A digitalized aquaculture environment monitoring method and system

Abstract

The present application relates to the technical field of environment detection, and relates to a digital aquaculture environment monitoring method and system, wherein the method comprises: continuously collecting underwater images, dissolved oxygen concentration and turbidity, and segmenting the underwater images to obtain water area images; obtaining a plurality of color moments of the water area images, the plurality of color moments comprising a first moment, a second moment and a third moment, and obtaining a plurality of color features of each color moment; constructing a gray level co-occurrence matrix according to the water area images; calculating a risk score; and performing early warning according to the risk score. The present application dynamically reflects the influence of environmental changes by calculating the risk score through standardized weighted summation, thereby improving the adaptability and flexibility of aquaculture environment management.

Description

Technical Field

[0001] This invention relates to the field of environmental monitoring technology. More specifically, this invention relates to a digital method and system for monitoring the environment in aquaculture. Background Technology

[0002] With the large-scale and intensive development of aquaculture, the demand for precise monitoring and management of the aquaculture environment is increasing. Currently, aquaculture environmental monitoring methods are gradually developing towards digitalization and intelligence, but they are still in a stage of continuous exploration and improvement. Traditional manual monitoring methods rely on experience-based judgment, which suffers from low data collection frequency and response delays, making it difficult to detect water quality anomalies or changes in fish behavior in a timely manner. Chinese patent application CN108828168A discloses an aquaculture environmental monitoring and maintenance device. This application enables real-time monitoring of the aquaculture environment, and can store and compare the collected environmental parameters, making real-time adjustments based on the current environment.

[0003] While existing monitoring systems can acquire some basic environmental parameters, such as water temperature and dissolved oxygen, they lack multimodal data fusion analysis, resulting in insufficient accuracy of risk warnings. Furthermore, existing systems are mostly based on static threshold triggering for warnings, which cannot adapt to dynamic scenarios such as seasonal changes and differences in aquaculture density. This leads to a high false alarm rate and makes it difficult to comprehensively, in real time, and accurately reflect changes in the aquaculture environment, thus failing to meet the needs of efficient and sustainable development in modern aquaculture.

[0004] Existing aquaculture environmental monitoring methods rely on single monitoring indicators and lack sufficient characteristic correlation analysis, failing to quantify the statistical correlation between environmental indicators and risk events. They depend on human experience to allocate weights, making it difficult to accurately assess the overall condition of the aquaculture environment. They have poor dynamic adaptability, lacking the ability to self-adjust and respond to sudden risks or long-term environmental changes. Their response mechanisms are simplistic and do not fully consider the complex relationship between the aquaculture environment and different risk events, resulting in delayed or inaccurate early warnings and an inability to prevent risk events in a timely and effective manner. Summary of the Invention

[0005] To address the technical problems that traditional aquaculture environmental monitoring methods may lead to delayed or inaccurate early warnings, this invention provides solutions in the following aspects.

[0006] In a first aspect, a digital aquaculture environment monitoring method is provided, comprising: continuously acquiring underwater images, dissolved oxygen concentration, and turbidity; segmenting the underwater images to obtain a water body region image; obtaining multiple color moments of the water body region image, the multiple color moments including a first moment, a second moment, and a third moment, and obtaining multiple color features of each color moment; constructing a gray-level co-occurrence matrix based on the water body region image, and obtaining multiple texture features of the gray-level co-occurrence matrix; standardizing the multiple color features, multiple texture features, dissolved oxygen concentration, and turbidity to obtain corresponding feature values; and performing a weighted summation of each feature value to obtain a risk score; wherein, the Pearson correlation coefficient between each feature value and the corresponding preset risk event is calculated to determine the weight of each feature value. , For the first i The weights of each feature value α The nonlinear amplification factor is a preset size, and α >1, For the first i The Pearson correlation coefficients corresponding to the eigenvalues, For the first j The Pearson correlation coefficients corresponding to the eigenvalues, i , j All values ​​are positive integers, where n is the number of all feature values; warnings are issued based on the risk score. The beneficial effects are: This invention uses multiple feature values ​​such as color moments, texture features, dissolved oxygen concentration, and turbidity, combined with weighted calculation of the risk score, to help comprehensively assess the overall condition of the aquaculture environment. This comprehensive assessment method avoids the limitations of a single indicator, enabling analysis of environmental risks from multiple perspectives and providing more comprehensive and accurate data support for risk management. This invention calculates the risk score through standardized weighted summation, dynamically reflecting the impact of environmental changes. Based on real-time data changes, the system can adaptively adjust the risk score, thereby responding to sudden risks or long-term environmental changes, enhancing the adaptability and flexibility of aquaculture environment management.

[0007] Preferably, a digital aquaculture environment monitoring method further includes correcting the weights of eigenvalues: in response to a Pearson correlation coefficient corresponding to an eigenvalue being greater than a preset first threshold, the eigenvalue is determined to be an important eigenvalue; eigenvalues ​​other than important eigenvalues ​​are determined to be unimportant eigenvalues; and the weights of the eigenvalues ​​are calculated. h Weights of important eigenvalues : ,in For the first h The Pearson correlation coefficients corresponding to the key eigenvalues For the first s The Pearson correlation coefficients corresponding to the key eigenvalues For a preset high proportion of weight; calculate the first q Weights of non-important eigenvalues : ,in For the first q Pearson correlation coefficients corresponding to non-significant eigenvalues For the first p Pearson correlation coefficients corresponding to non-significant eigenvalues For a preset low-proportion weight, where and .

[0008] This invention improves the sensitivity of weights to risk events by dividing relevant feature indicators corresponding to risk events into important groups (including important feature values) and unimportant groups (including unimportant feature values) through grouped competitive weights.

[0009] Preferably, calculating the risk score includes: obtaining multiple color features, wherein the multiple color features include: the mean of the R, G, and B channels in the first moment of the water body area image, the variance of the R, G, and B channels in the second moment, and the skewness of the R, G, and B channels in the third moment; obtaining multiple texture features, wherein the multiple texture features include: the contrast, energy, homogeneity, and correlation of the gray-level co-occurrence matrix; and standardizing the mean, variance, skewness, contrast, energy, homogeneity, correlation, oxygen desorption concentration, and turbidity of the R, G, and B channels, and then performing a weighted summation to obtain the risk score.

[0010] Preferably, a digital aquaculture environment monitoring method further includes: at preset intervals, adjusting the weights of each feature value using exponential decay, wherein the weights of the first feature value are adjusted... t The first moment i The formula for exponential decay of eigenvalues ​​is: , For the revised first t The first moment i 1 eigenvalue, For the first time before the correction t The first moment i 1 eigenvalue, For the first i The preset decay rate corresponding to each feature value T n The preset standard temperature value, T t For the first t The underwater temperature at any given time, t It can be any positive integer.

[0011] This invention avoids weight drift by attenuating the weights corresponding to each feature value.

[0012] Preferably, in response to the first t The first moment i The first characteristic value is dissolved oxygen concentration; calculate the... t The first moment i The decay rate corresponding to each characteristic value γ do The formula is:

[0013] ;

[0014] Among them in the first t time, This represents the mean of the G channel. Let G be the variance of the channel. α 1 and β 1 represents a preset adjustment coefficient. Let G be the mean of the values ​​of the G channel at all times up to time t. Let be the mean of the variances of the G channel for all times up to time t.

[0015] Preferably, obtain the current time's first... i The eigenvalues ​​include: the eigenvalues ​​at the current time for the eigenvalue. i Multiple samples are taken from each feature value to obtain multiple sample values; the mean of the multiple sample values ​​is determined as the th value at the current time. i Each feature value.

[0016] This invention performs multiple samplings and average calculations when obtaining feature values, reducing the random errors that may be caused by a single sampling, thereby improving the stability and accuracy of feature values.

[0017] Preferably, at the current moment, for the first i The eigenvalue and the eigenvalue i Each feature value corresponds to a preset risk event. m Sub-sampling, of which For the first i The eigenvalue of the th eigenvalue k Each sample value, For the first i The mean of all sampled values ​​of the feature value; at the current time, the mean of all sampled values ​​of the feature value; k The second sampling determines the first i When the preset risk event corresponding to each feature value does not occur, determine the risk event label. The value is 1, and the number of times the current time passes is 1. k The second sampling determines the first i When a preset risk event corresponding to a feature value occurs, the risk event label is determined. =0; calculate the current time step. i Pearson correlation coefficients corresponding to each eigenvalue The formula is: , The mean of the risk event labels, i.e. .

[0018] Preferably, at any given time, issuing an early warning based on the risk score includes: in response to the risk score being greater than or equal to a first early warning threshold and less than a second early warning threshold, increasing the frequency of underwater data collection, wherein the first early warning threshold is less than the second early warning threshold; in response to the risk score being greater than or equal to the second early warning threshold and less than a third early warning threshold, sending a control signal to the aerator to increase the aerator's operating power and sending an early warning signal to relevant personnel, wherein the second early warning threshold is less than the third early warning threshold; in response to the risk score being greater than or equal to the third early warning threshold and less than a fourth early warning threshold, sending a control signal to the aerator to make the aerator's operating power reach its rated power and sending an early warning signal to relevant personnel, wherein the third early warning threshold is less than the fourth early warning threshold; and in response to the risk score being greater than or equal to the fourth early warning threshold, sending a fish evacuation signal to relevant personnel.

[0019] This invention, through a graded response early warning mechanism, can provide hierarchical responses based on real-time risk scores, thereby improving the safety and management efficiency of aquaculture.

[0020] Preferably, the first warning threshold is 0.3, the second warning threshold is 0.5, the third warning threshold is 0.7, and the fourth warning threshold is 0.9.

[0021] In a second aspect, a digital aquaculture environment monitoring system includes a processor and a memory, the memory storing a computer program, and the processor executing the computer program to implement a digital aquaculture environment monitoring method as described in any one of the above inventions.

[0022] The beneficial effects of this invention are as follows:

[0023] This invention provides a comprehensive assessment of aquaculture environmental conditions by combining multiple feature values ​​such as color moments, texture characteristics, dissolved oxygen concentration, and turbidity, and employing a weighted risk score calculation method. This method overcomes the limitations of single indicators, enabling the analysis of environmental risks from multiple perspectives and providing more comprehensive and accurate data support for risk management. The standardized, weighted summation of the risk score dynamically reflects the impact of environmental changes and can adaptively adjust the score to address sudden risks or long-term environmental changes, thus improving the adaptability and flexibility of aquaculture environment management. Attached Figure Description

[0024] Figure 1 This is a schematic flowchart illustrating the steps of a digital aquaculture environment monitoring method according to an embodiment of the present invention;

[0025] Figure 2 This is a schematic diagram illustrating the structure of a digital aquaculture environment monitoring system according to this embodiment. Detailed Implementation

[0026] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0027] Figure 1 This is a schematic flowchart illustrating the steps of a digital aquaculture environment monitoring method according to an embodiment of the present invention.

[0028] like Figure 1 As shown, a digital aquaculture environment monitoring method includes steps S1 to S5.

[0029] Step S1: Continuously acquire underwater images, dissolved oxygen concentration, and turbidity, and segment the underwater images to obtain images of the water body area.

[0030] In one embodiment, underwater images, dissolved oxygen concentration, and turbidity are acquired at predetermined frequency intervals of 0.1 Hz.

[0031] This invention performs basic preprocessing on acquired underwater images, inputs them into a pre-trained U-Net model, and uses this model to segment the water region image. The final output is a binary mask (where 0 represents the background and 1 represents the Region of Interest (ROI). In aquaculture scenarios, the ROI is defined as the water region image. Only pixels within the mask are retained as the water region image for subsequent calculations, thus eliminating background interference and ensuring the calculation results accurately reflect the characteristics of the water region. Post-processing steps are performed on the segmented ROIs: first, a closing operation (kernel size set to 5×5) is used to fill the holes generated during segmentation; then, connected component analysis is performed to retain the largest connected regions, effectively eliminating sporadic noise points and further optimizing the segmentation results.

[0032] Step S2: Obtain multiple color moments of the water area image, including a first moment, a second moment, and a third moment, and obtain multiple color features of each color moment.

[0033] Specifically, the water area image is split into three independent single-channel images: red channel R, green channel G, and blue channel B, to facilitate accurate statistical analysis later. After splitting, all pixel values ​​(e.g., the 0-255 grayscale values ​​presented in channel G) constitute their own independent datasets within each channel. To eliminate the influence of illumination differences on the statistical results, the pixel values ​​in the dataset corresponding to each channel are mapped to the range [0,1]. After completing the above preparations, the color moments I (the first, second, and third moments are all referred to as color moments) are calculated.

[0034] The average values ​​for the three channels were calculated separately to obtain the average values ​​for the green, blue, and red channels. In actual aquaculture environments, these average values ​​can reflect a variety of water quality information: when algae density increases, chlorophyll reflection is enhanced, leading to a significant increase in the average value of the G channel; while when turbidity increases, suspended particles scatter blue light, causing the average value of the B channel to decrease. Simultaneously, the channel average values ​​can also reflect the organic matter content; for example, during red tides, the average value of the R channel will abnormally increase.

[0035] The variances of the three channels were calculated separately to obtain the variances of the green channel, blue channel, and red channel. The high variance in the green channel indicates uneven algal distribution, showing a clustered outbreak pattern; the high variance in the blue channel reflects drastic fluctuations in turbidity, such as rapid sediment deposition; and the high variance in the red channel usually suggests significant changes in organic matter concentration or abnormal aggregation of specific plankton in the water.

[0036] The green channel skewness, blue channel skewness, and red channel skewness were obtained. The green channel skewness reflects the symmetry of algal distribution; the blue channel skewness reflects the concentration trend of turbidity particle distribution; and the red channel skewness may be associated with organic matter or red tide phenomena.

[0037] Step S3: Construct a gray-level co-occurrence matrix based on the water area image to obtain multiple texture features of the gray-level co-occurrence matrix.

[0038] It should be noted that in aquaculture monitoring, in order to effectively detect abnormal fish behavior (such as aggregation degree, motion vector, etc.), texture features, including contrast, energy, homogeneity and correlation, are extracted within the defined ROI area using the gray-level co-occurrence matrix (GLCM).

[0039] In one embodiment, the water area image is converted into a grayscale image with values ​​ranging from 0 to 255. Next, grayscale quantization is performed, reducing the grayscale resolution, for example, quantizing it to 16 levels. After this preprocessing, the grayscale co-occurrence matrix (GLCM) is constructed: the distance is set to 1 pixel, focusing on adjacent pixel pairs. GLCM calculations are performed in four directions: 0°, 45°, 90°, and 135°, and the average is taken to improve robustness. Based on the GLCM, the contrast, energy, homogeneity, and correlation of the grayscale water area image are calculated and denoted as texture features. Increased contrast may indicate rapid swimming or dense collisions of fish, as sharp image texture edges and abrupt grayscale changes reflect rapid changes in the fish's spatial position; decreased contrast indicates stillness or dispersion of fish, corresponding to gradual grayscale changes, which can be used to monitor fish stress responses, such as a sudden increase in contrast when escaping due to oxygen deprivation. Increased energy indicates orderly fish movement, with uniform image texture and high consistency in movement direction and speed. Decreased energy suggests disordered fish distribution, such as foraging or panicked escape, with random image texture. This can be used to assess the effectiveness of feeding; observing whether energy stabilizes after feeding helps determine if fish behavior has returned to normal. Increased homogeneity means the fish are normally dispersed, with gradual grayscale changes and smooth texture. Decreased homogeneity is related to abnormal fish aggregation, such as fish huddling together during disease outbreaks, with localized grayscale abrupt changes in the image. This can be used for early disease warnings, and a continuous decrease in homogeneity triggers an alarm. Increased correlation indicates directional movement of the fish, with a clear directional pattern in image texture. Decreased correlation indicates random fish movement, with a lack of directional texture. This can be used to analyze the diurnal activity patterns of the fish, such as phototaxis leading to high diurnal correlation.

[0040] Step S4: After standardizing the multiple color features, multiple texture features, dissolved oxygen concentration and turbidity, the corresponding feature values ​​are obtained. The feature values ​​are then weighted and summed to obtain the risk score.

[0041] Specifically, the Pearson correlation coefficient between each feature value and its corresponding preset risk event is calculated to determine the weight of each feature value. , For the first i The weights of each feature value α The nonlinear amplification factor is a preset size, and α >1, For the first i The Pearson correlation coefficients corresponding to the eigenvalues, For the first j The Pearson correlation coefficients corresponding to the eigenvalues, i , jAll values ​​are positive integers, and n is the number of all feature values. In one embodiment, calculating the risk score includes: obtaining multiple color features, wherein the multiple color features include: the mean of the R, G, and B channels in the first moment of the water area image, the variance of the R, G, and B channels in the second moment, and the skewness of the R, G, and B channels in the third moment; obtaining multiple texture features, wherein the multiple texture features include: the contrast, energy, homogeneity, and correlation of the gray-level co-occurrence matrix; and performing a weighted summation after standardizing the mean, variance, skewness, contrast, energy, homogeneity, correlation, oxygenation concentration, and turbidity of the R, G, and B channels to obtain the risk score.

[0042] It should be noted that a higher real-time risk score indicates a higher overall risk based on the current environment and fish population status. To enhance the weighting's sensitivity to risk events, the formula employs non-linear amplification. Non-linear amplification refers to applying a non-linear amplification factor. α Apply index weighting to increase the weight of highly correlated feature indicators. α The value is greater than 1, depending on the specific scenario.

[0043] Furthermore, the present invention also includes correcting the weights of the eigenvalues: in response to a Pearson correlation coefficient corresponding to an eigenvalue being greater than a preset first threshold, the eigenvalue is determined to be an important eigenvalue; eigenvalues ​​other than important eigenvalues ​​are determined to be unimportant eigenvalues; the weights of the eigenvalues ​​are calculated. h Weights of important eigenvalues : ,in For the first h The Pearson correlation coefficients corresponding to the key eigenvalues For the first s The Pearson correlation coefficients corresponding to the key eigenvalues For a preset high proportion of weight; calculate the first q Weights of non-important eigenvalues : ,in For the first q Pearson correlation coefficients corresponding to non-significant eigenvalues For the first p Pearson correlation coefficients corresponding to non-significant eigenvalues For a preset low-proportion weight, where and .

[0044] It should be noted that, by using grouped competitive weighting, the relevant characteristic indicators corresponding to risk events are divided into an important group (important characteristic values) and an unimportant group (unimportant characteristic values), with the important group sharing a high proportion of weights. W1 (e.g., 90%), non-critical groups share the remaining weight (low percentage weight). W 2 (e.g., 10%).

[0045] In one embodiment, the invention further includes: adjusting the weights of each feature value by exponential decay at preset intervals, wherein the weights of the first feature value are adjusted by exponential decay at intervals of a predetermined number of time intervals. t The first moment i The formula for exponential decay of eigenvalues ​​is: , For the revised first t The first moment i 1 eigenvalue, For the first time before the correction t The first moment i 1 eigenvalue, For the first i The preset decay rate corresponding to each feature value T n The preset standard temperature value, T t For the first t The underwater temperature at any given time, t It can be any positive integer.

[0046] It should be noted that the decay rate increases in high-temperature environments, accelerating the replacement of old data weights and rapidly responding to the comprehensive risks caused by high temperatures. In low-temperature environments, environmental changes are slower, requiring a slower decay rate to extend data validity. The decay mechanism is designed to prevent weight drift; a preset number of values ​​is used, and exponential decay is performed at preset intervals.

[0047] Among them, in response to the t The first moment i The first characteristic value is dissolved oxygen concentration; calculate the... t The first moment i The decay rate corresponding to each characteristic value γ do The formula is:

[0048] ;

[0049] Among them in the first t time, This represents the mean of the G channel. Let G be the variance of the channel. α 1 and β 1 represents a preset adjustment coefficient. Let G be the mean of the values ​​of the G channel at all times up to time t. Let be the mean of the variances of the G channel for all times up to time t.

[0050] It should be noted that the adjustment coefficient is set based on the relationship between different algae and dissolved oxygen. This formula means that if the current... and If the dissolved oxygen weight decay rate exceeds the historical average, the weight of old data will decrease more quickly, highlighting the impact of real-time algal blooms.

[0051] Step S5: Issue an early warning based on the risk score.

[0052] It should be noted that each data collection moment (including underwater images, dissolved oxygen concentration, and turbidity) corresponds to a risk score, thereby enabling real-time monitoring of the aquaculture environment.

[0053] In one embodiment, the calculation of the current time's... i The eigenvalues ​​include: the eigenvalues ​​at the current time for the eigenvalue. i Multiple samples are taken from each feature value to obtain multiple sample values, and the mean of these multiple sample values ​​is determined as the feature value. For example, 20 samples of dissolved oxygen are taken from the underwater sample at the current moment, and the mean of the 20 dissolved oxygen samples is taken as the dissolved oxygen at the current moment.

[0054] Calculate the current time step. i Pearson correlation coefficient between each feature value and its corresponding preset risk event Including: at the current moment, the first i The eigenvalue and the eigenvalue i Each feature value corresponds to a preset risk event. m Sub-sampling, of which For the first i The eigenvalue of the th eigenvalue k Each sample value, For the first i The mean of all sampled values ​​of the feature value; at the current time, the mean of all sampled values ​​of the feature value; k The second sampling determines the first i When the preset risk event corresponding to each feature value does not occur, determine the risk event label. The value is 1, and the number of times the current time passes is 1. k The second sampling determines the first i When a preset risk event corresponding to a feature value occurs, the risk event label is determined. =0; calculate the current time step. i Pearson correlation coefficients corresponding to each eigenvalue The formula is: , The mean of the risk event labels, i.e. .

[0055] In one embodiment, at any given time, issuing an early warning based on the risk score includes: in response to the risk score being greater than or equal to a first early warning threshold and less than a second early warning threshold, increasing the frequency of underwater data collection, wherein the first early warning threshold is less than the second early warning threshold; in response to the risk score being greater than or equal to the second early warning threshold and less than a third early warning threshold, sending a control signal to the aerator to increase the aerator's operating power and sending an early warning signal to relevant personnel, wherein the second early warning threshold is less than the third early warning threshold; in response to the risk score being greater than or equal to the third early warning threshold and less than a fourth early warning threshold, sending a control signal to the aerator to make the aerator's operating power equal to its rated power and sending an early warning signal to relevant personnel, wherein the third early warning threshold is less than the fourth early warning threshold; and in response to the risk score being greater than or equal to the fourth early warning threshold, sending a fish evacuation signal to relevant personnel.

[0056] It should be noted that a risk score greater than or equal to 0.3 and less than 0.5 indicates that there are potential risk events in the aquatic environment, but they have not yet posed a significant threat. The system will automatically increase the monitoring frequency to once every 30 minutes, continuously record the fluctuation characteristics of sensor data such as dissolved oxygen and pH value, and generate detailed monitoring logs. At the same time, it will maintain a low-power operation state to ensure basic monitoring functions.

[0057] A risk score of 0.5 or higher but less than 0.7 indicates a significantly increased probability of a risk event. The system will then activate the aerator for pre-operation at 50% power to increase dissolved oxygen in the water and prevent a potential oxygen crisis. Simultaneously, the system will send an SMS alert to relevant personnel and trigger a rapid water quality testing process for a more comprehensive assessment of the water quality.

[0058] A risk score of 0.7 or higher but less than 0.9 indicates that the aquaculture environment is in a state where a high-risk event is likely to occur. The system will immediately run the aerators at full power to ensure sufficient dissolved oxygen in the water; at the same time, the feeding system in the high-risk area will be shut down to prevent improper feeding from further deteriorating the water quality. In addition, the system will trigger an audible and visual alarm to promptly notify the emergency response team so that swift countermeasures can be taken.

[0059] When the risk score is greater than or equal to 0.9, the system determines that the probability of a risk event occurring is extremely high and will automatically initiate the emergency evacuation procedure for the fish to ensure their safety; at the same time, it will switch to backup power to ensure the continuous operation of critical system equipment. In addition, the system will promptly report to regulatory authorities to coordinate resources from all parties to jointly address the crisis.

[0060] In summary, the hierarchical mechanism designed in this invention covers precise control throughout the entire lifecycle, from risk prevention to emergency response, effectively improving the scientific rigor, timeliness, and reliability of digital aquaculture environmental monitoring, and providing strong support for ensuring the healthy and stable development of aquaculture.

[0061] Figure 2 This is a schematic diagram illustrating the structure of a digital aquaculture environment monitoring system according to this embodiment.

[0062] This invention also provides a digital aquaculture environment monitoring system. For example... Figure 2 As shown, the system includes a processor and a memory, the memory storing computer program instructions, which, when executed by the processor, implement a digital aquaculture environment monitoring method according to the first aspect of the present invention.

[0063] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented using computer-readable / executable instructions that can be stored or otherwise maintained by such a computer-readable medium.

[0064] In the description of this specification, "multiple" or "several" means at least two, such as two, three or more, unless otherwise explicitly specified.

Claims

1. A digital method for monitoring aquaculture environments, characterized in that, include: The system continuously acquires underwater images, dissolved oxygen concentration, and turbidity, and segments the underwater images to obtain images of the water body area. Multiple color moments of the water body area image are obtained, including a first moment, a second moment, and a third moment, and multiple color features of each color moment are obtained. A gray-level co-occurrence matrix is ​​constructed based on the water area image to obtain multiple texture features of the gray-level co-occurrence matrix; After standardizing multiple color features, multiple texture features, dissolved oxygen concentration, and turbidity, corresponding feature values ​​are obtained. The risk score is obtained by weighted summation of each feature value. Specifically, the Pearson correlation coefficient between each feature value and its corresponding preset risk event is calculated to determine the weight of each feature value. , For the first i The weights of each feature value α The nonlinear amplification factor is a preset size, and α >1, For the first i The Pearson correlation coefficients corresponding to the eigenvalues, For the first j The Pearson correlation coefficients corresponding to the eigenvalues, i , j All are positive integers, where n is the number of all eigenvalues; At predetermined intervals, the weights of each feature value are adjusted using exponential decay, where the weights of the eigenvalues ​​are adjusted for the eigenvalues ​​at predetermined intervals. t The first moment i The formula for exponential decay of the weights of each feature value is: , For the revised first t The first moment i 1 eigenvalue, For the first time before the correction t The first moment i 1 eigenvalue, For the first i The preset decay rate corresponding to each feature value T n The preset standard temperature value, T t For the first t The underwater temperature at any given time, t It can be any positive integer; Early warnings are issued based on risk scores.

2. The digital aquaculture environment monitoring method according to claim 1, characterized in that, This also includes adjusting the weights of the eigenvalues: If the Pearson correlation coefficient corresponding to an eigenvalue is greater than a preset first threshold, the eigenvalue is determined to be an important eigenvalue. Identify eigenvalues ​​other than important eigenvalues ​​as unimportant eigenvalues. Calculate the first h Weights of important eigenvalues : ,in For the first h The Pearson correlation coefficients corresponding to the key eigenvalues For the first s The Pearson correlation coefficients corresponding to the key eigenvalues A high proportion of weights with a preset size; Calculate the first q Weights of non-important eigenvalues : ,in For the first q Pearson correlation coefficients corresponding to non-significant eigenvalues For the first p Pearson correlation coefficients corresponding to non-significant eigenvalues For a preset low-proportion weight, where and .

3. The digital aquaculture environment monitoring method according to claim 1, characterized in that, The calculation of the risk score includes: Multiple color features are obtained, including: the mean of the R, G, and B channels in the first moment of the water area image, the variance of the R, G, and B channels in the second moment, and the skewness of the R, G, and B channels in the third moment. Multiple texture features are obtained, including: contrast, energy, homogeneity, and correlation of the gray-level co-occurrence matrix; The risk score is obtained by standardizing the mean, variance, skewness, contrast, energy, homogeneity, correlation, dissolved oxygen concentration, and turbidity of the R, G, and B channels and then weighting and summing them.

4. The digital aquaculture environment monitoring method according to claim 3, characterized in that, In response to the t The first moment i The first characteristic value is dissolved oxygen concentration; calculate the... t The first moment i The decay rate corresponding to each characteristic value γ do The formula is: Among them, in the first t time, This represents the mean of the G channel. Let G be the variance of the channel. α 1 and β 1 represents a preset adjustment coefficient. For all times before time t G The mean of the channel mean, For all times before time t G The mean of the channel variance.

5. The digital aquaculture environment monitoring method according to claim 1, characterized in that, Get the current time's number i The eigenvalues ​​include: At the current moment, for the first i Multiple samples are taken from each feature value to obtain multiple sample values; the mean of the multiple sample values ​​is determined as the th value at the current time. i Each feature value.

6. The digital aquaculture environment monitoring method according to claim 5, characterized in that, Calculate the current time step. i Pearson correlation coefficient between each feature value and its corresponding preset risk event include: At the current moment, for the first i The eigenvalue and the eigenvalue i Each feature value corresponds to a preset risk event. m Sub-sampling, of which For the first i The eigenvalue of the th eigenvalue k Each sample value, For the first i The mean of all sampled values ​​of the feature value; at the current time, the mean of all sampled values ​​of the feature value; k The second sampling determines the first i When the preset risk event corresponding to each feature value does not occur, determine the risk event label. The value is 1, and the number of times the current time passes is 1. k The second sampling determines the first i When a preset risk event corresponding to a feature value occurs, the risk event label is determined. =0; calculate the current time step. i Pearson correlation coefficients corresponding to each eigenvalue The formula is: , The mean of the risk event labels, i.e. .

7. The digital aquaculture environment monitoring method according to claim 1, characterized in that, At any given time, issuing an early warning based on the risk score includes: In response to the risk score being greater than or equal to a first warning threshold and less than a second warning threshold, the frequency of underwater data collection is increased, wherein the first warning threshold is less than the second warning threshold; in response to the risk score being greater than or equal to the second warning threshold and less than a third warning threshold, a control signal is sent to the aerator to increase the aerator's operating power and a warning signal is sent to relevant personnel, wherein the second warning threshold is less than the third warning threshold; in response to the risk score being greater than or equal to the third warning threshold and less than a fourth warning threshold, a control signal is sent to the aerator to make the aerator's operating power reach its rated power and a warning signal is sent to relevant personnel, wherein the third warning threshold is less than the fourth warning threshold; in response to the risk score being greater than or equal to the fourth warning threshold, a fish evacuation signal is sent to relevant personnel.

8. The digital aquaculture environment monitoring method according to claim 7, characterized in that, The first warning threshold is 0.3, the second warning threshold is 0.5, the third warning threshold is 0.7, and the fourth warning threshold is 0.

9.

9. A digital aquaculture environment monitoring system, comprising a processor and a memory, wherein the memory stores a computer program, characterized in that, The processor executes the computer program to implement a digital aquaculture environment monitoring method as described in any one of claims 1-8.

Images