CN-121982348-A - Rice field heavy metal pollution risk assessment method and system based on image recognition

CN121982348ACN 121982348 ACN121982348 ACN 121982348ACN-121982348-A

Abstract

The invention relates to the technical field of image recognition, and provides a paddy field heavy metal pollution risk assessment method and system based on image recognition, wherein the method comprises the steps of acquiring multispectral image data of a paddy field region; the method comprises the steps of calculating relative spectral features among selected spectral bands related to heavy metal stress from multispectral image data aiming at each pixel point, comparing the relative spectral features with a pre-established heavy metal stress feature mode set to obtain feature comparison results, establishing the heavy metal stress feature mode set based on relative spectral feature modes corresponding to the heavy metal stress observed under different equipment degradation states, identifying heavy metal pollution areas based on similarity discrimination thresholds and spatial continuity according to the feature comparison results, and carrying out risk classification to obtain corresponding paddy field heavy metal pollution risk assessment information. The method has the effect of improving the accuracy of rice field heavy metal pollution risk assessment.

Inventors

LI LINFENG
WEN WEIFA
YAO YUE
LI YICHUN
LI QI
XIAO ANWEN

Assignees

广东省农业科学院农业资源与环境研究所

Dates

Publication Date: 20260505
Application Date: 20260126

Claims (10)

1. The rice paddy heavy metal pollution risk assessment method based on image recognition is characterized by comprising the following steps of: Acquiring multispectral image data of a paddy field area; Calculating relative spectral characteristics between selected spectral bands related to heavy metal stress for each pixel point from the multispectral image data; Comparing the relative spectral characteristics with a pre-established heavy metal stress characteristic mode set to obtain characteristic comparison results, wherein the heavy metal stress characteristic mode set is established based on the relative spectral characteristic modes corresponding to the heavy metal stress observed under different equipment degradation states; and identifying heavy metal pollution areas and carrying out risk classification based on similarity discrimination threshold and space continuity according to the characteristic comparison result to obtain corresponding paddy field heavy metal pollution risk assessment information.
2. The method for evaluating the risk of heavy metal pollution in paddy fields based on image recognition according to claim 1, wherein the step of establishing the heavy metal stress characteristic pattern set based on the relative spectrum characteristic patterns corresponding to the heavy metal stress observed in different equipment degradation states comprises the following steps: Acquiring an original digital value of a reference area according to a preset period, and calculating a differential change curve of an optical component of the equipment in a specific wave band according to the original digital values observed at different time points; Recording the image acquisition time and the environmental context information of an acquisition area in real time during each image acquisition; carrying out reverse differential calibration on the original digital value of the multispectral image data by combining the differential change curve to obtain the relative spectral characteristics of the calibrated device; Performing environmental effect stripping on the relative spectrum characteristics after equipment calibration by utilizing the image acquisition time and the environmental context information and combining a pre-established environmental mimicry effect library to obtain stress residual spectrum characteristics; and taking the stress residual spectrum characteristics as purification data, and establishing a heavy metal stress characteristic mode set.
3. The method for evaluating the risk of heavy metal pollution in paddy fields based on image recognition according to claim 2, wherein the step of using the stress residual spectrum features as the purification data for establishing the heavy metal stress feature pattern set comprises the following steps: analyzing similarity and difference between the stress residual spectrum characteristics and the existing modes in the heavy metal stress characteristic mode set; Marking the stress residual spectrum characteristics with the similarity with the existing mode lower than a preset threshold value as potential new mode samples; Searching whether an existing mode which is consistent with the environmental context information of the potential new mode sample exists in the heavy metal stress characteristic mode set or not by combining the rice growth stage, the season information and the soil trace element background data during the acquisition of the potential new mode sample, and obtaining a search result; if the search result indicates that the potential new mode sample exists, calculating the spectrum characteristic difference degree of the potential new mode sample and the existing mode; If the spectrum characteristic difference degree exceeds a preset threshold, triggering a mode evolution prompt to indicate that the existing mode has evolved, and carrying out ground verification by requiring the intervention of an agronomic expert; if the search result indicates that the environment model is not available, marking the potential new model sample as a potential new environment model sample, and triggering a new model discovery prompt to prompt an agronomic expert to perform ground verification; Receiving real labels and environmental context information returned after ground verification of an agronomic expert; and incrementally updating the heavy metal stress characteristic mode set according to the returned real label and environmental context information.
4. The method for evaluating the risk of heavy metal pollution to paddy fields based on image recognition according to claim 3, wherein the step of marking the stress residual spectrum features with the similarity with the existing pattern lower than a preset threshold value as potential new pattern samples comprises the following steps: Calculating similarity distribution of the stress residual spectrum characteristics and all existing modes in the heavy metal stress characteristic mode set; Calculating a classification uncertainty score of the stress residual spectrum feature according to the similarity distribution; calculating a spatial polymerization degree score of the stress residual spectrum characteristic by combining the geographic position of the stress residual spectrum characteristic; and comprehensively evaluating the stress residual spectrum characteristics according to the classification uncertainty score and the spatial aggregation degree score, and marking the stress residual spectrum characteristics as potential new mode samples.
5. The method for evaluating the risk of heavy metal pollution in paddy fields based on image recognition according to claim 1, wherein the step of calculating the relative spectral characteristics between the selected spectral bands related to heavy metal stress from the multispectral image data for each pixel point comprises: preprocessing the multispectral image data, and extracting original digital values corresponding to each spectrum wave band aiming at each pixel point in the paddy field area; Determining a characteristic wave band combination serving as a selected spectrum wave band based on physiological response characteristics of rice to heavy metal stress; and for each pixel point, calculating the relative spectral characteristics reflecting heavy metal stress based on the original digital values corresponding to the spectral bands in the characteristic band combination.
6. The method for evaluating the risk of heavy metal pollution in paddy fields based on image recognition according to claim 4, wherein the step of comprehensively evaluating the stress residual spectrum features and marking the stress residual spectrum features as potential new pattern samples according to the classification uncertainty score and the spatial aggregation score comprises the steps of: acquiring current growth stage and season information of rice and background data of soil trace elements; According to the growth stage, the season information and the soil trace element background data, obtaining classification uncertainty score weight and space aggregation score weight aiming at the current environment context from a preset dynamic weight adjustment rule; Calculating a weighted sum of the classification uncertainty score and the spatial aggregation score; And marking the stress residual spectrum characteristic as a potential new mode sample according to the weighted sum and combining a dynamically adjusted comprehensive threshold value.
7. The method for evaluating the risk of heavy metal pollution in paddy fields based on image recognition according to claim 6, wherein the step of obtaining the classification uncertainty score weight and the spatial aggregation score weight for the current environmental context from the preset dynamic weight adjustment rule according to the growth stage, the season information and the soil trace element background data comprises: continuously monitoring the change rate and amplitude of the growth stage, the season information and the soil trace element background data; Triggering an environment context updating mechanism when the change rate or amplitude exceeds a preset dynamic threshold value; In the environment context updating mechanism, according to the current growth stage, the current season information and the real-time value of the background data of the soil trace elements, a preset weight adjustment function is combined, and the classification uncertainty score weight and the space aggregation score weight aiming at the current environment context are calculated and output.
8. The method for evaluating the risk of heavy metal pollution in paddy fields based on image recognition according to claim 7, wherein the step of continuously monitoring the growth stage, the season information and the change rate and the amplitude of the background data of the trace elements of the soil comprises: According to the update frequency and precision of each sensor, performing time alignment and spatial interpolation processing on the growth stage, the season information and the soil trace element background data; according to the reliability level of each sensor, weighting and fusing the growth stage, season information and soil trace element background data after time alignment and spatial interpolation processing; performing redundancy analysis on the weighted and fused growth stage, season information and soil trace element background data to obtain redundancy-removed data; Setting an adaptive time window on the basis of redundancy data removal; calculating the change rate and fluctuation amplitude of redundancy elimination data in each self-adaptive time window; And adjusting the length of the subsequent self-adaptive time window according to the change characteristics of the redundancy elimination data in the self-adaptive time window.
9. The method for evaluating the risk of heavy metal pollution in paddy fields based on image recognition according to claim 8, wherein the step of adjusting the length of the subsequent adaptive time window according to the change characteristics of the redundancy elimination data in the adaptive time window comprises: dividing a rice field area into a plurality of sub-areas; analyzing the growth stage, the seasonal information and the change characteristics of soil trace element background data in the subareas aiming at each subarea to obtain the local change characteristics of the subareas; determining the length of the self-adaptive time window of each sub-region according to the local change characteristic of each sub-region; And taking the length of the self-adaptive time window of each sub-area as the length of the time window calculated by the growth stage, the seasonal information and the change rate and the change amplitude of the soil trace element background data in the sub-area.
10. A paddy field heavy metal pollution risk assessment system based on image recognition for performing paddy field heavy metal pollution risk assessment based on image recognition, comprising: The image data acquisition module is used for acquiring multispectral image data of the paddy field area; the spectrum characteristic calculation module is used for calculating relative spectrum characteristics between selected spectrum bands related to heavy metal stress aiming at each pixel point from the multispectral image data; The characteristic comparison execution module is used for comparing the relative spectral characteristics with a pre-established heavy metal stress characteristic mode set to obtain characteristic comparison results, wherein the heavy metal stress characteristic mode set is established based on the relative spectral characteristic modes corresponding to the heavy metal stress observed in different equipment degradation states; And the pollution risk assessment module is used for identifying heavy metal pollution areas and carrying out risk classification based on the similarity discrimination threshold and the space continuity according to the characteristic comparison result so as to obtain corresponding paddy field heavy metal pollution risk assessment information.

Description

Rice field heavy metal pollution risk assessment method and system based on image recognition Technical Field The invention relates to the technical field of image recognition, in particular to a paddy field heavy metal pollution risk assessment method and system based on image recognition. Background In modern agricultural production, high-efficiency and accurate monitoring and risk assessment of paddy field heavy metal pollution are important. Traditional monitoring methods are inefficient, difficult to cover large areas, and have long data acquisition and analysis cycles, resulting in a lag in pollution risk assessment. Therefore, an intelligent analysis method based on image recognition is commonly adopted in the industry, and multispectral images of a paddy field area are acquired through unmanned aerial vehicle or satellite remote sensing equipment so as to realize rapid recognition and risk assessment of potential pollution areas. However, in practice, such systems are subjected to difficult-to-detect physical changes in equipment and complex environmental factors during long-term operation, which may interact with the patterns of human operator behavior, thereby creating a complex series of specific technical dilemmas that cause the system to deviate in early, low-concentration heavy metal pollution identification and assessment. In a specific agricultural environment such as a paddy field, a multispectral camera lens carried by an unmanned aerial vehicle is exposed to high humidity and attack of trace pesticide mist for a long period of time in daily flight operation. This environment is not extreme, but its persistent effect may result in microstructural changes in the anti-reflective coating of the lens surface. This change is not a macroscopic scratch or break, but rather a slow, cumulative degradation at the molecular level. This degradation does not immediately cause the camera to fail, nor trigger a conventional fault alarm in the device self-test system, as it is far from reaching the normally set physical damage threshold. Due to inaccuracy in heavy metal stress feature extraction, the system cannot effectively match rice features in early and low concentration pollution areas when compared with a preset heavy metal stress feature pattern set. This is because the feature patterns in the information set are established based on accurate spectral data, and the features extracted by the current system deviate from these patterns due to distortion. Therefore, the threshold parameters preset by the system to classify the pollution risk cannot be effectively triggered, so that the risk assessment result of the system on the early polluted areas is generally low, and even is reported as no risk by mistake. This means that the system fails to give an effective warning in the phase where contamination just begins to accumulate, most easily intervene. In view of the above, there is a need in the art for improvements. Disclosure of Invention The application discloses a paddy heavy metal pollution risk assessment method and system based on image identification, and aims to solve the technical problems that in the paddy heavy metal pollution risk assessment method based on image identification in the prior art, due to interaction of physical changes of equipment and complex environmental factors, early and low-concentration heavy metal pollution identification and assessment deviate, and the pollution risk assessment result is generally low or even the system reports no risk in error. The technical scheme of the application is as follows: in a first aspect, the application discloses a paddy field heavy metal pollution risk assessment method based on image recognition, which comprises the following steps: Acquiring multispectral image data of a paddy field area; calculating relative spectral features between selected spectral bands related to heavy metal stress for each pixel point from the multispectral image data; Comparing the relative spectral features with a pre-established heavy metal stress feature pattern set to obtain feature comparison results, wherein the heavy metal stress feature pattern set is established based on the relative spectral feature patterns corresponding to the heavy metal stress observed in different equipment degradation states; And identifying heavy metal pollution areas and carrying out risk classification based on similarity discrimination threshold and space continuity according to the characteristic comparison result to obtain corresponding paddy field heavy metal pollution risk assessment information. Through the technical scheme, the problem that early heavy metal pollution identification is inaccurate due to equipment degradation and environmental complexity in the prior art can be effectively solved, and accuracy and timeliness of rice field heavy metal pollution risk assessment are remarkably improved. Further, the application also provides that the step of establish