A wheat red spider mites detection sample collection and statistical system and method

By designing the wheat Red Spider Measurement and Reporting Sample Collection and Statistics System, using smartphones and servers for image recognition and statistics, the problems of inaccurate counting, large workload and difficult sample retention in the existing technology are solved, and the rapid and accurate samples of Red Spider Measurement are achieved. Collect and efficient statistics.

CN113570588BActive Publication Date: 2024-12-03临汾市植保植检站临汾市农药监督管理站
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202110883528.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-08-03
Publication Date
2024-12-03
Estimated Expiration
2041-08-03

AI Technical Summary

Technical Problem

The existing wheat red spider measurement and reporting methods have problems such as inaccurate counting, large workload and low efficiency, and difficulty in sample retention.

Method used

A wheat red spider measurement and reporting sample collection and statistics system is designed, including a sample collection device and a sample statistics device. The collection device collects red spider through the sampling box and the box lid component, and uses the smartphone client and server to take images and pre-process them. , identify and count, and use convolutional neural network and support vector machine algorithm for image recognition and classification.

Benefits of technology

The rapid collection and accurate counting of Red Spider samples are achieved, which improves work efficiency, reduces the time and labor intensity of manual collection, and improves the sample retention rate.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113570588B_ABST
    Figure CN113570588B_ABST
Patent Text Reader

Abstract

The invention discloses a wheat red spider monitoring sample collection and statistics system and method, comprising a sample collection device and a sample statistics device; the sample collection device comprises a sampling box, a box cover assembly, and a vibration controller; the sample statistics device comprises a smart phone client and a server, the smart phone client is used to photograph the red spiders in the sample collection device, and transmit the collected images to the server with image recognition and statistics functions, and the server performs image preprocessing, image recognition and classification statistics; the novel wheat red spider monitoring sample collection and statistics system can realize the rapid collection, sample retention and accurate counting of wheat red spider samples in wheat fields, save time and labor, and improve sample collection and statistics efficiency.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of wheat red spider mite detection and reporting, and in particular relates to a wheat red spider mite detection and reporting sample collection and statistics system and method directly carried out in a wheat field. Background Art

[0002] At present, the wheat spider mites are detected and reported by manual survey, which is time-consuming and labor-intensive, and the counting is inaccurate. Its shortcomings are mainly reflected in the following three points:

[0003] 1. Inaccurate counting. At present, the number of wheat spider mites is manually surveyed by patting the wheat with hands to shake the spider mites on the plants onto white paper, white plastic cloth or white porcelain plates spread on the wheat rhizosphere to obtain samples. The disturbed spider mites move around quickly on the objects from which the samples are obtained, and the individuals are small, so it is very difficult to count them visually, and it is often impossible to count them accurately.

[0004] 2. Heavy workload and low efficiency. According to the technical specifications for wheat spider mites monitoring and reporting by the Ministry of Agriculture, a survey is conducted every 5 days from the time wheat turns green to the heading stage. The surveyed fields involve wheat fields with different ecological environments. 2 to 3 fields of each type need to be surveyed, and 5 diagonal sampling points are used for each field. The survey time is limited to 8 to 10 a.m. or 16 to 18 p.m. on the same day. Records are taken while surveying. The time is short, the task is heavy, the workload is heavy, and the work efficiency is low.

[0005] 3. Difficulty in retaining samples. Due to the large number of samples collected and the strong mobility of live mites, it is very difficult to retain samples.

[0006] The present invention comprehensively solves all the shortcomings of the existing methods. Summary of the invention

[0007] The technical problem to be solved by the present invention is to provide a wheat red spider detection sample collection and statistics system and method in view of the deficiencies in the prior art.

[0008] The technical solution of the present invention is as follows:

[0009] A wheat red spider mite detection and reporting sample collection and statistics system, comprising a sample collection device and a sample statistics device; the sample collection device comprises a sampling box (1), a box cover assembly (2), and a vibration controller (3); the sample statistics device comprises a smart phone client and a server, the smart phone client being used to photograph red spiders in the sample collection device and transmit the collected images to a server having image recognition and statistics functions, and performing image preprocessing, image recognition, and classification statistics on the server;

[0010] The sampling box (1) comprises a box body (11), a rotating shaft seat (12), and a box cover rotating shaft insertion guide groove (13); the box body (11) is a rectangular parallelepiped box body lacking a top surface and a side surface adjacent to the top surface, the missing side surface being an open end of the sampling box (1); when in use, the open end is aligned with the root of the wheat seedling; the missing top surface is used to accommodate a box cover (21) of a box cover assembly (2); when in use, the box cover (21) closes the missing top surface; two side surfaces on both sides of the open end are A rotating shaft seat (12) is provided near the top of the surface, the rotating shaft seat (12) being a circular hole, the diameter of the circular hole being slightly larger than the outer diameter of the circular cutting rotating shaft (22), and the circular cutting rotating shaft (22) being able to rotate freely in the circular hole; the box cover rotating shaft insertion guide groove (13) being a rectangular opening, the rectangular opening being communicated with the rotating shaft seat (12), the circular cutting rotating shaft (22) being inserted into the rotating shaft seat (12) through the rectangular opening, and the size of the rectangular opening being smaller than the outer diameter of the circular cutting rotating shaft (22).

[0011] The wheat red spider mite monitoring and reporting sample collection and statistical system, the box cover assembly (2) comprises a box cover (21), a circular rotating shaft (22), a battery and controller box (23) integrated with the box cover, a flapping plate guide column (24), a guide column limit retaining ring (25) and a flapping plate (26), four flapping plate guide columns (24) are fixed under the box cover (21), a vibration element (28) is also fixed under the box cover (21) and is located at the center of the four flapping plate guide columns (24), the flapping plate (26) is fixed on the vibration element (28), the flapping plate (26) has flapping plate ribs (27) under the flapping plate (26), and circular holes are provided at the four corners of the flapping plate (26), the four circular holes are penetrated on the four flapping plate guide columns (24), and the flapping plate (26) can vibrate freely on the four flapping plate guide columns (24); the guide column limit retaining ring (25) is fixed at the bottom end of the four flapping plate guide columns (24) for limiting the flapping plate (26).

[0012] In the wheat red spider monitoring sample collection and statistical system, the battery and controller box (23) is fixed together with a box cover (21) and one side of a flapping plate (26), and a circular cutting shaft (22) is arranged on the upper part of the battery and controller box (23). After the circular cutting shaft (22) is inserted into the shaft seat (12), when the box cover (21) and the flapping plate (26) are rotated to be parallel to the bottom surface of the box body (11) or perpendicular to the bottom surface of the box body (11), the battery and controller box (23) can block the open end of the sampling box (1).

[0013] In the wheat red spider detection and reporting sample collection and statistical system, the cross-section of the battery and the controller box (23) is in the shape of a (90) degree fan, and the open end is well sealed by the outer arc, and the arc of the battery and the controller box (23) will not cause damage to the wheat plants during the rotation process.

[0014] The sample collection method according to any of the sample collection and statistics systems comprises the following steps:

[0015] A1. Place the open end of the sampling box (1) directly against the root of the wheat seedling (6) to be sampled;

[0016] A2. Place the box cover assembly (2) with the vibrating element vertically, then insert the circular rotating shaft (22) on the side of the box cover into the guide groove (13) along the rotating shaft of the box cover and push it into the bottom of the groove of the sampling box (1) so that the rotating shaft reaches the rotating shaft seat (12);

[0017] A3. Closing the box cover assembly (2). During the installation of the box cover assembly (2), the battery and the controller box of the box cover assembly (2) push the upright wheat seedlings to press against the bottom edge of the sampling box (1). During the closing process, the arc-shaped battery and the controller box ensure that the wheat seedlings (6) are nearly horizontal when the cover is closed. After the cover is closed, the flapping plate and the flapping plate ribs (27) at the bottom of the box cover assembly are in contact with the wheat seedlings (6);

[0018] A4, turning on the vibration controller (3) to switch on the vibration element (28), the vibration element (28) drives the beating plate and the beating plate ribs to vibrate up and down quickly along the beating plate guide column (24) to beat the wheat;

[0019] A5. By vibrating and beating the wheat seedlings, the wheat spider mites fall into the bottom of the sampling box (1);

[0020] A6. After the beating is completed, turn off the power supply and remove the box cover assembly (2) to complete the sample collection.

[0021] According to any of the above-mentioned sample collection and statistical systems, the identification and statistical method comprises the following steps:

[0022] B1. Image collection: Collect red spider mites in the wild, and use the smartphone client to take photos of the red spider mites in the sample collection device;

[0023] B2. Image preprocessing and graphic transformation: The server receives the image to be processed transmitted by the smartphone client, removes the background, grays it, removes noise, and binarizes it, then uses a sliding window to locate the target and marks the target to be detected with a box;

[0024] (1) Input an image, use the lazy snapping method to remove the background of the image and grayscale it; then use Gaussian filtering to binarize the image;

[0025] (2) Use the selective search algorithm to extract and generate 2000 candidate regions from the binary image;

[0026] (3) Dilate the edge of each candidate region by p pixels, where p = 16, and frame the candidate region after dilation; transform the size of each candidate region to 227 x 227;

[0027] B3. Image recognition:

[0028] The tensorflow framework is used to construct a parallelized AlexNet convolutional neural network model. Each candidate region is input into the AlexNet convolutional neural network to obtain a 4096-dimensional feature vector to determine whether it is a spider or not, and the boxes of spiders are left for the next step.

[0029] The AlexNet convolutional neural network has two input ports. The input data is processed by the convolution layer conv, the activation function relu, the pooling layer pool and the batch normalization BN. The output data is cascaded merged by the parallel convolution layer conv and then connected by the fully connected layer fc to obtain the forward propagation result logits.

[0030] AlexNet convolutional neural network adopts transfer learning strategy, first pre-training on ImageNet, using convolutional neural network to fully learn shallow features, using SGD algorithm during training, initial learning rate of 0.001, batch_size of 128; then remove the last fully connected layer of pre-training, modify the number of categories, and users perform large-scale training on small-scale data sets. When user annotation data is insufficient, AlexNet convolutional neural network is fine-tuned to improve the ability to identify object types;

[0031] B4, import the feature vector derived from the neural network into the SVM classifier to generate the category score;

[0032] B5. Set the IoU indicator threshold (>0.3), calculate the IoU indicator, use non-maximum suppression, remove overlapping boxes based on the box with the highest score, and obtain the candidate box with the highest score in this category;

[0033] B6. Classification and statistics: Classification is to identify the two types of spiders and then count their numbers;

[0034] B7. Result output: The classification and statistical results are transmitted back to the smart phone client through the server communication module, and the smart phone client stores the identification and statistical results.

[0035] The use of a new wheat red spider monitoring sample collection and statistical system can realize the rapid collection, sample retention and accurate counting of wheat red spider samples in wheat fields, saving time and labor and improving sample collection and statistical efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 This is a schematic diagram of the composition of the wheat red spider mites monitoring sample collection and statistical system;.

[0037] Figure 2 This is a schematic diagram of the sampling box structure;

[0038] Figure 3 It is a schematic diagram of the structure of the box cover assembly;

[0039] Figure 4 The following is a schematic diagram of the method of use;

[0040] Figure 5 This is a schematic diagram of installation and usage;

[0041] Figure 6 It is a flowchart of the method for the smartphone client;

[0042] Figure 7 is a flow chart of the server method;

[0043] 1. Sampling box, including: 11. Box body, 12. Rotating shaft seat, 13. Box cover rotating shaft insertion guide groove; 2. Box cover assembly, including: 21. Box cover, 22. Circular cutting rotating shaft, 23. Battery and controller box, 24. Beating plate guide column, 25. Guide column limit retaining ring, 26. Beating plate, 27. Beating plate rib, 28. Vibration element; 3. Vibration controller; 4. Battery; 5. Battery box cover; 6. Wheat seedlings. DETAILED DESCRIPTION

[0044] The present invention is described in detail below in conjunction with specific embodiments.

[0045] like Figure 1-5 As shown, the wheat red spider detection and reporting sample collection and statistics system includes a sample collection device and a sample statistics device; the sample collection device includes a sampling box 1, a box cover assembly 2, and a vibration controller 3; the sample statistics device includes a smart phone client and a server, the smart phone client is used to take pictures of red spiders in the sample collection device, and transmit the collected images to a server with image recognition and statistics functions, and perform image preprocessing, image recognition and classification statistics on the server;

[0046] The sampling box 1 includes a box body 11, a rotating shaft seat 12, and a box cover rotating shaft insertion guide groove 13; the box body 11 is a rectangular box body lacking a top surface and a side surface adjacent to the top surface, the missing side surface is the open end of the sampling box 1, when in use, the open end is aligned with the roots of the wheat seedlings, the missing top surface is used to accommodate the box cover 21 of the box cover assembly 2, when in use, the box cover 21 closes the missing top surface; the two side surfaces on both sides of the open end are provided with a rotating shaft seat 12 near the top, the rotating shaft seat 12 is a circular hole, the diameter of the circular hole is slightly larger than the outer diameter of the circular cutting shaft 22, and the circular cutting shaft 22 can rotate freely in the circular hole; the box cover rotating shaft insertion guide groove 13 is a rectangular opening, the rectangular opening is connected to the rotating shaft seat 12, the circular cutting shaft 22 is inserted into the rotating shaft seat 12 through the rectangular opening, and the size of the rectangular opening is smaller than the outer diameter of the circular cutting shaft 22.

[0047] The box cover assembly 2 includes a box cover 21, a circular rotating shaft 22, a battery and controller box 23 integrated with the box cover, a flapping plate guide column 24, a guide column limit retaining ring 25 and a flapping plate 26. Four flapping plate guide columns 24 are fixed under the box cover 21. A vibration element 28 is also fixed under the box cover 21 and is located at the center of the four flapping plate guide columns 24. The flapping plate 26 is fixed on the vibration element 28. The flapping plate 26 has flapping plate ribs 27 (which are in contact with the wheat seedlings and apply vibration to the wheat seedlings) under the flapping plate 26. Circular holes are provided at the four corners of the flapping plate 26. The four circular holes are penetrated on the four flapping plate guide columns 24, so that the flapping plate 26 can vibrate freely on the four flapping plate guide columns 24. The guide column limit retaining ring 25 is fixed at the bottom end of the four flapping plate guide columns 24 to limit the flapping plate 26.

[0048] The battery and controller box 23 is fixed together with one side of the box cover 21 and the slapping plate 26. A circular cutting shaft 22 is set on the upper part of the battery and controller box 23. After the circular cutting shaft 22 is inserted into the shaft seat 12, when the box cover 21 and the slapping plate 26 are rotated to be parallel to the bottom surface of the box body 11 or perpendicular to the bottom surface of the box body 11, the battery and controller box 23 can block the open end of the sampling box 1 to prevent red spider mites from escaping. Preferably, the cross-section of the battery and controller box 23 is a 90-degree fan-shaped, and the outer circular arc is used to well block the open end to prevent red spider mites from escaping, and the battery and controller box 23 will not cause damage to the wheat plants during rotation.

[0049] The vibration element 28 is realized by a fixed coil and a spring, and a movable iron core; the vibration controller 3 is composed of an oscillating electronic circuit and a driving circuit, the frequency of the oscillating circuit is adjustable to control the beating frequency, and the output current of the driving circuit is adjustable to control the beating intensity; the battery 4 is composed of a rechargeable battery pack; the smart phone is selected with a high-definition camera with a magnification of 20 times.

[0050] 1. Sample collection

[0051] 1. Place the open end of the sampling box 1, which meets the sampling specifications and has a net bottom surface of 333mm×200mm, directly at the root of the wheat seedling 6 to be sampled (see Figure 4 );

[0052] 2. Place the box cover assembly 2 with the vibrating element vertically, then insert the circular rotating shaft 22 on the side of the box cover into the guide groove 13 along the rotating shaft of the box cover and push it into the bottom of the groove of the sampling box 1, so that the rotating shaft reaches the rotating shaft seat 12 (see Figure 4 );

[0053] 3. Close the box cover assembly 2. During the installation of the box cover assembly 2, the battery and controller box of the box cover assembly 2 push the upright wheat seedlings to press against the bottom edge of the sampling box 1. During the closing process, the arc-shaped battery and controller box ensure that the wheat seedlings 6 are nearly horizontal when the cover is closed. After the cover is closed, the flapping plate and the flapping plate ribs 27 at the bottom of the box cover assembly contact the wheat seedlings 6 (see Figure 5 );

[0054] 4. Turn on the vibration controller 3 to connect the vibration element 28. The vibration element 28 drives the beating plate and the beating plate ribs to vibrate up and down quickly along the beating plate guide column 24 to beat the wheat. The beating frequency and speed of the vibration element 28 are adjusted and controlled by the vibration controller 3, and the power of the vibration controller 3 is supplied by the battery 4.

[0055] 5. By vibrating and beating the wheat seedlings, the wheat spider mites fall into the bottom of the sampling box 1;

[0056] 6. After the beating is completed, turn off the power, remove the box cover assembly 2, and complete the sample collection.

[0057] 2. Image Acquisition, Recognition and Statistics

[0058] 1. Image acquisition: Red spiders are small in size, and image recognition requires a smartphone or digital camera with a high-definition lens with high magnification and close focal length and a high-resolution camera. Under appropriate lighting conditions, the red spiders in the sample collection device are photographed. When collecting red spiders in the wild, the smartphone client usually uses an App phone with image remote transmission and data return functions. The collected images are transmitted via the Internet to a server with image preprocessing, graphic transformation, image recognition, and statistical functions for image processing.

[0059] 2. Image preprocessing and graphic transformation: The server receives the image to be processed transmitted by the smartphone client, and first performs background removal, grayscale, noise removal, image binarization and other processes, then uses a sliding window to locate the target and marks the target to be detected with a box.

[0060] (1) Input a picture, use the lazy snapping method to remove the background of the image and grayscale it; then use Gaussian filtering to binarize the image;

[0061] (2) Use the selective search algorithm to extract and generate about 2000 candidate regions from the binary image;

[0062] (3) Dilate the edge of each candidate region by p = 16 pixels, add a frame to the candidate region after dilation, and transform the size of each candidate region to 227 x 227.

[0063] 3. Image recognition:

[0064] The tensorflow framework is used to construct a parallel AlexNet convolutional neural network model. Each candidate area is input into the AlexNet convolutional neural network to obtain a 4096-dimensional feature vector to determine whether it is a spider, and the boxes of spiders are left for the next step of processing.

[0065] The AlexNet convolutional neural network has two input ports. The input data is processed by the convolution layer conv, the activation function relu, the pooling layer pool and the batch normalization BN. The output data is cascaded merged by the parallel convolution layer conv and then connected by the fully connected layer fc to obtain the forward propagation result logits.

[0066] The AlexNet convolutional neural network adopts a transfer learning strategy. It is first pre-trained on ImageNet, using the convolutional neural network to fully learn shallow features. The SGD algorithm is used during training, with an initial learning rate of 0.001 and a batch_size of 128. Then the last fully connected layer of the pre-training is removed, and the number of categories is modified (3 categories = 2 targets + 1 background). Users perform large-scale training on small-scale data sets. When user annotation data is insufficient, the AlexNet convolutional neural network is fine-tuned to improve the ability to recognize object types.

[0067] 4. Import the feature vector derived from the neural network into the SVM classifier to generate category scores;

[0068] 5. Set the IoU indicator threshold (>0.3), calculate the IoU indicator, use non-maximum suppression, remove overlapping boxes based on the highest-scoring box, and obtain the candidate box with the highest score in this category.

[0069] 6. Classification and statistics: Classification is to identify the two types of spiders and then count their numbers.

[0070] 7. Result output: The classification and statistical results are transmitted back to the smart phone client through the server communication module, and the smart phone client stores the recognition and statistical results.

[0071] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all these improvements and changes should fall within the scope of protection of the appended claims of the present invention.

Claims

1. A wheat red spider mites detection and reporting sample collection and statistics system, It is characterized in that The invention comprises a sample collection device and a sample statistics device; the sample collection device comprises a sampling box (1), a box cover assembly (2), and a vibration controller (3); the sample statistics device comprises a smart phone client and a server, the smart phone client being used to photograph red spiders in the sample collection device and transmit the collected images to a server having image recognition and statistics functions, and performing image preprocessing, image recognition, and classification statistics on the server; The sampling box (1) comprises a box body (11), a rotating shaft seat (12), and a box cover rotating shaft insertion guide groove (13); the box body (11) is a rectangular parallelepiped box body lacking a top surface and a side surface adjacent to the top surface, the missing side surface being an open end of the sampling box (1); when in use, the open end is aligned with the root of the wheat seedling; the missing top surface is used to accommodate a box cover (21) of a box cover assembly (2); when in use, the box cover (21) closes the missing top surface; two side surfaces on both sides of the open end are A rotating shaft seat (12) is provided near the top of the surface, the rotating shaft seat (12) being a circular hole, the diameter of the circular hole being slightly larger than the outer diameter of the circular cutting rotating shaft (22), and the circular cutting rotating shaft (22) being able to rotate freely in the circular hole; the box cover rotating shaft insertion guide groove (13) being a rectangular opening, the rectangular opening being communicated with the rotating shaft seat (12), the circular cutting rotating shaft (22) being inserted into the rotating shaft seat (12) through the rectangular opening, and the size of the rectangular opening being smaller than the outer diameter of the circular cutting rotating shaft (22); The box cover assembly (2) comprises a box cover (21), a circular cutting shaft (22), a battery and controller box (23) integrated with the box cover, a flapping plate guide column (24), a guide column limit retaining ring (25) and a flapping plate (26). Four flapping plate guide columns (24) are fixed below the box cover (21). A vibration element (28) is also fixed below the box cover (21) and is located at the center of the four flapping plate guide columns (24). The flapping plate (26) is fixed on the vibration element (28). The flapping plate (26) has flapping plate ribs (27) below. Circular holes are provided at four corners of the flapping plate (26). The four circular holes are penetrated on the four flapping plate guide columns (24) so ​​that the flapping plate can vibrate freely on the four flapping plate guide columns (24). The guide column limit retaining ring (25) is fixed at the bottom end of the four flapping plate guide columns (24) for limiting the flapping plate (26).

2. The wheat red spider mite detection and reporting sample collection and statistical system according to claim 1, It is characterized in that The battery and controller box (23) is fixed together with the box cover (21) and one side of the flapping plate (26). A circular cutting shaft (22) is arranged on the upper part of the battery and controller box (23). After the circular cutting shaft (22) is inserted into the shaft seat (12), when the box cover (21) and the flapping plate (26) are rotated to be parallel to the bottom surface of the box body (11) or perpendicular to the bottom surface of the box body (11), the battery and controller box (23) can block the open end of the sampling box (1).

3. The wheat red spider mite detection and reporting sample collection and statistical system according to claim 1, It is characterized in that The cross section of the battery and controller box (23) is in the shape of a 90-degree sector, and the open end is well sealed by using the circular arc of the outer periphery, and the circular arc of the battery and controller box (23) will not cause damage to the wheat plants during the rotation process.

4. The sample collection method of the sample collection and statistics system according to any one of claims 1 to 3, It is characterized in that The following steps are involved: A1. Place the open end of the sampling box (1) directly against the root of the wheat seedling (6) to be sampled; A2. Place the box cover assembly (2) with the vibrating element vertically, then insert the circular rotating shaft (22) on the side of the box cover into the guide groove (13) along the rotating shaft of the box cover and push it into the bottom of the groove of the sampling box (1) so that the rotating shaft reaches the rotating shaft seat (12); A3. Closing the box cover assembly (2). During the installation of the box cover assembly (2), the battery and the controller box of the box cover assembly (2) push the upright wheat seedlings to press against the bottom edge of the sampling box (1). During the closing process, the arc-shaped battery and the controller box ensure that the wheat seedlings (6) are nearly horizontal when the cover is closed. After the cover is closed, the flapping plate and the flapping plate ribs (27) at the bottom of the box cover assembly are in contact with the wheat seedlings (6); A4, turning on the vibration controller (3) to switch on the vibration element (28), the vibration element (28) drives the beating plate and the beating plate ribs to vibrate rapidly up and down along the beating plate guide column (24) to beat the wheat; A5. By vibrating and beating the wheat seedlings, the wheat spider mites fall into the bottom of the sampling box (1); A6. After the beating is completed, turn off the power supply and remove the box cover assembly (2) to complete the sample collection.

5. The identification and statistical method of the sample collection and statistical system according to any one of claims 1 to 3, It is characterized in that The following steps are involved: B1. Image collection: Collect red spider mites in the wild, and use the smartphone client to take photos of the red spider mites in the sample collection device; B2. Image preprocessing and graphic transformation: The server receives the image to be processed transmitted by the smartphone client, removes the background, grays it, removes noise, and binarizes it, then uses a sliding window to locate the target and marks the target to be detected with a box; (1) Input an image and use the lazy snapping method to remove the background and grayscale the image; Then use Gaussian filtering to binarize the image; (2) Use the selective search algorithm to extract and generate 2000 candidate regions from the binary image; (3) Dilate the edge of each candidate region by p pixels, where p = 16, and frame the candidate region after dilation; transform the size of each candidate region to 227 x 227; B3. Image recognition: The tensorflow framework is used to construct a parallelized AlexNet convolutional neural network model. Each candidate region is input into the AlexNet convolutional neural network to obtain a 4096-dimensional feature vector to determine whether it is a spider or not, and the boxes of spiders are left for the next step. The AlexNet convolutional neural network has two input ports. The input data is processed by the convolution layer conv, the activation function relu, the pooling layer pool and the batch normalization BN. The output data is cascaded merged by the parallel convolution layer conv and then connected by the fully connected layer fc to obtain the forward propagation result logits. AlexNet convolutional neural network adopts transfer learning strategy, first pre-training on ImageNet, using convolutional neural network to fully learn shallow features, using SGD algorithm during training, initial learning rate of 0.001, batch_size of 128; then remove the last fully connected layer of pre-training, modify the number of categories, and users perform large-scale training on small-scale data sets. When user annotation data is insufficient, AlexNet convolutional neural network is fine-tuned to improve the ability to identify object types; B4. Import the feature vector derived from the AlexNet convolutional neural network into the SVM classifier to generate category scores; B5. Set the IoU indicator threshold, calculate the IoU indicator, use non-maximum suppression, remove overlapping boxes based on the box with the highest score, and obtain the candidate box with the highest score in this category; B6. Classification and statistics: Classification is to identify the two types of spiders and then count their numbers; B7. Result output: The classification and statistical results are transmitted back to the smart phone client through the server communication module, and the smart phone client stores the identification and statistical results.

Citation Information

Patent Citations

  • Wheat red spider forecasting sample collecting device

    CN215302552U