A method and system for intelligently identifying the life state and activity trajectory of a silkworm

Abstract

The application discloses a kind of silkworm life state and activity track intelligent identification method and identification system, the method includes: obtaining the growth video of silkworm, the silkworm is under real breeding environment and intensive breeding condition;The activity track of each silkworm in the preset time period is obtained from growth video;According to activity track, the behavior data of each silkworm is determined, and the behavior data includes the relative crawling distance of silkworm in the preset time period, head shaking frequency, body shaking frequency and body activity degree;The behavior data of each silkworm is assigned to behavior vector and compared with the standard value of healthy silkworm obtained in advance, to identify the life state of silkworm.The application mainly according to the abnormal behavior characteristics shown when silkworm is ill or poorly developed, for example: frenzied crawling, refuse mulberry leaf and not in time hibernate etc., using deep learning obtains and quantifies silkworm behavior, and then by contrast with standard value, the life state of silkworm is intelligently identified.

Description

Technical Field

[0001] This invention belongs to the field of behavior recognition technology, specifically, it relates to a method and system for intelligent recognition of the life state and activity trajectory of silkworms. Background Technology

[0002] Silkworms are intensively farmed insects, making them highly susceptible to pathogens. Silkworm diseases are typically highly contagious, and because silkworms have short lifespans, diseased silkworms are difficult to cure with medication. Often, they are simply fed directly or fail to spin cocoons, resulting in significant annual losses of cocoons. Furthermore, improper rearing methods can easily lead to stunted silkworm development, promoting disease growth and affecting cocoon quality.

[0003] Existing literature uses deep learning to identify silkworm diseases, mainly based on the morphological characteristics of diseased silkworms. This requires data annotation on the basis of manual identification, which cannot achieve early identification of diseases and is not effective under intensive breeding conditions. Summary of the Invention

[0004] To address the problem of the inability to effectively identify silkworm diseases in existing silkworm farming practices, this invention provides an intelligent identification method and system for silkworm life status and activity trajectory. This system is based on the abnormal behavioral characteristics exhibited by silkworms when they are diseased or under other abnormal conditions. By extracting the corresponding behavioral data and analyzing the behavioral data, the life status of silkworms can be identified.

[0005] To achieve the above-mentioned technical objectives, the technical solution adopted by the present invention is as follows: Firstly, this application provides a method for intelligent recognition of the life state and activity trajectory of silkworms, including: Obtain videos of silkworm growth, where the silkworms are in a real breeding environment and under intensive breeding conditions; The activity trajectory of each silkworm within a preset time period is obtained from the growth video; The behavioral data of each silkworm is determined based on its activity trajectory. The behavioral data includes the relative crawling distance, head shaking frequency, body shaking frequency, and body activity level of the silkworm within the preset time period. The behavioral data of each silkworm is assigned to a behavioral vector and compared with the pre-obtained standard values ​​for healthy silkworms to identify the life status of the silkworms.

[0006] Preferably, the step of obtaining the activity trajectory of each silkworm within a preset time period from the growth video specifically includes: The growth video is converted into frame images according to the time series and then preprocessed. Starting from the first frame image, key point detection is performed on each silkworm according to the time sequence to obtain the key point coordinates of each silkworm. The key points include five key points: the head, tail, and three markings on the back of the silkworm. Based on the key point coordinates of each silkworm obtained from the time series, the activity trajectory of each silkworm at each key point can be obtained.

[0007] Preferably, the step of performing keypoint detection on each silkworm according to a time sequence, starting from the first frame image, to obtain the keypoint coordinates of each silkworm, specifically includes: Each silkworm is numbered from the first frame of the image; In each frame of the image, feature extraction is performed on the numbered silkworms to obtain the coordinates of key points.

[0008] Preferably, the step of extracting key point coordinates from the numbered silkworms in each frame of the image specifically includes: Use a convolutional neural network to perform feature calculations on each frame of silkworm images to obtain the probability of each pixel belonging to each key point of each silkworm. The pixel with the highest probability value is selected as the key point coordinate of the silkworm.

[0009] Preferably, the step of extracting key point coordinates from the numbered silkworms in each frame of the image specifically includes: extracting features from each frame of the silkworm image using a sequence model to obtain the key point coordinates for each type of key point.

[0010] Preferably, the step of obtaining the key point coordinates of each silkworm based on the time series, thereby obtaining the activity trajectory of each silkworm at each key point, specifically includes: Connect each key point of a single silkworm in the first frame image with a line. Select the nearest adjacent silkworm in the next frame and match each of its key points with each key point of the previous frame, until the last frame. By connecting the key point coordinates of each silkworm from the first frame to the last frame, the activity trajectory of each key point of the silkworm can be obtained.

[0011] Preferably, determining the behavioral data of each silkworm based on its activity trajectory specifically includes: The relative crawling distance of the silkworm within a preset time period is obtained based on the behavioral vectors of two key points: the head and the tail. The frequency of head shaking of silkworms within a preset time period is obtained by using the behavioral vectors of three key points, namely two spots on the head and back of the silkworm. The frequency of the silkworm's body swaying within a preset time period was obtained based on the behavioral vectors of three key points: one of the markings on the silkworm's head, tail, and back; and The level of physical activity of silkworms within a preset time period is obtained by analyzing the behavioral vectors of the silkworm's head, tail, back, and two key points of its markings.

[0012] Preferably, the pre-obtained healthy silkworm behavior data includes: the behavior data of healthy silkworms at each age within a preset period.

[0013] Secondly, this application also provides an intelligent recognition system for the life status and activity trajectory of silkworms, including: The image acquisition module is used to acquire videos of the growth of silkworms, which are in a real breeding environment and under intensive breeding conditions; The activity trajectory acquisition module is used to acquire the activity trajectory of each silkworm within a preset time period from the growth video; The behavior data acquisition module determines the behavior data of each silkworm based on its activity trajectory. The behavior data includes the relative crawling distance, head shaking frequency, body shaking frequency, and body activity level of the silkworm within the preset time period. The identification module assigns the behavioral data of each silkworm to a behavioral vector and compares it with the pre-obtained standard values ​​for healthy silkworms to identify the life status of the silkworms.

[0014] Thirdly, this application also provides an intelligent recognition system for the life status and activity trajectory of silkworms, including a camera device, a processor, and a memory. The processor is interconnected with the camera device and the memory, respectively. The camera device is used to collect growth videos of silkworms, and the memory is used to store a computer program. The computer program includes program instructions, and the processor is configured to call the program instructions to execute the method as described in any one of claims 1 to 8.

[0015] Compared with the prior art, the present invention has the following advantages: Based on the different behavioral characteristics that silkworms exhibit when they are sick or in abnormal condition, we can use deep learning to quantify silkworm behavior to carry out intelligent identification of their life status. This eliminates the need for manual data labeling, making the identification more objective and accurate, and enabling early identification of diseases. Attached Figure Description

[0016] Figure 1 This is a flowchart of an intelligent recognition method for the life status and activity trajectory of a silkworm according to an embodiment of the present invention; Figure 2 A flowchart for obtaining the activity trajectory of each silkworm within a preset time period from growth videos; Figure 3 Flowchart of the method for obtaining key point coordinates for silkworms; Figure 4 A flowchart for obtaining the activity trajectory of each silkworm at each key point based on the coordinates of key points; Figure 5 A schematic diagram of an embodiment of an intelligent recognition system for the life status and activity trajectory of silkworms; Figure 6 This is a schematic diagram of another embodiment of an intelligent recognition system for the life status and activity trajectory of silkworms. Detailed Implementation

[0017] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to embodiments and accompanying drawings. The content mentioned in the embodiments is not intended to limit the present invention.

[0018] This application provides an intelligent method for recognizing the life status and activity trajectory of silkworms in a real-world farming environment, including identifying disease and stunted growth. The recognition principle is based on the abnormal behavioral characteristics exhibited by sick or stunted silkworms, such as frantic crawling, refusal to eat mulberry leaves, and failure to molt promptly. Furthermore, considering that silkworms in actual farming environments are neither constantly obscured by mulberry leaves nor constantly overlapping, the method utilizes a corresponding algorithm to obtain quantified behavioral characteristics of the silkworms. These characteristics are then compared with pre-established standard values, which are pre-acquired behavioral data of healthy silkworms. The life status of the silkworms is identified based on the comparison results.

[0019] like Figure 1 As shown, the intelligent identification method for the life status of silkworms specifically includes the following steps: S100: Obtain a video of silkworm growth, wherein the silkworms are in a real breeding environment and under intensive breeding conditions.

[0020] Due to the high density of silkworms, high-definition cameras were used to capture their growth videos. The cameras were positioned vertically above the silkworm trays, at a fixed height to establish identification standards. The recording time was approximately one hour to avoid capturing images immediately after feeding when the silkworms are still diseased and obstructed by mulberry leaves. Therefore, the optimal recording time was before feeding, after the mulberry leaves in the trays had been eaten, ensuring the silkworms were not obscured by leaves, which is beneficial for subsequent image analysis.

[0021] S200: Obtain the activity trajectory of each silkworm within a preset time period from the growth video.

[0022] The preset time period can be the duration of the growth video capture, such as 1 hour, or it can be a specific segment of the growth video.

[0023] The activity trajectory is the path that silkworms follow in a growth video or within a selected time period, used to determine the activity status of silkworms.

[0024] The specific steps are as follows: Figure 2 As shown: S210: Convert the grown video into frame images according to the time series and preprocess them.

[0025] After the silkworm growth video is collected, image processing methods can be used to convert the video into frame images. When the video is short, all images can be saved; when the video is long, one frame can be captured every second, and then the frame images can be saved in time sequence.

[0026] Preprocessing of the converted frame image includes necessary size segmentation and brightness transformation.

[0027] S220: Starting from the first frame image, key point detection is performed on each silkworm according to the time sequence to obtain the key point coordinates of each silkworm. The key points include three spots on the head, tail and back of the silkworm.

[0028] This step mainly involves key point detection of the silkworm in the frame image to obtain the coordinates of each key point of the silkworm in each frame image. These key point coordinates can be used to determine the movement process of the silkworm, thereby obtaining the corresponding behavior data of the silkworm.

[0029] The specific selection of key points is related to the analysis of behavioral data. For example, to analyze the relative crawling distance of silkworms, it is necessary to obtain key point information of the silkworm's head and tail. In this embodiment, a total of five key points of silkworms were selected, namely the head key point, the tail key point, and the three pattern key points on the back.

[0030] The method for obtaining the key point coordinates of silkworms is shown below: S221: Number each silkworm from the first frame image. The numbering rules are not unique, but the basic requirement is to ensure that each silkworm corresponds to a unique number. At the same time, a single frame image will contain multiple silkworms, with the density consistent with the actual breeding environment.

[0031] S222: In each frame of the image, feature extraction is performed on the numbered silkworms to obtain the coordinates of key points.

[0032] Feature extraction is performed on each silkworm in each frame of the image. The feature extraction method can be a convolutional neural network or a sequence model. The feature extraction process also includes feature operation, feature fusion, attention mechanism, residual connection, deconvolution and other methods commonly used in deep learning.

[0033] When a convolutional neural network is used for feature extraction, the network outputs a heatmap of the original image for the extracted features. This heatmap represents the probability that each pixel in the original image belongs to each key point of each silkworm. The pixel with the highest probability value is taken as the individual key point of the silkworm.

[0034] When the feature extraction network is a sequence model, the network output is an array of coordinate points for each class of key points, that is, the key point coordinates are directly output.

[0035] The training process for a keypoint detection network should also include the standard procedures for training deep learning models, such as calculating the loss between the network's predicted values ​​and the actual values, updating parameters, and performing iterative operations.

[0036] In this step, regardless of whether the algorithm uses a convolutional neural network or a sequence model, it needs to be trained using sample images. In the corresponding algorithm, the input is a frame image, and the output is the coordinates of key points. In addition to the algorithms provided above, existing classic algorithms, such as Hourglass, Cascaded, and HRNet, can be used for structural fine-tuning and structural modifications based on the silkworm dataset for training silkworm key point detection models.

[0037] During the training process of the model for detecting key points of silkworms in frame images, it is necessary to continuously verify and adjust the corresponding parameters to achieve the ideal detection accuracy. The key point detection model is then tested on different datasets, and the method is optimized and improved based on the test results to achieve an average detection accuracy of over 95% for silkworm key points.

[0038] Then, a keypoint detection model is used to detect keypoints in each frame of the image, obtain the keypoint coordinates of each silkworm, and match these keypoints.

[0039] S230: Based on the key point coordinates of each silkworm obtained from the time series, the activity trajectory of each key point of each silkworm is obtained.

[0040] This step mainly estimates the movement trajectory of the silkworm's five key points by using the coordinates of five key points of the same silkworm number in each frame image. The specific steps are shown in Figure 4: S231: Connect each key point of a single silkworm into a line on the first frame image.

[0041] The key points of the silkworm captured in the first frame of the image are used as the starting point of the activity trajectory. Connecting the five key points of the corresponding silkworm with a line helps to determine the silkworm that the line corresponds to.

[0042] S232: Select the nearest adjacent silkworm in the next frame image and match each of its key points with each key point of the previous frame until the last frame image.

[0043] Since the silkworm does not undergo sudden positional changes in adjacent frames, key point positional changes of the same silkworm can be matched through consecutive frame images.

[0044] S233: Connect the key point coordinates of each silkworm from the first frame to the last frame in sequence to obtain the activity trajectory of each key point of the silkworm.

[0045] Finally, based on the time-series frame images, the activity trajectory of each silkworm can be obtained at five key points. These activity trajectories are important data for judging the behavior of silkworms.

[0046] S300: Determine the behavioral data of each silkworm based on its activity trajectory. The behavioral data includes the relative crawling distance, head shaking frequency, body shaking frequency, and body activity level of the silkworm within the preset time period.

[0047] Based on the previously obtained key point activity trajectories of silkworms, the corresponding behavioral data can be obtained using the following method, the specific process of which is as follows: As previously known, the behavioral data of silkworms in this embodiment mainly considers the relative crawling distance, head shaking frequency, body shaking frequency, and body activity level of silkworms in the growth video. These parameters, after years of research, can reflect the behavioral data of silkworms to a certain extent.

[0048] Specifically, when assessing the relative crawling distance of silkworms within a preset time period, the activity trajectories of the two key points, the head and tail of the silkworms, are mainly considered.

[0049] When evaluating the frequency of head shaking in silkworms within a preset time period, the main consideration is the shaking of the head. Therefore, the activity trajectories of two key spots (three in total) on the head and back are selected as the evaluation criteria.

[0050] When assessing the frequency of body shaking of silkworms within a preset time period, the shaking of the head and tail is mainly considered. In order to select reference points, one of the pattern key points on the back can be added as a consideration in addition to the key points on the head and tail.

[0051] When assessing the physical activity level of silkworms within a preset time period, the activity trajectories of five key points can be taken into account.

[0052] S400: The behavioral data of each silkworm is assigned to a behavioral vector and compared with the pre-obtained standard value of a healthy silkworm to identify the life status of the silkworm.

[0053] Since behavioral data includes multiple parameter values, it can be grouped into a set of behavioral vectors and compared with the pre-obtained standard values ​​for healthy silkworms (standard values ​​containing multiple behavioral data parameters). Silkworms that exceed the standard values ​​within a certain range are identified as diseased or underdeveloped silkworms, and the identification results are output.

[0054] Specifically, the acquisition of the standard values ​​for judging healthy silkworm behavior data is similar to steps S100~S300 above, except that the standard values ​​need to be calculated and obtained in advance for healthy silkworms. The specific process can be as follows: (1) Collect videos of silkworm growth. The silkworms must be fed by professionals in a standard breeding environment. The silkworms must grow healthily and be free from disease or poor development. Ensure that the final behavioral data can be used as a standard value for judgment.

[0055] (2) After the acquisition is completed, the video is converted into frame images according to the time series, preprocessed, and then the features of the silkworm are extracted by the key point detection model. Thus, the activity trajectory is analyzed from the time series frame images, and finally the corresponding behavioral data is obtained from the activity trajectory. The parameters of the behavioral data are also calculated to obtain standard values ​​in step S300.

[0056] Analyzing the activity trajectory of a silkworm from time-series frame images can be done by referring to steps S231-S233. After keypoint detection on consecutive image frames, in the first frame, each silkworm is numbered, and the keypoints of each silkworm are connected to form a line. Next, in the next frame, based on the fact that the silkworm's position does not change abruptly, nor does it suddenly appear or disappear in the video, the silkworm closest to its neighbor in the previous frame is matched as the same silkworm, and each of its keypoints is matched against each other. This process of numbering and matching keypoints is repeated in the next frame until the last frame of the video data. Finally, the activity trajectory of the silkworm is obtained by printing and connecting the keypoint coordinates of each silkworm sequentially.

[0057] It should be noted that the discrimination criteria established in this embodiment include behavioral data for each day of the silkworm's first instar to the seventh day of the fifth instar (a total of 20 sets of values). The criteria are based on each day of the silkworm's instar, and the behavioral data of silkworms during the day and night are different. When the silkworms are small, a physical magnification method is used; that is, during the 1st and 2nd instars, an electron microscope is used to enlarge the silkworms for image acquisition, and the acquired silkworm growth video data is also magnified data. Furthermore, the discrimination criteria are established for different silkworm varieties and under various rearing environments. The larger the dataset used when training the keypoint detection model, the wider the applicability of the method.

[0058] Considering the physiological differences between different silkworm varieties, the above training was conducted for multiple varieties to broaden its applicability. For specific silkworm varieties, appropriate behavioral quantification methods should be adopted.

[0059] It should also be noted that in actual breeding processes, silkworms may be obscured by mulberry leaves, and they may overlap. To address this, our method involves manually marking the approximate location of invisible keypoints during the establishment of discrimination criteria. These keypoints are then marked as invisible for use in keypoint detection, further used to train the keypoint detection model and enhance its predictive ability under partial occlusion conditions. In extreme cases, if a silkworm is completely obscured, its data is not labeled, and it will not contribute to training.

[0060] After key point detection, key point matching is performed to obtain the silkworm's activity trajectory by observing the positional changes of key points on consecutive video frames.

[0061] Regarding the phenomenon of silkworms being blocked by mulberry leaves or other silkworms during the use of the method provided in this embodiment, since silkworms are insects with their own activities and have the physiological characteristic of climbing upwards, they will not be blocked all the time. This method is based on video datasets, and a single silkworm will not be blocked all the time in a video segment, but will be identified in continuous detection. Therefore, this invention can overcome the problem of silkworm occlusion in actual breeding environments.

[0062] like Figure 5 As shown, this embodiment also provides an intelligent recognition system for the life status and activity trajectory of silkworms, including an image acquisition module 51, an activity trajectory acquisition module 52, a behavior data acquisition module 53, and a recognition module 54.

[0063] The system includes an image acquisition module 51 for acquiring growth videos of silkworms in a real breeding environment and under intensive breeding conditions; an activity trajectory acquisition module 52 for acquiring the activity trajectory of each silkworm within a preset time period from the growth videos; a behavior data acquisition module 53 for determining the behavior data of each silkworm based on the activity trajectory, including the relative crawling distance, head shaking frequency, body shaking frequency, and body activity level of the silkworm within the preset time period; and an identification module 54 for assigning the behavior data of each silkworm to a behavior vector and comparing it with the pre-obtained standard value for healthy silkworms. When the behavior data of a silkworm differs significantly from the standard value, it can be identified as diseased or underdeveloped.

[0064] like Figure 6As shown, this embodiment also provides an intelligent recognition system for the life status and activity trajectory of silkworms, including a camera device 61, a processor 62, and a memory 63. The processor 62 is interconnected with the camera device 61 and the memory 63. The camera device 61 is used to collect growth videos of silkworms, and the memory 63 is used to store computer programs. The computer programs include program instructions, and the processor 62 is configured to call the program instructions to execute the above-described intelligent recognition method for the life status of silkworms.

[0065] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the foregoing technical solutions, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the various technical solutions of the present invention.

Claims

1. A method for intelligent recognition of the life state and activity trajectory of silkworms, characterized in that, include: Obtain videos of silkworm growth, where the silkworms are in a real breeding environment and under intensive breeding conditions; The process of obtaining the activity trajectory of each silkworm within a preset time period from the growth video includes: converting the growth video into frame images according to the time sequence and preprocessing them; starting from the first frame, performing keypoint detection on each silkworm according to the time sequence to obtain the keypoint coordinates of each silkworm, wherein the keypoints include five keypoints: the head, tail, and three markings on the back of the silkworm; based on the keypoint coordinates of each silkworm obtained from the time sequence, obtaining the activity trajectory of each keypoint of each silkworm, specifically including: connecting each keypoint of a single silkworm into a line on the first frame; selecting the nearest adjacent silkworm on the next frame and matching each of its keypoints with each keypoint of the previous frame, until the last frame; and connecting the keypoint coordinates of each silkworm from the first frame to the last frame sequentially to obtain the activity trajectory of each keypoint of the silkworm. The behavioral data of each silkworm is determined based on its activity trajectory. The behavioral data includes the relative crawling distance, head shaking frequency, body shaking frequency, and body activity level of the silkworm within the preset time period. The behavioral data of each silkworm is assigned to a behavioral vector and compared with the pre-obtained standard values ​​for healthy silkworms to identify the life status of the silkworms.

2. The intelligent recognition method for the life state and activity trajectory of silkworms according to claim 1, characterized in that, Starting from the first frame of the image, keypoint detection is performed on each silkworm according to a time sequence to obtain the keypoint coordinates of each silkworm. Specifically, this includes: Each silkworm is numbered from the first frame of the image; In each frame of the image, feature extraction is performed on the numbered silkworms to obtain the coordinates of key points.

3. The intelligent recognition method for the life state and activity trajectory of silkworms according to claim 2, characterized in that, The step of extracting key point coordinates from the numbered silkworms in each frame of the image specifically includes: Use a convolutional neural network to perform feature calculations on each frame of silkworm images to obtain the probability of each pixel belonging to each key point of each silkworm. The pixel with the highest probability value is selected as the key point coordinate of the silkworm.

4. The intelligent recognition method for the life status and activity trajectory of silkworms according to claim 3, characterized in that, The behavioral data for each silkworm, determined based on its activity trajectory, specifically includes: The relative crawling distance of the silkworm within a preset time period is obtained based on the behavioral vectors of two key points: the head and the tail. The frequency of head shaking of silkworms within a preset time period is obtained by using the behavioral vectors of three key points, namely two spots on the head and back of the silkworm. The frequency of the silkworm's body swaying within a preset time period was obtained based on the behavioral vectors of three key points: one of the markings on the silkworm's head, tail, and back; and The level of physical activity of silkworms within a preset time period is obtained by analyzing the behavioral vectors of the silkworm's head, tail, back, and two key points of its markings.

5. The intelligent recognition method for the life state and activity trajectory of silkworms according to claim 1, characterized in that, The pre-obtained healthy silkworm behavior data includes: the behavior data of healthy silkworms at each age within a preset period.

6. A smart recognition system for the life status and activity trajectory of silkworms, characterized in that, include: The image acquisition module is used to acquire videos of the growth of silkworms, which are in a real breeding environment and under intensive breeding conditions; The activity trajectory acquisition module is used to acquire the activity trajectory of each silkworm within a preset time period from the growth video. Specifically, it includes: converting the growth video into frame images according to the time sequence and preprocessing them; starting from the first frame image, performing key point detection on each silkworm according to the time sequence to obtain the key point coordinates of each silkworm, wherein the key points include five key points: the head, tail, and three markings on the back of the silkworm; based on the key point coordinates of each silkworm obtained from the time sequence, obtaining the activity trajectory of each key point of each silkworm, specifically including: connecting each key point of a single silkworm into a line on the first frame image; selecting the nearest adjacent silkworm on the next frame image and matching each of its key points with each key point of the previous frame, until the last frame image; connecting the key point coordinates of each silkworm from the first frame to the last frame sequentially to obtain the activity trajectory of each key point of the silkworm. The behavior data acquisition module determines the behavior data of each silkworm based on its activity trajectory. The behavior data includes the relative crawling distance, head shaking frequency, body shaking frequency, and body activity level of the silkworm within the preset time period. The identification module assigns the behavioral data of each silkworm to a behavioral vector and compares it with the pre-obtained standard values ​​for healthy silkworms to identify the life status of the silkworms.

7. A smart recognition system for the life status and activity trajectory of silkworms, characterized in that, The device includes a camera, a processor, and a memory, wherein the processor is interconnected with the camera and the memory, respectively. The camera is used to capture videos of the growth of silkworms, the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to invoke the program instructions to execute the method as described in any one of claims 1 to 5.

Images