CN-121392608-B - Crop pest identification method and system based on visual identification

Abstract

The invention discloses a crop pest identification method and system based on visual identification, and relates to the technical field of pest identification. A crop pest identification system based on visual identification comprises a crop visual processing module and a crop pest identification module. The method and the device construct a multisource information cooperative judgment mechanism by fusing visual image characteristics, volatile gas concentration and environmental parameters, effectively distinguish physiological abnormality from pathogenic diseases, reduce misjudgment and missed judgment caused by environmental interference, remarkably improve the recognition accuracy and result reliability, analyze the respiratory characteristics and the mature state of crops by combining gas sensing data, can recognize the physiological abnormality of the crops before visual symptoms appear, realize early discovery and early warning of insect pests, dynamically adjust recognition threshold and sensitivity according to different growth stages, support targeted control decision and improve control timeliness.

Inventors

• TAO XINGZHEN
• ZENG JUN
• LI HAIPING
• HUANG HUIYAN

Assignees

• 江西应用技术职业学院

Dates

Publication Date

20260505

Application Date

20251121

Claims (5)

1. 1. The crop pest identification method based on visual identification is characterized by comprising the following steps of: the method comprises the steps of acquiring image data of crops in real time based on visual equipment to obtain crop surface image data, constructing a crop visual identification model, processing the crop surface image data based on the crop visual identification model to obtain abnormal crop surface images and crop growth stage characteristics; The method comprises the steps of synchronously collecting the concentration of volatile gas around crops based on gas sensing equipment, identifying the state of the crops to obtain the ripening characteristics of the crops, wherein the ripening characteristics of the crops are used for representing the ripening stage and the expected ripening time of the crops; Carrying out pest identification on crops based on the crop environmental characteristics, the crop maturation characteristics and the abnormal surface images of the crops to obtain a crop pest identification result; The specific steps for identifying the crop state of the volatile gas concentration comprise the following steps: comparing the gas concentration mean value of the standard time sequence gas concentration data with a volatile gas threshold interval of corresponding crops to obtain a basic maturity grade of the crops; Meanwhile, extracting the first gas concentration and the second gas concentration from the standard time sequence gas concentration data, and comparing to obtain an active gas ratio; judging that the corresponding crops are in a respiratory transition period when the active gas ratio exceeds a preset gas ratio, otherwise, the corresponding crops are in a respiratory stationary period, and analyzing the daily change rule of the third gas concentration to obtain the release characteristics of volatile substances; Establishing a gas concentration change trend model according to the respiratory jump period or the respiratory stationary period, and continuously monitoring the change slopes of the first gas concentration, the second gas concentration and the third gas concentration to obtain the characteristics of predicting the ripening progress of crops; carrying out characteristic fusion according to the basic maturity grade of the crops, the release characteristics of volatile substances and the characteristics of predicting the crop maturation process to obtain the crop maturation characteristics; the specific steps for identifying insect pests of crops based on crop environmental characteristics, crop maturation characteristics and abnormal surface images of the crops comprise the following steps: carrying out association analysis based on the environmental characteristics of the crops and the mature characteristics of the crops to obtain external characteristics of crop diseases; The growth state of the current crop and the interrelation between the current crop and the environment are known by analyzing the factors of temperature, humidity, illumination intensity and soil humidity in the environment where the crop is positioned and combining the maturity characteristics of the crop; The external characteristics of the diseases comprise correlation between the marking changes on the surfaces of crops and the types of the insect pests, and the identification sensitivity of the corresponding insect pests is matched by combining the existing insect pest data; Carrying out space correlation analysis on the environmental characteristics of the crops and the corrected pest distribution, and identifying the crop pests by combining the mature characteristics of the crops to obtain a crop pest identification result; The method comprises the steps of identifying a pest identification threshold value of an abnormal surface image of a crop according to identification sensitivity, carrying out adaptive adjustment on pest identification requirements of different environments, different crop varieties and different maturity stages by dynamically adjusting the pest identification threshold value, obtaining a pest distribution map after the pest identification threshold value is corrected, and reducing the phenomena of false identification and missing identification; the method comprises the steps of analyzing and identifying the distribution rule of insect pests in farmlands or orchards through spatial association, comprehensively evaluating the distribution rule of the insect pests in the farmlands or orchards by combining the environmental characteristics, correcting the insect pest distribution and the spatial data of the mature characteristics, and comprehensively identifying the insect pests of the crops to obtain crop insect pest identification results, wherein the crop insect pest identification results specifically represent whether the insect pests exist on the surfaces or in the surrounding environment of the crops or not and the types, the positions, the severity and the influence areas of the insect pests by analyzing the visual images, the gas concentration data, the environmental characteristics and the maturity of the crops; The setting of the volatile gas threshold interval and the preset gas ratio is based on the physiological characteristics of crops and the natural association of the physiological characteristics with environmental changes, and is not a numerical value set at will by people.
2. 2. A crop pest identification method based on visual identification as claimed in claim 1, characterized by the specific step of processing crop surface image data based on a crop visual identification model, comprising: the crop visual recognition model comprises an image dividing module, an abnormal characteristic enhancing module and a growth characteristic judging module; the method comprises the steps of carrying out abnormal feature division on crop surface image data in an image division module, extracting a crop to-be-identified abnormal image containing abnormal features and a crop to-be-identified background image containing a large-scale background; The method comprises the steps of carrying out crop growth stage identification on a background image to be identified of crops in a growth characteristic judging module to obtain crop growth stage characteristics, specifically, carrying out layer-by-layer characteristic extraction on a crop area of the background image to be identified of the crops by utilizing multi-scale convolution characteristic extraction in the growth characteristic judging module to obtain crop growth characteristics, extracting chromaticity information in the crop growth characteristics to obtain crop growth chromaticity characteristics, and judging the crop growth chromaticity characteristics by utilizing a trained classifier to obtain the crop growth stage characteristics.
3. 3. The visual recognition-based crop pest recognition method according to claim 2, wherein the specific step of constructing the abnormal feature enhancement module in the crop visual recognition model comprises: Introducing an ECA attention mechanism into a Backbone network of the basic model to serve as a convolution feature extraction unit, adding a multi-task learning branch into a Head part to serve as a crop feature classification unit, and using a CSPNet structure into a Neck part to serve as a crop feature fusion unit; And processing the abnormal image to be identified of the crop in the abnormal characteristic enhancement module to obtain an abnormal surface image of the crop.
4. 4. A crop pest identification method based on visual identification as claimed in claim 3, wherein the first gas, the second gas and the third gas respectively represent different kinds of gas for different kinds of crops.
5. 5. A crop pest identification system based on visual identification, wherein the system applies a crop pest identification method based on visual identification as claimed in any one of claims 1 to 4, comprising: The crop visual processing module comprises a visual processing unit and an auxiliary processing unit, wherein the visual processing unit is used for acquiring image data of crops in real time based on visual equipment to obtain crop surface image data, constructing a crop visual recognition model, and processing the crop surface image data based on the crop visual recognition model to obtain abnormal crop surface images and crop growth stage characteristics; the auxiliary processing unit is used for synchronously collecting the concentration of volatile gas around crops based on the gas sensing equipment, and identifying the state of the crops to obtain the ripening characteristics of the crops, wherein the ripening characteristics of the crops are used for representing the ripening stage and the expected ripening time of the crops; the crop pest identification module comprises a pest identification unit, wherein the pest identification unit is used for acquiring crop environment characteristics, and carrying out pest identification on crops based on the crop environment characteristics, crop maturation characteristics and abnormal surface images of the crops to obtain crop pest identification results.

Description

Crop pest identification method and system based on visual identification Technical Field The invention relates to the technical field of pest identification, in particular to a crop pest identification method and system based on visual identification. Background In the prior art, the crop pest identification method based on vision mainly depends on an image processing and deep learning model, and realizes automatic identification to a certain extent, but has obvious limitations that firstly physiological diseases and infectious diseases are difficult to distinguish by means of image features, are easily interfered by environmental factors such as illumination and shielding to cause misjudgment and missed judgment, secondly, dynamic perception of the growth state of crops is lacking, comprehensive judgment can not be carried out by combining physiological indexes such as maturity and respiratory characteristics, identification accuracy and timeliness are insufficient, and thirdly, early warning and accurate prevention and control of pests are difficult to realize due to the fact that biochemical information such as gas volatile matters can not be effectively fused. Accordingly, there is a need to provide a crop pest identification method and system based on fusion of vision and gas sensing. Disclosure of Invention The invention aims to provide a crop pest identification method and a crop pest identification system based on visual identification, which can dynamically adjust pest identification sensitivity and threshold value, realize accurate identification and early warning of pest types in different growth stages and different environmental conditions, and remarkably improve identification accuracy. A crop pest identification method based on visual identification comprises the following steps: the method comprises the steps of acquiring image data of crops in real time based on visual equipment to obtain crop surface image data, constructing a crop visual identification model, processing the crop surface image data based on the crop visual identification model to obtain abnormal crop surface images and crop growth stage characteristics; The method comprises the steps of synchronously collecting the concentration of volatile gas around crops based on gas sensing equipment, identifying the state of the crops to obtain the ripening characteristics of the crops, wherein the ripening characteristics of the crops are used for representing the ripening stage and the expected ripening time of the crops; And carrying out pest identification on the crops based on the crop environmental characteristics, the crop maturation characteristics and the abnormal surface images of the crops to obtain a crop pest identification result. As a preferred technical scheme of the present invention, the specific steps for processing the crop surface image data based on the crop visual recognition model include: the crop visual recognition model comprises an image dividing module, an abnormal characteristic enhancing module and a growth characteristic judging module; the method comprises the steps of carrying out abnormal feature division on crop surface image data in an image division module, extracting a crop to-be-identified abnormal image containing abnormal features and a crop to-be-identified background image containing a large-scale background; The method comprises the steps of carrying out crop growth stage identification on a background image to be identified of crops in a growth characteristic judging module to obtain crop growth stage characteristics, specifically, carrying out layer-by-layer characteristic extraction on a crop area of the background image to be identified of the crops by utilizing multi-scale convolution characteristic extraction in the growth characteristic judging module to obtain crop growth characteristics, extracting chromaticity information in the crop growth characteristics to obtain crop growth chromaticity characteristics, and judging the crop growth chromaticity characteristics by utilizing a trained classifier to obtain the crop growth stage characteristics. As a preferable technical scheme of the invention, the specific steps of constructing the abnormal characteristic enhancement module in the crop visual recognition model comprise the following steps: Introducing an ECA attention mechanism into a Backbone network of the basic model to serve as a convolution feature extraction unit, adding a multi-task learning branch into a Head part to serve as a crop feature classification unit, and using a CSPNet structure into a Neck part to serve as a crop feature fusion unit; And processing the abnormal image to be identified of the crop in the abnormal characteristic enhancement module to obtain an abnormal surface image of the crop. As a preferable technical scheme of the invention, the specific steps for identifying the crop state by the concentration of the volatile gas comprise the following step