CN117612155A - Method and system for litchi pedicel moth detection and eclosion rate calculation - Google Patents

Method and system for litchi pedicel moth detection and eclosion rate calculation Download PDF

Info

Publication number
CN117612155A
CN117612155A CN202311343299.6A CN202311343299A CN117612155A CN 117612155 A CN117612155 A CN 117612155A CN 202311343299 A CN202311343299 A CN 202311343299A CN 117612155 A CN117612155 A CN 117612155A
Authority
CN
China
Prior art keywords
moth
litchi
incubator
cocoons
adults
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202311343299.6A
Other languages
Chinese (zh)
Inventor
陆健强
袁家俊
杨继国
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Lingnan Modern Agricultural Science And Technology Guangdong Provincial Laboratory Heyuan Sub Center
South China Agricultural University
Original Assignee
Lingnan Modern Agricultural Science And Technology Guangdong Provincial Laboratory Heyuan Sub Center
South China Agricultural University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Lingnan Modern Agricultural Science And Technology Guangdong Provincial Laboratory Heyuan Sub Center, South China Agricultural University filed Critical Lingnan Modern Agricultural Science And Technology Guangdong Provincial Laboratory Heyuan Sub Center
Priority to CN202311343299.6A priority Critical patent/CN117612155A/en
Publication of CN117612155A publication Critical patent/CN117612155A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/68—Food, e.g. fruit or vegetables
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/09—Supervised learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Multimedia (AREA)
  • General Health & Medical Sciences (AREA)
  • Software Systems (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Molecular Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Measuring Or Testing Involving Enzymes Or Micro-Organisms (AREA)

Abstract

The invention relates to a litchi pedicle borer detection technology, in particular to a method and a system for litchi pedicle borer detection and eclosion rate calculation, wherein the method is realized based on an incubator assembly comprising an incubator body, a camera assembly, a lamp control assembly and an in-box environment regulation and control module; acquiring a picture containing the litchi Di-moth cocoons and the adults shot by the shooting assembly in real time, and identifying and recording the Di-moth cocoons and the adults in the picture by adopting an identification model to obtain the Di-moth cocoons and the Di-moth quantities; calculating the current emergence rate of the Di-moth according to the Di-moth cocoons and the quantity of the Di-moth; the total fruit moth pest severity of the orchard is estimated and the optimal timing of the insecticide is estimated by referring to the quantity of fruit borers contained in the falling fruit and the current emergence rate of the fruit borers. According to the method, the collected Di-moth cocoons and Di-moth adults generated by litchi falling fruits are accurately identified, the emergence rate of Di-moth is calculated to calculate the integral pest damage condition of an orchard, and objective basis is provided for pesticide application of fruit farmers.

Description

Method and system for litchi pedicel moth detection and eclosion rate calculation
Technical Field
The invention relates to a litchi pedicle moths detection technology, in particular to a method and a system for litchi pedicle moths detection and eclosion rate calculation.
Background
In recent years, with the rapid development of artificial intelligence technology, the identification of crop diseases and insect pests has obtained new methods and applications of ideas. The efficient image recognition technology is utilized to improve recognition efficiency, reduce cost and improve recognition precision, so that accuracy and feasibility of agricultural pest and disease recognition are continuously improved. Under the background, expert scholars at home and abroad conduct a great deal of research, and deep learning is used as an important point for exploration. The application of deep learning in crop pest identification can significantly reduce workload and shorten identification time. With the rapid development of machine learning, more and more researches prove that the deep convolutional neural network is excellent in plant science, and a plurality of successful application examples are also available in plant disease and pest detection.
Based on image processing and computer vision technology, researchers develop various image feature extraction, image segmentation and target detection algorithms for extracting feature information from pest images so as to realize automatic detection and classification of pests. The method comprises color features, texture features, shape features and the like, and different types of plant diseases and insect pests can be effectively identified by analyzing and comparing the features; the introduction of deep learning algorithm, especially the application of Convolutional Neural Network (CNN), brings great improvement to the identification of plant diseases and insect pests. Through large-scale dataset training, CNN can learn and extract the characteristic in the image automatically, carries out high-efficient classification and discernment, and this makes the pest and disease damage discernment can go on a larger scale and higher precision, provides more advanced, high-efficient and accurate pest and disease damage discernment solution for agricultural production, helps realizing healthy management and the increase in production efficiency of crops.
Litchi pedicel moth is one of the main pests which harm litchi, and the main control method at present is pesticide spraying. The method mainly comprises the steps of collecting litchi fruits from the field, observing the cocoon formation quantity and the adult quantity of the litchi borers, measuring the eclosion rate of the litchi borers, and predicting the best pesticide spraying time.
Because the litchi pedicel moth cocoons are smaller and transparent, cocooning positions are often positioned in the included angle gaps and are not easy to distinguish from the background; the adult Di-moth has small size, similar color to oxidized fruits and is easy to fly to change positions, so that the difficulty of manually counting Di-moth cocoons and Di-moth adults is high, the efficiency is low, the subjectivity is high, the statistics is easy to be wrong, the correct time for spraying the pesticide cannot be accurately judged, the litchi Di-moth can not be effectively killed, and the pesticide is easy to be excessively used, so that the problems of agricultural waste, environmental pollution, fruit pollution and the like are caused.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides a method for detecting the litchi pedicle borers and calculating the emergence rate, which simulates the field environment through an incubator assembly, observes and shoots pictures of different growth periods, automatically identifies the litchi pedicle borers cocoons and adults in the pictures through an improved identification model, calculates the emergence rate, calculates the optimal opportunity of pesticide spraying, breaks the bottleneck that the traditional manual statistics of the quantity and the efficiency of the litchi pedicle borers is low and the error rate is high, and solves the problems that agricultural materials are wasted, the environment and the fruit are polluted and the like easily when pesticide spraying is performed through subjective experience.
It is another object of the present invention to provide a system for litchi pedicel moth detection and eclosion rate calculation to improve the efficiency and accuracy of pedicel moth eclosion rate detection.
The method is realized by the following technical scheme: a method for litchi pedicel moth detection and eclosion rate calculation, the method being implemented based on an incubator assembly; the incubator assembly comprises an incubator body, a camera assembly, a lamp control assembly and an in-box environment regulation and control module, wherein the camera assembly, the lamp control assembly and the in-box environment regulation and control module are positioned in the incubator body; the lamp control assembly is used for changing the light environment in the incubator and providing the same light environment as that in the field for the Di-moths in the incubator; the in-box environment regulation and control module monitors the temperature and humidity in the box, compares the temperature and humidity with the field temperature and humidity data in real time, and regulates the temperature and humidity in the box to be similar to the field temperature and humidity;
the method comprises the following steps:
s1, collecting a batch of fallen fruits from a litchi orchard, and placing the fallen fruits into an incubator;
s2, starting a lamp control assembly and an in-box environment regulation and control module, and performing simulation comparison on the lamplight environment, the temperature and humidity environment and the field environment in the incubator so that the environment in the incubator is suitable for normal growth activities of Di-moths;
s3, acquiring a picture containing the litchi Di moth cocoons and the adults shot by the shooting assembly in real time, and identifying and recording the Di moth cocoons and the adults in the picture by adopting an identification model to obtain the Di moth cocoons and the Di moth quantities;
s4, calculating the current emergence rate of the Di-moths according to the Di-moths cocoons and the quantity of the Di-moths;
s5, estimating the severity of the total fruit moth insect pests of the orchard and estimating the optimal timing of insecticide spraying according to the quantity of the fruit moth contained in the fallen fruit and the current emergence rate of the fruit moth.
Preferably, the recognition model is used for constructing a training data set by using pictures containing the Di-moth cocoons and adults under different lamplight backgrounds according to the characteristics of small body size, similar color to oxidized fruits and transparent Di-moth cocoons, constructing a feature extraction network, extracting the texture, size, shape and color features of the Di-moth cocoons and adults by convolution operation, and finally realizing the recognition and positioning of the Di-moth cocoons and adults in the pictures by establishing a multi-layer neural network model.
Preferably, the recognition model performs feature fusion in a channel dimension and a space dimension such that channel attention is focused on color, texture and shape features of the litchi Di-moth cocoons and adults, and space attention is focused on position information of the litchi Di-moth cocoons and adults.
Preferably, step S3 includes:
s31, shooting the growth condition of the litchi pedicle moths in the incubator at regular time through a camera shooting assembly and a lamp control assembly in the incubator, and obtaining picture samples of the litchi pedicle moths under different light effects; labeling the litchi pedicel borer cocoons and adults in the pictures, constructing a data set, and dividing the data set into a training set and a test;
s32, constructing a recognition model of the litchi Di moth cocoons and adults;
carrying out attention feature fusion on the channel dimension and the space dimension of the recognition model, so that the attention of the channel is focused on the color, texture and shape features of the litchi Di-moth cocoons and adults, and the space attention is focused on the position information of the litchi Di-moth cocoons and adults; and a bidirectional information flow structure from top to bottom and from bottom to top is added in the improved model, so that the transmission of the characteristic information of the Di-moth between different network layers is enhanced.
Further preferably, in the recognition model constructed in step S32, a 4×4 target detection layer is employed; and constructing a combined cascade feature fusion structure and a parallel feature structure to promote fusion of low-level image features of the Di-moths and high-level image features of the Di-moths, wherein the low-level image features comprise pixel values, colors and edge features, and the high-level image features comprise position, shape and relation features.
The system of the invention is realized by the following technical scheme: a system for litchi pedicle borer detection and eclosion rate calculation comprises an incubator assembly and an identification model;
the incubator assembly comprises an incubator body, a camera assembly, a lamp control assembly and an in-box environment regulation and control module, wherein the camera assembly, the lamp control assembly and the in-box environment regulation and control module are positioned in the incubator body; the lamp control assembly is used for changing the light environment in the incubator and providing the same light environment as that in the field for the Di-moths in the incubator; the in-box environment regulation and control module monitors the temperature and humidity in the box, compares the temperature and humidity with the field temperature and humidity data in real time, and regulates the temperature and humidity in the box to be similar to the field temperature and humidity;
the lamplight environment, the temperature and humidity environment and the field environment in the incubator are simulated and compared through the lamp control assembly and the in-box environment regulation and control module, so that the environment in the incubator is suitable for the normal growth activity of the Di-moths;
acquiring a picture containing the litchi Di-moth cocoons and the adults shot by the shooting assembly in real time, and identifying and recording the Di-moth cocoons and the adults in the picture by adopting an identification model to obtain the Di-moth cocoons and the Di-moth quantities; calculating the current emergence rate of the Di-moth according to the Di-moth cocoons and the quantity of the Di-moth; the total fruit moth pest severity of the orchard is estimated and the optimal timing of the insecticide is estimated by referring to the quantity of fruit borers contained in the falling fruit and the current emergence rate of the fruit borers.
Preferably, the incubator body comprises an outer frame, an inner frame, a continuous slope structure, a gentle slope truncated cone structure, a camera hole and a frame connecting piece, wherein the outer frame and the inner frame are connected through the frame connecting piece; the camera shooting assembly is fixed at the top end of the internal framework, the camera shooting hole is positioned above the internal framework, and the lamp control assembly is embedded above the external framework of the incubator body; the continuous slope-shaped structure is arranged at the bottom of the incubator body, and the included angle between slopes provides cocoon setting angle conditions for the litchi pedicel borers; the gentle slope round platform structure is positioned in the middle of the bottom surface of the incubator body and is higher than the continuous slope structure and used for filling the observation dead angle at the joint of the gentle slope round platform structure and the continuous slope structure.
Compared with the prior art, the invention has the following beneficial effects:
1. the incubator assembly provided by the invention can collect the cocoons and adults of the litchi pedicle moths in a batch of fallen litchi fruits in the incubator, and simulate the field environment for the litchi pedicle moths, wherein the environment comprises an angle condition suitable for the cocoons of the litchi pedicle moths, a temperature and humidity environment suitable for the normal activities of the litchi pedicle moths, a day and night light condition of a normal field and the like, so that the acquired data is more scientific.
2. The litchi pedicle borer detection system can automatically acquire pictures containing litchi pedicle borer cocoons and litchi pedicle borer adults under different background lights, can count the numbers of the litchi pedicle borer cocoons and the litchi pedicle borer adults in the pictures in real time, automatically counts the eclosion rate of the litchi pedicle borers in the box on the same day every day, further calculates the optimal pesticide spraying time, and has higher efficiency and accuracy compared with the traditional manual statistics of the litchi pedicle borers and the manual calculation eclosion rate, thereby saving time and manpower; compared with fruit growers which rely on self experience to infer the pesticide spraying time, the method has more scientific accuracy.
3. According to the litchi Di moth identification improved model, identification of Di moth cocoons and Di moth adults in pictures is realized by constructing a characteristic extraction network and a multilayer neural network; the main body framework, the neck part, the head part and the like of the model are improved, so that the small target semantic information extraction capability and the small target detection performance of the model are enhanced, and the model is more suitable for identifying the litchi pedicle borers.
Drawings
FIG. 1 is a top view of the internal housing of the incubator of the present invention;
FIG. 2 is a schematic diagram of an incubator of the present invention;
FIG. 3 is a front view of the incubator of the present invention;
FIG. 4 is a side view of the incubator of the present invention;
FIG. 5 is a flowchart of a method for litchi Di moth detection and feathering rate calculation in an embodiment of the invention;
FIG. 6 is a diagram of a modified model for litchi pedicel moth identification in accordance with the present invention;
FIG. 7 is a diagram illustrating examples of picture samples used in the present invention, wherein the diagrams (a), (b), (c), and (d) are respectively picture samples under different light effects;
FIG. 8 is a schematic diagram of the improved model channel dimension and spatial dimension feature fusion of the present invention;
FIG. 9 is a schematic diagram of a bidirectional information flow structure in the improved model of the present invention;
FIG. 10 is a schematic diagram of a portion of a loss function in a modified model of the present invention;
the reference numerals in the drawings are: 11 camera holes, 21 lamp control components, 22 temperature and humidity monitoring modules, 23 continuous slope structures, 24 gentle slope round platform structures and 31 isolating gauze.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Example 1
The present embodiment provides a method for pedicle screw detection and calculation of emergence rate, the method being implemented based on an incubator assembly. The incubator assembly has a structure shown in fig. 1-4, and comprises an incubator body, a camera shooting assembly, a lamp control assembly 21 and an in-box environment regulation and control module. The incubator body is made of resin materials, so that the stability is good.
In this embodiment, the incubator body comprises an outer frame, an inner frame, a continuous slope structure 23, a gentle slope truncated cone structure 24, an isolation gauze 31, a camera hole 11 and frame connectors; the incubator box takes a triangular prism framework as an external framework, takes a triangular pyramid framework as an internal framework, takes a thin cylinder with a camera shooting hole as the top of the internal framework, and connects the external framework and the internal framework through framework connecting pieces. The camera shooting assembly is fixed at the top end of the internal framework, and the camera shooting assembly in a working state can automatically carry out full-coverage shooting on the box body at each set time interval according to the set time and store the full-coverage shooting. Referring to fig. 1, the camera hole 11 is located above the center of the internal framework of the incubator, and the size of the camera hole is the same as the size of the camera assembly, so that the camera can pass through the camera hole to shoot the whole bottom of the incubator without dead angles.
The light control assembly 21 is embedded above the outer frame of the incubator body, is provided with an LED lamp strip embedded at the upper end of the outer frame of the incubator body, and is used for changing the light environment in the incubator, different light effects can be shown according to a set time period, white light is irradiated in a daytime period to simulate the daytime, all light simulation black nights are closed in a black night period, and the same light environment as that in the field is provided for the borers in the incubator; when working in coordination with the camera assembly, the lamp control assembly can also provide different light effects for the camera assembly to ensure the shooting quality of the borer cocoons and the borer adults.
The in-box environment regulation and control module comprises a temperature and humidity regulation and control module (not shown) and a temperature and humidity monitoring module 22, and the temperature and humidity in the box can be monitored by the temperature and humidity monitoring module in a working state and compared with temperature and humidity data acquired from a field sensor in real time; when the difference between the environment in the box and the external environment is larger than the set threshold, the temperature and humidity regulation module in the box sends out alarm prompt sound, and simultaneously starts to work, so that the temperature and humidity in the box are regulated to be close to the temperature and humidity in the field, and the indoor environment is suitable for normal growth activities of the Di-moths. Referring to fig. 2, the temperature and humidity monitoring module 22 is located at the bottom side of the incubator body, when the system is used normally, the temperature and humidity monitoring module 22 is used for monitoring the temperature and humidity conditions inside the incubator body and transmitting the temperature and humidity conditions to the terminal, and when the temperature and humidity conditions inside the incubator body are greatly different from the field temperature and humidity environments, the module will send out an alarm prompt sound.
The continuous slope-shaped structure 23 is arranged at the bottom of the incubator body, the included angle between slopes is 120 degrees, and a proper cocoon forming angle condition is provided for the litchi pedicel borers, so that most cocoons are concentrated on the bottom surface of the incubator body, and observation is facilitated.
The gentle slope round platform structure 24 is positioned in the middle of the bottom surface of the box body and slightly higher than the continuous slope structure, and the gentle slope is designed for filling the observation dead angle at the joint of the gentle slope round platform structure and the continuous slope structure. Specifically, the gentle slope round platform structure 24 is located at the right center of the bottom surface of the incubator body and is used for placing litchi fruits, and the height of the gentle slope round platform structure is slightly higher than that of the continuous slope structure 23, so as to avoid the situation that litchi shields the cocoons of thecaryopsis; the gentle slope round platform structure aims at avoiding the condition that a small included angle gap exists at the joint of the round platform and the continuous slope structure so as to generate a dead angle for shooting.
Referring to fig. 3 and 4, the isolating gauze 31 is located on the side of the inner box body of the incubator, and covers the inner frame of the incubator box body, that is, the triangular pyramid structure, so as to seal the incubator and prevent the litchi pedicel moth from climbing out of the incubator, thus causing inaccurate data. The isolation gauze is made of transparent and breathable materials, so that the condition inside the incubator body can be conveniently observed from the outside, and the phenomenon that the inside of the incubator is too sultry and the normal activity of the litchi pedicel borers is influenced is prevented. In this embodiment, the spacer screen is sewn from three-sided translucent breathable material for confining the pedicel moth in the box and for ease of viewing. The isolation gauze is also provided with an invisible zipper, so that the litchi falling fruits can be conveniently placed and taken, and the joint of the bottom surface and the box body is provided with an adhesive tape, so that the position of the gauze can be conveniently adjusted, and the gauze is free of wrinkles.
As shown in fig. 5, the method for detecting the pedicel moth and calculating the emergence rate according to the present embodiment specifically includes the following steps:
s1, collecting a batch of fallen fruits from a litchi orchard, and after 80-100 litchi fallen fruits are placed on a gentle slope circular table structure in an incubator, confirming whether the isolation gauze is completely sealed.
S2, starting the lamp control assembly and the in-box environment regulation and control module, and performing simulation comparison on the lamplight environment, the temperature and humidity environment and the field environment in the box so that the environment in the incubator is suitable for normal growth activities of the Di-moth.
S3, acquiring a picture containing the litchi and the Thiotis borer cocoons and the Thiotis borer adults shot by the shooting assembly in real time, accurately identifying and recording the Thiotis borer cocoons and the Thiotis borer adults in the picture by adopting an improved identification model, and obtaining the number of the Thiotis borer cocoons and the Thiotis borer adults.
The improved recognition model takes pictures, which are shot by the shooting assembly, of the box body and contain the worm cocoons and the adults of the worm, as training samples, captures and learns textures, shapes and color characteristics of the worm cocoons and the adults, and finally accurately recognizes and counts the worm cocoons and the adults of the worm in real time.
In the embodiment, the improved recognition model is used for constructing a training data set by using pictures containing the Di-moth cocoons and the adults under different lamplight backgrounds aiming at the characteristics of small Di-moth adults, similar colors to oxidized fruits and transparent Di-moth cocoons, constructing a feature extraction network, extracting the characteristics of textures, sizes, shapes, colors and the like of the Di-moth cocoons and the adults by convolution operation, and finally realizing the recognition and positioning of the Di-moth cocoons and the adults in the pictures by establishing a multi-layer neural network model.
In order to improve the attention degree of the self characteristic information and the position characteristic information of the litchi Di-moth cocoons and the litchi Di-moth adults, the improved recognition model performs characteristic fusion in the channel dimension and the space dimension, so that the channel attention is focused on the color, texture and shape characteristics of the litchi Di-moth cocoons and the adults, and the space attention is focused on the position information of the litchi Di-moth cocoons and the adults.
As shown in fig. 6, the improved recognition model accurately recognizes the borer cocoons and adults, and mainly uses the following steps and algorithm flow:
s31, referring to FIG. 7, through the camera module and the lamp control module in the incubator, the growth condition of the litchi pedicel borer in the incubator is shot regularly, the picture sample of the litchi pedicel borer is obtained, and the sub-pictures (a), (b), (c) and (d) in FIG. 7 are respectively the picture samples shot under different light effects.
Labeling the litchi Di moth cocoons and adults in the pictures by using LabelImg, constructing a data set, and dividing the data set into a training set and a testing set by a Python script, wherein the proportion is 8:2.
after the data set is marked and divided, the model is prevented from being trained and fitted through image enhancement operations such as overturning, scaling, blurring and the like on the training set.
S32, constructing an improved identification model of the litchi Di moth cocoons and adults.
The improved recognition model comprises a backbone network, a neck network and a detection layer which are connected in sequence. The improvement on the original model in this particular example includes:
s321, in order to improve the attention degree of the model to the self characteristic information and the position characteristic information of the litchi Di-moth cocoons and the litchi Di-moth adults, attention characteristic fusion is carried out on the channel dimension and the space dimension of the model, so that the channel attention is focused on the color, texture and shape characteristics of the litchi Di-moth cocoons and the adults, and the space attention is focused on the position information of the litchi Di-moth cocoons and the adults.
Specifically, referring to fig. 8, the main process of attention feature fusion is to perform fusion processing on the input feature map in channel dimension and space dimension, namely, for the input feature map F e R C×H×W Generating one-dimensional channel attention map M using a channel attention module C ，M C ∈R C×1×1 Generating a two-dimensional spatial attention pattern M using a spatial attention module S ，M S ∈R 1×H×W Therefore, the content and the position information of the litchi pedicel borers in the graph are focused on by the model, and more details are acquired.
S322, aiming at the characteristic of small sizes of the cocoons and adults of the litchi Di-moth, replacing a large target detection layer 32 multiplied by 32 in an original model with a small target detection layer 4 multiplied by 4 which is more suitable for the Di-moth detection, so that the model can identify the Di-moth more accurately; and constructing a combined cascade characteristic fusion structure and a parallel characteristic structure, promoting the fusion of low-level image characteristics (such as pixel values, colors, edges and the like) of the Di-moths and high-level image characteristics (such as position, shape, relationship and the like) of the Di-moths, and reducing the loss of the characteristic information of the Di-moths.
S323, because the Di-moths in the incubator are different in distance from the camera, the Di-moths with different sizes can appear in the picture, in order to improve the detection performance of the model, the characteristic information transmission of the Di-moths between different network layers is enhanced by adding a two-way information flow structure from top to bottom and from bottom to top, and the capability of the model to solve the identification problems of inconsistent sizes, inconsistent colors and the like of the litchi Di-moths in the picture is improved, as shown in fig. 9.
Specifically, the bidirectional information flow structure can capture effective information from a plurality of feature graphs with different scales, and integrate the effective information into a unified feature set, and the bidirectional information flow from top to bottom and from bottom to top is helpful for capturing context information between features, so that the model can better understand the relationship between the Di-moth and the environment in the box, and the accuracy of target detection and segmentation tasks is improved.
S324, because the identification results of the litchi Di moth cocoons and the litchi Di moth adults are required to be obtained in real time, the reasoning speed of the model is required to be improved, the network parameter quantity is reduced by realizing the light weight of the model, and the network reasoning speed of the model is further improved.
Specifically, a plurality of original input channels are linearly combined to form a new virtual channel, so that the number of the input channels is far smaller than that of the original input channels; and meanwhile, the weight parameters of all the original input channels are shared, independent parameters are not allocated to each original input channel, the number of parameters of the model is reduced, the weight of the model is reduced, and the network reasoning speed of the model is increased.
S325, because the positions of the litchi pedicle moths in the incubator are not fixed and have different orientations, the vector angle between the real frame and the predicted frame of the detection frame needs to be further considered, so that the calculation of the vector angle between the real frame and the predicted frame is increased, and the speed of model training and the accuracy of reasoning are improved.
Specifically, the loss function of the improved model considers vector angles between the required regressions, further defines penalty indicators, redefines the associated loss functions, including distance loss functions, shape loss functions, IOU loss functions, as shown in FIG. 10, as compared to the loss functions of the original model.
The loss function calculation formula is as follows:
wherein Δ is a distance loss function value, Ω is a shape loss value, ioU is a IoU loss function value
And S326, finally, identifying the litchi Di moth cocoons and the adults contained in the input pictures by using the improved model.
To evaluate the superiority of the improved model in the recognition of Di-moths, the original model Yolov5, the classical model Faster R-CNN, the mainstream models Yolov7, yolov8 and the improved model of the invention were tested under the same data set and the results are shown in Table 1. As shown in table 1, this example is superior to other models that participated in the comparison in both the recognition accuracy of the litchi Di moth cocoons and adults.
Table 1 comparison of performance of improved models with other test models
S4, calculating the current emergence rate of the Di-moths according to the Di-moths cocoons and the quantity of the Di-moths.
S5, estimating the severity of the total fruit moth insect pests of the orchard and calculating the optimal timing of insecticide spraying by referring to the quantity of the fruit moth contained in the batch of fallen fruits and the current emergence rate of the fruit moth.
Example 2
Based on the same inventive concept as in example 1, this example provides a system for pedicle screw detection and emergence rate calculation, comprising an incubator assembly and an identification model.
The incubator assembly comprises an incubator body, a camera assembly, a lamp control assembly and an in-box environment regulation and control module, wherein the camera assembly, the lamp control assembly and the in-box environment regulation and control module are positioned in the incubator body; the lamp control assembly is used for changing the light environment in the incubator and providing the same light environment as that in the field for the Di-moths in the incubator; the in-box environment regulation and control module monitors the temperature and humidity in the box, compares the temperature and humidity data with the field temperature and humidity data in real time, and regulates the temperature and humidity in the box to be close to the field temperature and humidity.
The lamplight environment, the temperature and humidity environment and the field environment in the incubator are simulated and compared through the lamp control assembly and the in-box environment regulation and control module, so that the environment in the incubator is suitable for the normal growth activity of the Di-moths.
Acquiring a picture containing the litchi Di-moth cocoons and the adults shot by the shooting assembly in real time, and identifying and recording the Di-moth cocoons and the adults in the picture by adopting an identification model to obtain the Di-moth cocoons and the Di-moth quantities; calculating the current emergence rate of the Di-moth according to the Di-moth cocoons and the quantity of the Di-moth; the total fruit moth pest severity of the orchard is estimated and the optimal timing of the insecticide is estimated by referring to the quantity of fruit borers contained in the falling fruit and the current emergence rate of the fruit borers.
The method comprises the steps of constructing a training data set by using a pattern comprising the Di-moth cocoons and adults under different lamplight backgrounds, constructing a characteristic extraction network, extracting texture, size, shape and color characteristics of the Di-moth cocoons and adults by convolution operation, and finally realizing the identification and positioning of the Di-moth cocoons and adults in the pattern by establishing a multi-layer neural network model.
The recognition model performs feature fusion in the channel dimension and the space dimension, so that the channel attention is focused on the color, texture and shape features of the lichee pedicle worm cocoons and adults, and the space attention is focused on the position information of the lichee pedicle worm cocoons and adults.
In the embodiment, the incubator body comprises an outer frame, an inner frame, a continuous slope structure, a gentle slope truncated cone structure, a camera hole and a frame connecting piece, wherein the outer frame and the inner frame are connected through the frame connecting piece; the camera shooting assembly is fixed at the top end of the internal framework, the camera shooting hole is positioned above the internal framework, and the lamp control assembly is embedded above the external framework of the incubator body; the continuous slope-shaped structure is arranged at the bottom of the incubator body, and the included angle between slopes provides cocoon setting angle conditions for the litchi pedicel borers; the gentle slope round platform structure is positioned in the middle of the bottom surface of the incubator body and is higher than the continuous slope structure and used for filling the observation dead angle at the joint of the gentle slope round platform structure and the continuous slope structure.
For further details regarding the incubator assembly and identification model of this embodiment, reference is made to the relevant description of embodiment 1.
The identification model disclosed by the invention can automatically identify the litchi pedicle moth cocoons and the litchi pedicle moth adults in the pictures, count the number of the litchi pedicle moth cocoons and the litchi pedicle moth adults in a batch of fallen fruits in real time, calculate the eclosion rate, calculate the damage degree and the optimal pesticide spraying time of the integral litchi pedicle moth in an orchard, break the bottleneck that the traditional manual statistics of the number of the litchi pedicle moth cocoons is low and the error rate is high, and solve the problems that agricultural materials are wasted, the environment and the fruits are polluted and the like easily when pesticide spraying is carried out by virtue of subjective experience.
According to the method, the collected Di-moth cocoons and Di-moth adults generated by litchi falling fruits are accurately identified, the number of the Di-moth cocoons and the number of the adults are counted in real time, and the emergence rate of the Di-moth is automatically calculated day by day, so that the integral pest damage condition of an orchard can be calculated, and objective basis is provided for pesticide application of fruit farmers; the invention is beneficial to improving the detection efficiency and reducing the manpower.
The foregoing is illustrative of the present invention, and is not to be construed as limiting thereof, but rather as various changes, modifications, substitutions, combinations, and simplifications which may be made without departing from the spirit and principles of the invention are intended to be included within the scope of the invention.

Claims (10)

1. A method for litchi pedicel moth detection and eclosion rate calculation, characterized in that the method is implemented based on an incubator assembly; the incubator assembly comprises an incubator body, a camera assembly, a lamp control assembly and an in-box environment regulation and control module, wherein the camera assembly, the lamp control assembly and the in-box environment regulation and control module are positioned in the incubator body; the lamp control assembly is used for changing the light environment in the incubator and providing the same light environment as that in the field for the Di-moths in the incubator; the in-box environment regulation and control module monitors the temperature and humidity in the box, compares the temperature and humidity with the field temperature and humidity data in real time, and regulates the temperature and humidity in the box to be similar to the field temperature and humidity;
the method comprises the following steps:
s1, collecting a batch of fallen fruits from a litchi orchard, and placing the fallen fruits into an incubator;
s2, starting a lamp control assembly and an in-box environment regulation and control module, and performing simulation comparison on the lamplight environment, the temperature and humidity environment and the field environment in the incubator so that the environment in the incubator is suitable for normal growth activities of Di-moths;
s3, acquiring a picture containing the litchi Di moth cocoons and the adults shot by the shooting assembly in real time, and identifying and recording the Di moth cocoons and the adults in the picture by adopting an identification model to obtain the Di moth cocoons and the Di moth quantities;
s4, calculating the current emergence rate of the Di-moths according to the Di-moths cocoons and the quantity of the Di-moths;
s5, estimating the severity of the total fruit moth insect pests of the orchard and estimating the optimal timing of insecticide spraying according to the quantity of the fruit moth contained in the fallen fruit and the current emergence rate of the fruit moth.
2. The method for litchi fruit moth detection and eclosion rate calculation according to claim 1, wherein the identification model aims at the characteristics of small size, color similar to that of oxidized fruits and transparent characteristic of fruit moth cocoons, a training data set is constructed by pictures containing fruit moth cocoons and adults under different lamplight backgrounds, a feature extraction network is constructed, texture, size, shape and color characteristics of the fruit moth cocoons and adults are extracted by convolution operation, and finally identification and positioning of the fruit moth cocoons and adults in the pictures are realized by building a multi-layer neural network model.
3. A method for litchi fruit moth detection and feathering rate calculation as claimed in claim 2 wherein the identification model performs feature fusion in a channel dimension and a spatial dimension such that channel attention is focused on color, texture and shape features of litchi fruit moth cocoons and adults and spatial attention is focused on positional information of litchi fruit moth cocoons and adults.
4. A method for litchi fruit moth detection and calculation of the emergence rate as claimed in claim 1, wherein step S3 comprises:
s31, shooting the growth condition of the litchi pedicle moths in the incubator at regular time through a camera shooting assembly and a lamp control assembly in the incubator, and obtaining picture samples of the litchi pedicle moths under different light effects; labeling the litchi pedicel borer cocoons and adults in the pictures, constructing a data set, and dividing the data set into a training set and a test;
s32, constructing a recognition model of the litchi Di moth cocoons and adults;
carrying out attention feature fusion on the channel dimension and the space dimension of the recognition model, so that the attention of the channel is focused on the color, texture and shape features of the litchi Di-moth cocoons and adults, and the space attention is focused on the position information of the litchi Di-moth cocoons and adults; and a bidirectional information flow structure from top to bottom and from bottom to top is added in the improved model, so that the transmission of the characteristic information of the Di-moth between different network layers is enhanced.
5. A method for litchi fruit moth detection and feathering rate calculation as claimed in claim 4, wherein in the recognition model constructed in step S32, a 4 x 4 target detection layer is used; and constructing a combined cascade feature fusion structure and a parallel feature structure to promote fusion of low-level image features of the Di-moths and high-level image features of the Di-moths, wherein the low-level image features comprise pixel values, colors and edge features, and the high-level image features comprise position, shape and relation features.
6. The method for litchi fruit moth detection and eclosion rate calculation as claimed in claim 1, wherein the incubator box comprises an outer frame, an inner frame, a continuous slope structure, a gentle slope truncated cone structure, a camera hole and a frame connector, the outer frame and the inner frame being connected by the frame connector; the camera shooting assembly is fixed at the top end of the internal framework, the camera shooting hole is positioned above the internal framework, and the lamp control assembly is embedded above the external framework of the incubator body; the continuous slope-shaped structure is arranged at the bottom of the incubator body, and the included angle between slopes provides cocoon setting angle conditions for the litchi pedicel borers; the gentle slope round platform structure is positioned in the middle of the bottom surface of the incubator body and is higher than the continuous slope structure and used for filling the observation dead angle at the joint of the gentle slope round platform structure and the continuous slope structure.
7. A method for litchi fruit moth detection and calculation of emergence rate as defined in claim 6, wherein the incubator box further comprises a spacer screen covering the internal frame of the incubator box for closing the internal frame of the incubator box.
8. A system for litchi pedicel moth detection and eclosion rate calculation, comprising an incubator assembly and an identification model;
the incubator assembly comprises an incubator body, a camera assembly, a lamp control assembly and an in-box environment regulation and control module, wherein the camera assembly, the lamp control assembly and the in-box environment regulation and control module are positioned in the incubator body; the lamp control assembly is used for changing the light environment in the incubator and providing the same light environment as that in the field for the Di-moths in the incubator; the in-box environment regulation and control module monitors the temperature and humidity in the box, compares the temperature and humidity with the field temperature and humidity data in real time, and regulates the temperature and humidity in the box to be similar to the field temperature and humidity;
the lamplight environment, the temperature and humidity environment and the field environment in the incubator are simulated and compared through the lamp control assembly and the in-box environment regulation and control module, so that the environment in the incubator is suitable for the normal growth activity of the Di-moths;
acquiring a picture containing the litchi Di-moth cocoons and the adults shot by the shooting assembly in real time, and identifying and recording the Di-moth cocoons and the adults in the picture by adopting an identification model to obtain the Di-moth cocoons and the Di-moth quantities; calculating the current emergence rate of the Di-moth according to the Di-moth cocoons and the quantity of the Di-moth; the total fruit moth pest severity of the orchard is estimated and the optimal timing of the insecticide is estimated by referring to the quantity of fruit borers contained in the falling fruit and the current emergence rate of the fruit borers.
9. The system for litchi fruit moth detection and eclosion rate calculation as claimed in claim 8, wherein the identification model aims at the characteristics of small size, color similar to that of oxidized fruits and transparent characteristic of fruit moth cocoons, a training data set is constructed by pictures containing fruit moth cocoons and adults under different lamplight backgrounds, a feature extraction network is constructed, texture, size, shape and color characteristics of the fruit moth cocoons and adults are extracted by convolution operation, and finally identification and positioning of the fruit moth cocoons and adults in the pictures are realized by building a multi-layer neural network model;
the recognition model performs feature fusion in the channel dimension and the space dimension, so that the channel attention is focused on the color, texture and shape features of the lichee pedicle worm cocoons and adults, and the space attention is focused on the position information of the lichee pedicle worm cocoons and adults.
10. The system for litchi fruit moth detection and feathering rate calculation as claimed in claim 8, wherein the incubator box comprises an outer frame, an inner frame, a continuous ramp structure, a gentle-ramp truncated cone structure, a camera hole and frame connection members, the outer frame and the inner frame being connected by the frame connection members; the camera shooting assembly is fixed at the top end of the internal framework, the camera shooting hole is positioned above the internal framework, and the lamp control assembly is embedded above the external framework of the incubator body; the continuous slope-shaped structure is arranged at the bottom of the incubator body, and the included angle between slopes provides cocoon setting angle conditions for the litchi pedicel borers; the gentle slope round platform structure is positioned in the middle of the bottom surface of the incubator body and is higher than the continuous slope structure and used for filling the observation dead angle at the joint of the gentle slope round platform structure and the continuous slope structure.
CN202311343299.6A 2023-10-17 2023-10-17 Method and system for litchi pedicel moth detection and eclosion rate calculation Pending CN117612155A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311343299.6A CN117612155A (en) 2023-10-17 2023-10-17 Method and system for litchi pedicel moth detection and eclosion rate calculation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311343299.6A CN117612155A (en) 2023-10-17 2023-10-17 Method and system for litchi pedicel moth detection and eclosion rate calculation

Publications (1)

Publication Number Publication Date
CN117612155A true CN117612155A (en) 2024-02-27

Family

ID=89958532

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311343299.6A Pending CN117612155A (en) 2023-10-17 2023-10-17 Method and system for litchi pedicel moth detection and eclosion rate calculation

Country Status (1)

Country Link
CN (1) CN117612155A (en)

Similar Documents

Publication Publication Date Title
CN109029588A (en) A kind of Grain Growth Situation prediction technique based on climatic effect
WO2021250444A1 (en) Insect monitoring system and method
CN114818909A (en) Weed detection method and device based on crop growth characteristics
CN114460080A (en) Rice disease and pest intelligent monitoring system
CN115294518B (en) Intelligent monitoring method and system for precise greenhouse cultivation of horticultural plants
CN108829762A (en) The Small object recognition methods of view-based access control model and device
CN115861721B (en) Livestock and poultry breeding spraying equipment state identification method based on image data
CN114898405B (en) Portable broiler chicken anomaly monitoring system based on edge calculation
CN116543347A (en) Intelligent insect condition on-line monitoring system, method, device and medium
Varley et al. Parasites and predators
Mortensen et al. The oil radish growth dataset for semantic segmentation and yield estimation
WO2021198731A1 (en) An artificial-intelligence-based method of agricultural and horticultural plants' physical characteristics and health diagnosing and development assessment.
CN110245604B (en) Mosquito identification method based on convolutional neural network
CN115358440B (en) Tree coverage prediction method based on angular second moment and deep learning
CN117612155A (en) Method and system for litchi pedicel moth detection and eclosion rate calculation
CN115756037A (en) Greenhouse vegetable management and control method and system
CN114511801A (en) Insect pest recognition model training method, insect pest monitoring method and electronic equipment
CN113361377A (en) Plant growth control model construction method, electronic device and storage medium
CN115272943B (en) Livestock and poultry feeding abnormity identification method based on data processing
TWI804060B (en) Surveillance method for plant disease and pest and surveillance system for plant disease and pest
CN108288057B (en) Portable poultry life information detection device
Qu et al. Deep learning based detection of plant nutrient deficiency symptom and design of multi-layer greenhouse system
Krogh Mortensen et al. The oil radish growth dataset for semantic segmentation and yield estimation
US20230214997A1 (en) Training method, evaluation method, electronic device and storage medium
Sahu et al. ARTIFICIAL INTELLIGENCE IN AGRICULTURE

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination