel corresponding to the target land parcel and the image features corresponding to each target image, wherein the global features of the land parcel are used to characterize the overall features of all the target images corresponding to the target land parcel;
[0033] The first result acquisition unit is used to obtain the straw detection result of the target plot and the image straw detection result of each target image based on the global features of the plot and the image features.
[0034] Optionally, the detection module includes:
[0035] The second feature acquisition unit is used to acquire the cluster global feature corresponding to each cluster, wherein the cluster global feature is used to characterize the overall features of all target images contained in the corresponding cluster;
[0036] The third feature acquisition unit is used to obtain the global features of the target land parcel and the image features of each target image based on the cluster and the global features of the cluster. The global features of the land parcel are used to characterize the overall features of all the target images corresponding to the target land parcel.
[0037] The second result acquisition unit is used to obtain the straw detection result of the target plot and the image straw detection result of each target image based on the global features of the plot and the image features.
[0038] Optionally, the plot straw detection results and image straw detection results include at least one of the following: straw type, straw quantity, soil condition, and straw burning status.
[0039] Optionally, the loss function of the straw detection model for the plot includes at least one of the cross-entropy loss function and the mean squared error loss function.
[0040] This disclosure also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and used to run on the processor, wherein the processor executes the computer program to implement the straw detection method based on the plot straw detection model described above.
[0041] This disclosure also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the straw detection method based on a plot straw detection model described above.
[0042] This disclosure also provides a computer program product, including a computer program, characterized in that, when the computer program is executed by a processor, it implements the straw detection method based on the plot straw detection model as described above.
[0043] Based on common knowledge in the field, the above-mentioned preferred conditions can be combined arbitrarily to obtain various preferred embodiments of this disclosure.
[0044] The positive and progressive effects of this disclosure are as follows:
[0045] This disclosure involves clustering several target images of a target plot, then detecting the resulting clusters to obtain the straw detection results for the target plot and the straw detection results for each target image. This method clusters complex target images with multiple features into single-feature clusters, overcoming the interference between features of different target images. Only the target image corresponding to the single feature of the cluster needs to be detected, improving detection accuracy and robustness. Furthermore, it allows for simultaneous detection of batches of images, comprehensively identifying the entire plot without requiring manual judgment, thus improving detection efficiency. Attached Figure Description
[0046] Figure 1 A flowchart of a straw detection method based on a plot straw detection model provided in Embodiment 1 of this disclosure;
[0047] Figure 2 A specific example diagram of a straw detection method based on a plot straw detection model provided in Embodiment 1 of this disclosure;
[0048] Figure 3 The flowchart of step S102 in the straw detection method based on the straw detection model of the plot provided in Embodiment 1 of this disclosure;
[0049] Figure 4 This is a specific example diagram of step S102 in a straw detection method based on a plot straw detection model provided in Embodiment 1 of this disclosure;
[0050] Figure 5 This is the first flowchart of step S103 in a straw detection method based on a plot straw detection model provided in Embodiment 1 of this disclosure;
[0051] Figure 6 This is a first specific example diagram of step S103 in a straw detection method based on a plot straw detection model provided in Embodiment 1 of this disclosure;
[0052] Figure 7 This is a second flowchart of step S103 in a straw detection method based on a plot straw detection model provided in Embodiment 1 of this disclosure;
[0053] Figure 8 This is a specific example diagram of the first detection sub-model in a straw detection method based on a plot straw detection model provided in Embodiment 1 of this disclosure;
[0054] Figure 9 This is a specific example diagram of the second detection sub-model in a straw detection method based on a plot straw detection model provided in Embodiment 1 of this disclosure;
[0055] Figure 10 This is a specific example diagram illustrating a straw detection method based on a plot straw detection model provided in Embodiment 1 of this disclosure, where the straw percentage is zero.
[0056] Figure 11 The figure shows a specific example of a straw detection method based on a plot straw detection model provided in Embodiment 1 of this disclosure, where the proportion of straw is scattered to a small amount.
[0057] Figure 12 This is a specific example diagram illustrating the proportion of straw in the straw detection method based on a plot straw detection model provided in Embodiment 1 of this disclosure, where the straw percentage is the amount of stubble standing.
[0058] Figure 13 The first specific example diagram shows the proportion of straw in the straw detection method based on the straw detection model of the plot provided in Embodiment 1 of this disclosure, where the straw is fully covered.
[0059] Figure 14 This is a schematic diagram of the first module of a straw detection system based on a plot straw detection model provided in Embodiment 2 of this disclosure;
[0060] Figure 15 This is a schematic diagram of the second module of a straw detection system based on a plot straw detection model provided in Embodiment 2 of this disclosure;
[0061] Figure 16 This is a schematic diagram of the structure of an electronic device provided in Embodiment 3 of this disclosure. Detailed Implementation
[0062] The present disclosure is further illustrated below by way of embodiments, but the present disclosure is not limited to the scope of the embodiments described herein.
[0063] The prefixes such as "first" and "second" used in this disclosure are merely for distinguishing different descriptive objects and do not limit the position, order, priority, quantity, or content of the described objects. The use of ordinal numbers and other prefixes to distinguish descriptive objects in this disclosure does not constitute a limitation on the described objects. The description of the described objects is given in the context of the embodiments, and the use of such prefixes should not constitute unnecessary restrictions. Furthermore, in the description of this embodiment, unless otherwise stated, "multiple" means two or more.
[0064] Example 1
[0065] This disclosure provides a straw detection method based on a plot straw detection model, such as... Figure 1 As shown, the straw detection method includes:
[0066] S101. Obtain several target images of the target plot;
[0067] S102. Clustering is performed on several target images using a pre-trained plot straw detection model to obtain several clusters, each cluster including several target images with a single straw feature.
[0068] S103. Detect several clusters using the plot straw detection model to obtain the plot straw detection results corresponding to the target plot and the image straw detection results corresponding to each target image. The plot straw detection results are used to characterize the straw features that match the target plot, and the image straw detection results are used to characterize the straw features of the corresponding target image.
[0069] Specifically, such as Figure 2 As shown, several target images A1 of the target plot are acquired using image acquisition devices, including but not limited to cameras and mobile phones. The acquired target images may contain straw in the target plot, or they may be images unrelated to straw.
[0070] A pre-trained straw detection model is used to cluster several target images to obtain several clusters. The collected target images are input into the straw detection model, and its clustering sub-model divides the target images into multiple clusters. Each cluster contains multiple target images with a single straw feature. For example, the clusters could be cluster B1 (first cluster) or cluster B2 (second cluster). The straw features include at least one of the following: straw type, straw percentage, soil condition, and straw burning status.
[0071] The straw detection model detects straw in several clusters of crops, obtaining straw detection results for the target crop and straw detection results for each target image. Specifically, the clustering features corresponding to each cluster are input into the straw detection model to output the straw detection result for each cluster. Multiple clustering features include, for example, the first clustering feature C1 and the second clustering feature C2, and multiple straw detection results include, for example, the first clustering straw detection result D1 and the second clustering straw detection result D2. The clustering features corresponding to each cluster and the global feature C3 corresponding to the target crop are input into the straw detection model to output the straw detection result D3 for the target crop and straw detection results for each target image. Multiple straw detection results include, for example, the first image straw detection result D4 and the second image straw detection result D5.
[0072] The number of clusters is output by the plot straw detection model based on the actual situation; the number 2 mentioned above is just an example.
[0073] The straw detection model for land parcels uses clustering information to supplement the feature information of the entire land parcel, enabling the model to comprehensively identify the results of the entire land parcel without the need for manual judgment.
[0074] In this scheme, several target images of the target plot are clustered, and the resulting clusters are then detected to obtain the straw detection results for the target plot and the straw detection results for each target image. This method clusters complex target images with multiple features into single-feature clusters, overcoming the interference between features of different target images. Only the target images with single features corresponding to the clusters need to be detected, improving the accuracy and robustness of the detection. Furthermore, batch images can be detected simultaneously, enabling comprehensive identification of the entire plot without manual judgment, thus improving detection efficiency.
[0075] Optionally, such as Figure 3 As shown, step S102 includes:
[0076] S1021. Extract the feature vector corresponding to each target image;
[0077] S1022. Based on feature vectors, obtain the similarity between different target images;
[0078] S1023. Based on similarity, cluster several target images to obtain several clusters.
[0079] Specifically, such as Figure 4 As shown, during the clustering process of the straw detection model, a cluster is first input, which includes n target images A2, where n is a positive integer. The dimension of this cluster is [n, 3, 256, 256], where the size of each target image is [n, 3, 256, 256]. The 3 indicates that the target image is a color image with 3 color channels, and the two 256s represent the height and width of the target image, respectively, as 256 pixels. The image size is just an example and can be set according to actual conditions. Therefore, the input data dimension of the straw detection model is [n, 3, 256, 256].
[0080] Then, the n target images are combined into a set and input into the backbone (feature extraction backbone network) to obtain the feature vector C0, which has the dimension [n, d], where d represents the dimension of the feature vector, and d is, for example, 768.
[0081] Next, the similarity between different target images is calculated using feature vectors, with dimensions [n, n]. Finally, the cluster to which each target image belongs is determined using the similarity, with the cluster corresponding to a dimension [n, m], where m represents the number of clusters. The similarity matrix can be obtained by calculating the cosine of the angle between each pair of feature vectors.
[0082] Clustering can be performed using the connected component method, as follows: The similarity matrix is converted into a binary adjacency matrix: if S[i][j] ≥ threshold T, then a connection edge is established between i and j, where S represents the binary adjacency matrix, i and j represent the indexes of the target images in the binary adjacency matrix, and S[i][j] represents the similarity between the first target image i and the second target image j; in graph theory, this forms an undirected graph where nodes are target images and edges represent "sufficient similarity"; all connected components in the graph are found using a depth-first search or breadth-first search algorithm; each connected component corresponds to a cluster.
[0083] In this scheme, target images are clustered based on the similarity between different target images, ensuring the accuracy and reliability of the clusters.
[0084] In a feasible solution, such as Figure 5 As shown, step S103 includes:
[0085] S1031. Based on several clusters, obtain the global features of the target land parcel and the image features of each target image. The global features of the land parcel are used to characterize the overall features of all target images corresponding to the target land parcel.
[0086] S1032. Based on the global features of the land parcel and the image features, obtain the land parcel straw detection results corresponding to the target land parcel and the image straw detection results corresponding to each target image.
[0087] Specifically, such as Figure 6 As shown, each cluster obtained through clustering is used as the input of the plot straw detection model, with the feature vector C0 corresponding to each cluster as the input of the plot straw detection model. The input of the plot straw detection model consists of m feature vectors with dimensions [n1, d], where n1 is the number of target images in the cluster.
[0088] A global feature C3 of dimension [1, d] is generated and fused into the feature vectors of all clusters to obtain cluster features, which are m feature vectors of dimension [n1+1, d]. The cluster features are input into the Transformer (a type of neural network) module of the plot straw detection model for feature fusion and mutual learning, resulting in the global feature C7 corresponding to the target plot and the image features corresponding to each target image. Examples of image features for multiple target images are C8 and C9.
[0089] Finally, the global features of the target plot and the image features of each target image are input into the fully connected layer to obtain the straw detection result D3 of the target plot and the straw detection result of each target image. For example, the straw detection results of multiple images are the first image straw detection result D4 and the second image straw detection result D5.
[0090] In this scheme, the straw detection results for the target plot and the image straw detection results for each target image are obtained by using the global features of the target plot and the image features of each target image, thus ensuring the accuracy and reliability of the straw detection results for the plot and the image straw detection results.
[0091] In another feasible option, such as Figure 7 As shown, step S103 includes:
[0092] S1033. Obtain the global clustering feature corresponding to each cluster. The global clustering feature is used to characterize the overall features of all target images contained in the corresponding cluster.
[0093] S1034. Based on clustering clusters and global clustering features, obtain the global features of the target plot and the image features of each target image. The global features of the plot are used to characterize the overall features of all target images corresponding to the target plot.
[0094] S1035. Based on the global features of the land parcel and the image features, obtain the land parcel straw detection results corresponding to the target land parcel and the image straw detection results corresponding to each target image.
[0095] Specifically, the crop straw detection model includes a clustering sub-model, a first detection sub-model, and a second detection sub-model. The clustering sub-model performs clustering processing on several target images to obtain multiple clusters. The first detection sub-model uses the global clustering features corresponding to each cluster to obtain the clustered crop straw detection result for each cluster. The second detection sub-model inputs the global clustering features corresponding to each cluster and the global crop straw features corresponding to the target crop into the crop straw detection model to obtain the crop straw detection result for the target crop and the image crop straw detection result for each target image.
[0096] like Figure 8 As shown, each cluster obtained through clustering is used as the input of the first detection sub-model by the feature vector C0 corresponding to each cluster. The input of the plot straw detection model is m feature vectors with dimensions [n1, d], where n1 is the number of target images in the cluster.
[0097] A global feature C3 of dimension [1, d] is generated and fused into the feature vector of each cluster to obtain cluster features of dimension [m, n1+1, d]. These cluster features are the first cluster feature C1 and the second cluster feature C2. The cluster features are input into the first Transformer module of the first detection sub-model to obtain the global cluster feature corresponding to each cluster, which has dimension [m, d]. For example, the global cluster feature is the first global cluster feature C4 and the second global cluster feature C5.
[0098] Finally, the global clustering features corresponding to each cluster are input into a fully connected layer to obtain the clustered straw detection results for each cluster, with dimensions [m, c]. Multiple clustered straw detection results are, for example, the first clustered straw detection result D1 and the second clustered straw detection result D2. Here, c represents the straw features, thus obtaining m clustered straw detection results in sequence. Each clustered straw detection result includes the straw category, the proportion of straw, the soil condition, and the burning status of the straw.
[0099] like Figure 9 As shown, the clustering global feature C6, with dimensions [m+n, d], is obtained through the clustering sub-model and the first detection sub-model, where m represents the number of clusters and n represents the number of target images. Then, a global feature C3' with dimensions [1, d] is generated and fused into the clustering feature, with dimensions [m+n+1, d]. This is then fed into the second Transformer module of the second detection sub-model for feature fusion and mutual learning, resulting in the global feature C7 for the entire land parcel and the image features for each target image. Examples of image features for multiple target images are C8 and C9.
[0100] Finally, a fully connected layer is used to obtain the straw detection result D3 for the entire target plot and the straw detection result for each target image, with dimensions [1, c] and [n, c], respectively, where 1 represents the result for the entire plot and c represents the straw features. Multiple image straw detection results are, for example, the first image straw detection result D4 and the second image straw detection result D5.
[0101] In this scheme, the straw detection results for the target plot and the straw detection results for each target image are obtained by using the global clustering features corresponding to each cluster and the image features corresponding to each target image, thus ensuring the accuracy and reliability of the straw detection results for the plot and the straw detection results for the images.
[0102] In one feasible approach, the straw detection results and images of the plot include at least one of the following: straw type, straw quantity, soil condition, and straw burning status.
[0103] Straw categories include corn, soybeans, rice, wheat, etc. Straw percentage includes no straw, scattered to a small amount, scattered to a medium amount, scattered to a large amount, standing stubble, contiguous to a small amount, contiguous to a medium amount, contiguous to a large amount, and covered. Soil conditions include untouched soil, deep loosening, stubble removal, rotary tillage, harrowing, and ridging. Straw burning status includes no burning, burning in a few pictures, burning in nearly half of the pictures, burning in most pictures, and suspected burning. Figure 10 The image shows a specific example where the percentage of straw is zero. Figure 11 The image shows a specific example where the proportion of straw is scattered to a small amount. Figure 12 This is a specific example diagram showing the proportion of straw in the stubble. Figure 13 This is a specific example diagram showing the percentage of straw used to cover the entire surface.
[0104] This solution uses clustered straw detection results, such as straw type, straw proportion, soil condition, and straw burning status, to reflect multiple straw characteristics of the target plot and improve the practicality of the clustered straw detection results.
[0105] In one feasible approach, the loss function of the crop straw detection model includes at least one of the cross-entropy loss function and the mean squared error loss function.
[0106] Specifically, a training sample set is obtained to train the plot straw detection model. The training sample set includes several training sample data, each of which includes a plot sample image, the corresponding sample cluster, the corresponding sample cluster straw detection result, the corresponding plot straw detection result, and the corresponding image straw detection result.
[0107] The loss function of the straw detection model includes a clustering loss function, a first detection loss function, and a second detection loss function, and the corresponding calculation formulas are as follows:
[0108] ;
[0109] ;
[0110] ;
[0111] in, The loss function represents the straw detection model for land parcels; This represents the clustering loss function, corresponding to the clustering sub-model, which uses the cross-entropy loss function; This represents the first detection loss function, which corresponds to the first detection sub-model and uses the mean squared error loss function. This represents the second detection loss function, corresponding to the second detection sub-model; n represents the number of training sample data, y iThis represents the sample cluster of the i-th training sample data. Let z represent the predicted cluster of the i-th training sample data. i This represents the straw detection result of the clustering of the i-th training sample data. This represents the predicted clustering straw detection result for the i-th training sample data.
[0112] The second detection loss function includes the loss function L for the entire target plot. block_straw and the loss function L for each target image image_i_straw The corresponding calculation formula is as follows:
[0113] ;
[0114] Taking into account the type of straw, the proportion of straw, soil conditions, and the state of straw burning, in , , Of these three loss functions, with For example, the corresponding calculation formula is as follows:
[0115] ;
[0116] Wherein, a, b, c, and d represent the weighting coefficients corresponding to the type of straw, the proportion of straw, the soil condition, and the burning state of straw, respectively, and can be adjusted or set according to actual needs. , , , These represent the loss functions corresponding to the type of straw, the proportion of straw, the soil condition, and the burning state of straw, respectively. The specific form of the loss function is similar to that of the mean squared error loss function mentioned above, and will not be repeated here.
[0117] In this scheme, the loss function of the straw detection model is determined by the cross-entropy loss function and the mean square error loss function. The clustering process and detection process of the straw detection model are taken into account, which improves the accuracy and reliability of the straw detection model.
[0118] In this embodiment, several target images of the target plot are clustered, and the resulting clusters are detected to obtain the straw detection results for the target plot and the straw detection results for each target image. This method clusters complex target images with multiple features into clusters with single features, overcoming the interference between features of different target images. Only the target images with single features corresponding to the clusters need to be detected, improving the accuracy and robustness of the detection. Batch images can be detected simultaneously, and the results of the entire plot can be comprehensively identified without manual judgment, thus improving detection efficiency.
[0119] Example 2
[0120] Corresponding to the aforementioned embodiments of the straw detection method based on the plot straw detection model, this disclosure also provides embodiments of the straw detection system based on the plot straw detection model.
[0121] like Figure 14 As shown, the straw detection system includes:
[0122] Image acquisition module 1 is used to acquire several target images of the target plot;
[0123] Clustering module 2 is used to cluster several target images using a pre-trained plot straw detection model to obtain several clusters, each cluster including several target images with a single straw feature;
[0124] The detection module 3 is used to detect several clusters through the plot straw detection model, and obtain the plot straw detection results corresponding to the target plot and the image straw detection results corresponding to each target image. The plot straw detection results are used to characterize the straw features that match the target plot, and the image straw detection results are used to characterize the straw features of the corresponding target image.
[0125] In a feasible solution, such as Figure 15 As shown, clustering module 2 includes:
[0126] The feature vector extraction unit 21 is used to extract the feature vector corresponding to each target image;
[0127] The similarity acquisition unit 22 is used to obtain the similarity between different target images based on feature vectors;
[0128] Clustering unit 23 is used to perform clustering processing on several target images based on similarity to obtain several clusters.
[0129] In one feasible solution, the detection module 3 includes:
[0130] The first feature acquisition unit 31 is used to obtain the global features of the target land parcel and the image features of each target image based on several clusters. The global features of the land parcel are used to characterize the overall features of all target images corresponding to the target land parcel.
[0131] The first result acquisition unit 32 is used to obtain the straw detection results of the target plot and the image straw detection results of each target image based on the global features of the plot and the image features.
[0132] In one feasible solution, the detection module 3 includes:
[0133] The second feature acquisition unit 33 is used to acquire the cluster global features corresponding to each cluster. The cluster global features are used to characterize the overall features of all target images contained in the corresponding cluster.
[0134] The third feature acquisition unit 34 is used to obtain the global features of the target plot and the image features of each target image based on the cluster and global features of the cluster. The global features of the plot are used to characterize the overall features of all target images corresponding to the target plot.
[0135] The second result acquisition unit 35 is used to obtain the straw detection results of the target plot and the image straw detection results of each target image based on the global features of the plot and the image features.
[0136] In one feasible approach, the straw detection results and images of the plot include at least one of the following: straw type, straw quantity, soil condition, and straw burning status.
[0137] In one feasible approach, the loss function of the crop straw detection model includes at least one of the cross-entropy loss function and the mean squared error loss function.
[0138] For the system embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs.
[0139] In this embodiment, several target images of the target plot are clustered, and the resulting clusters are detected to obtain the straw detection results for the target plot and the straw detection results for each target image. This method clusters complex target images with multiple features into clusters with single features, overcoming the interference between features of different target images. Only the target images with single features corresponding to the clusters need to be detected, improving the accuracy and robustness of the detection. Batch images can be detected simultaneously, and the results of the entire plot can be comprehensively identified without manual judgment, thus improving detection efficiency.
[0140] Example 3
[0141] Figure 16This is a schematic diagram of the structure of an electronic device according to an example embodiment of the present disclosure. The electronic device includes a memory, a processor, and a computer program stored in the memory and used to run on the processor. When the processor executes the computer program, it implements the straw detection method based on the plot straw detection model described in any of the above embodiments. Figure 16 The electronic device 90 shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments disclosed herein.
[0142] like Figure 16 As shown, the electronic device 90 can be manifested as a general-purpose computing device, such as a server device. The components of the electronic device 90 may include, but are not limited to: at least one processor 91, at least one memory 92, and a bus 93 connecting different system components (including memory 92 and processor 91).
[0143] Bus 93 includes a data bus, an address bus, and a control bus.
[0144] The memory 92 may include volatile memory, such as random access memory (RAM) 921 and / or cache memory 922, and may further include read-only memory (ROM) 923.
[0145] The memory 92 may also include a program tool 925 (or utility) having a set (at least one) program module 924, such program module 924 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each or some combination of these examples may include an implementation of a network environment.
[0146] The processor 91 executes various functional applications and data processing by running computer programs stored in the memory 92, such as the straw detection method based on the plot straw detection model provided in any of the above embodiments.
[0147] Electronic device 90 can also communicate with one or more external devices 94 (e.g., keyboard, pointing device, etc.). This communication can be performed via input / output (I / O) interface 95. Furthermore, electronic device 90 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 96. Figure 16 As shown, network adapter 96 communicates with other modules of electronic device 90 via bus 93. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with electronic device 90, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (disk array) systems, tape drives, and data backup storage systems.
[0148] It should be noted that although several units / modules or sub-units / modules of the electronic device have been mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to embodiments of this disclosure, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided and embodied by multiple units / modules.
[0149] Example 4
[0150] This disclosure also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the straw detection method based on the plot straw detection model provided in any of the above embodiments.
[0151] The readable storage medium may be more specifically adopted, including but not limited to: portable disk, hard disk, random access memory, read-only memory, erasable programmable read-only memory, optical storage device, magnetic storage device, or any suitable combination thereof.
[0152] Example 5
[0153] This disclosure also provides a computer program product, including a computer program that, when executed by a processor, implements the straw detection method based on the plot straw detection model described above.
[0154] The program code for executing the computer program product of this disclosure can be written in any combination of one or more programming languages, and the program code can be executed entirely on a user device, partially on a user device, as a stand-alone software package, partially on a user device and partially on a remote device, or entirely on a remote device.
[0155] While specific embodiments of this disclosure have been described above, those skilled in the art should understand that these are merely illustrative examples, and the scope of protection of this disclosure is defined by the appended claims. Those skilled in the art can make various changes or modifications to these embodiments without departing from the principles and essence of this disclosure, but all such changes and modifications fall within the scope of protection of this disclosure.
Claims
1. A straw detection method based on a plot straw detection model, characterized in that, The straw detection method includes: Acquire several target images of the target plot; The target images are clustered using a pre-trained plot straw detection model to obtain several clusters, each cluster comprising several target images with a single straw feature. The straw detection model is used to detect several clusters to obtain straw detection results for the target plot and straw detection results for each target image. The straw detection results for the plot are used to characterize the straw features that match the target plot, and the straw detection results for the image are used to characterize the straw features of the corresponding target image.
2. The straw detection method based on the plot straw detection model as described in claim 1, characterized in that, The step of clustering several target images using a pre-trained land straw detection model to obtain several clusters includes: Extract the feature vector corresponding to each of the target images; Based on the feature vectors, the similarity between different target images is obtained; Based on the similarity, several target images are clustered to obtain several clusters.
3. The straw detection method based on the plot straw detection model as described in claim 1, characterized in that, The step of detecting straw in several clusters using the plot straw detection model to obtain the plot straw detection result corresponding to the target plot and the image straw detection result corresponding to each target image includes: Based on several clusters, global features of the land parcel corresponding to the target land parcel and image features corresponding to each target image are obtained. The global features of the land parcel are used to characterize the overall features of all target images corresponding to the target land parcel. Based on the global features of the land parcel and the image features, the straw detection results of the target land parcel and the straw detection results of each target image are obtained.
4. The straw detection method based on the plot straw detection model as described in claim 1, characterized in that, The step of detecting straw in several clusters using the plot straw detection model to obtain the plot straw detection result corresponding to the target plot and the image straw detection result corresponding to each target image includes: Obtain the global clustering feature corresponding to each cluster, and the global clustering feature is used to characterize the overall features of all target images contained in the corresponding cluster; Based on the cluster and the global cluster features, the global features of the target plot and the image features of each target image are obtained. The global plot features are used to characterize the overall features of all the target images corresponding to the target plot. Based on the global features of the land parcel and the image features, the straw detection results of the target land parcel and the straw detection results of each target image are obtained.
5. The straw detection method based on a plot straw detection model as described in any one of claims 1-4, characterized in that, The straw detection results and images of the plots include at least one of the following: straw type, straw quantity, soil condition, and straw burning status.
6. The straw detection method based on the plot straw detection model as described in claim 5, characterized in that, The loss function of the straw detection model for the plot includes at least one of the cross-entropy loss function and the mean square error loss function.
7. A straw detection system based on a plot straw detection model, characterized in that, The straw detection system includes: The image acquisition module is used to acquire several target images of the target plot; The clustering module is used to perform clustering processing on several target images using a pre-trained plot straw detection model to obtain several clusters, each of which includes several target images with a single straw feature; The detection module is used to detect several clusters using the plot straw detection model to obtain plot straw detection results corresponding to the target plot and image straw detection results corresponding to each target image. The plot straw detection results are used to characterize the straw features that match the target plot, and the image straw detection results are used to characterize the straw features of the corresponding target image.
8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and for running on the processor, characterized in that, When the processor executes the computer program, it implements the straw detection method based on the plot straw detection model as described in any one of claims 1-6.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the straw detection method based on the plot straw detection model as described in any one of claims 1-6.
10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the straw detection method based on the plot straw detection model as described in any one of claims 1-6.
Citation Information
Patent Citations
Image clustering method and device, computer equipment and storage medium
CN112749762A
Image retrieval method and device, computer equipment and storage medium
CN112784087A
Method for detecting abnormal straw picture and detection device
CN114743048A