An intelligent monitoring system for aquaculture

Abstract

The present application relates to the technical field of intelligent monitoring of aquaculture, and particularly relates to an intelligent monitoring system for aquaculture. The system comprises a visual perception module, a light intensity factor acquisition module, a region division module, a region correction module and a water body abnormal region acquisition module. The visual perception module is used to acquire water surface images and corresponding water surface regions; the light intensity factor acquisition module is used to calculate a light intensity factor based on minimum standard voltage, maximum standard voltage and real-time standard voltage under outdoor light intensity; the region division module is used to analyze the water surface regions to obtain reflective regions and non-reflective regions; the region correction module is used to correct the brightness values of the regions according to the calculated gamma correction index; and the water body abnormal region acquisition module is used to acquire water body abnormal regions in the corrected water surface regions. The present application corrects the water surface regions of water surface cultivation, improves the accuracy of abnormal detection of the water surface regions, realizes real-time monitoring and eliminates the interference of dynamic objects.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent monitoring of aquaculture, and particularly relates to an intelligent monitoring system for aquaculture. BACKGROUND

[0002] In aquaculture, due to the richness of organic matter in the water and the imbalance of the nitrogen and phosphorus ratio, abnormal water quality and water color may occur, such as white turbid water, turbid water, red water, euglena water, blue-green algae and green algae. These may cause the water quality of the aquaculture water to deteriorate, and in severe cases, may deplete the oxygen in the water or produce toxins to cause the death of fish. Therefore, timely detection and real-time monitoring of the abnormal water in aquaculture is of great significance to improve the yield of aquaculture.

[0003] At present, a common method for detecting abnormalities in aquaculture water is to analyze the collected water surface image to achieve abnormal detection. However, due to the different light changes within a day and the interference of dynamic objects on the water surface, the accuracy of directly analyzing the water surface image to achieve abnormal detection is low. SUMMARY

[0004] In order to solve the above technical problems, the purpose of the present application is to provide an intelligent monitoring system for aquaculture, which comprises the following modules:

[0005] A visual perception module is configured to collect a water surface image, pre-process the water surface image to obtain an HSI image, and segment a water surface area in the HSI image.

[0006] An intensity factor acquisition module is configured to acquire a minimum standard voltage and a maximum standard voltage corresponding to outdoor light intensity, and calculate an intensity factor based on the difference between the minimum standard voltage, the maximum standard voltage, and a real-time standard voltage corresponding to real-time outdoor light intensity.

[0007] A region division module is configured to divide the luminance values corresponding to each pixel point in the water surface area to obtain an optimal luminance value segmentation threshold, and to form a reflective region from the pixel points whose luminance values are greater than the optimal luminance value segmentation threshold. The non-reflective region is obtained by subtracting the reflective region from the water surface area.

[0008] A region correction module is configured to construct a correction energy function based on the intensity factor, the local uniformity of the non-reflective region, the definition of the non-reflective region, the image information entropy of the non-reflective region, and the luminance difference of the non-reflective region before and after correction. The gamma correction index is updated based on the correction energy function, and the luminance values of the non-reflective region are corrected based on the updated gamma correction index.

[0009] The water body abnormal area acquisition module is configured to acquire similarities between each pixel point in the corrected water surface area and each pixel point in the standard water surface area, and take a pixel point with a similarity less than a preset similarity threshold as an initial growing point, and perform region growing on the initial growing point to obtain a water body abnormal area.

[0010] Preferably, the light intensity factor is calculated based on a difference between the minimum standard voltage, the maximum standard voltage and a real-time standard voltage corresponding to a real-time outdoor light intensity, and the light intensity factor comprises:

[0011] a difference between the real-time standard voltage and the minimum standard voltage is calculated as a first difference value, a difference between the maximum standard voltage and the minimum standard voltage is calculated as a second difference value, and a ratio of the first difference value to the second difference value is taken as a voltage adjustment coefficient, and a ratio of the voltage adjustment coefficient to a preset voltage threshold is taken as the light intensity factor.

[0012] Preferably, the correction energy function is constructed based on the light intensity factor, the local uniformity of the non-reflective region, the definition of the non-reflective region, the image information entropy of the non-reflective region and a brightness difference value of the non-reflective region before and after correction, and the correction energy function comprises:

[0013] the correction energy function is:

[0014]

[0015] wherein, J is the correction energy function, q is the light intensity factor, V is the brightness difference value of the non-reflective region before and after correction, I in is a brightness mean value of the non-reflective region before correction, θ is the definition, C is the local uniformity, and E is the image information entropy.

[0016] Preferably, the gamma correction index is updated based on the correction energy function, and the updating comprises:

[0017] two golden section points between a preset first end point and a preset second end point of the gamma correction index are acquired by using the golden section method, the two golden section points comprise a first golden section point and a second golden section point, and the first golden section point is less than the second golden section point.

[0018] respectively, the first correction energy function value and the second correction energy function value corresponding to the first golden section point and the second golden section point are calculated; when the first correction energy function value is greater than the second correction energy function value, the first endpoint remains unchanged, and the second endpoint is updated as the second golden section point; when the second correction energy function value is greater than the first correction energy function value, the second endpoint remains unchanged, and the first endpoint is updated as the first golden section point; the first endpoint and the second endpoint are repeatedly updated until the distance between the updated first endpoint and the second endpoint is less than or equal to the preset exponential error precision, and the gamma correction index is updated as the mean value of the first endpoint and the second endpoint.

[0019] Preferably, the brightness value of the non-reflective region is corrected based on the updated gamma correction index, including:

[0020] The calculation formula of the brightness value of the corrected non-reflective region is:

[0021]

[0022] Wherein, I is the i-th pixel point of the non-reflective region, and Y is the brightness value of the i-th pixel point of the non-reflective region. Y is the brightness value of the i-th pixel point of the non-reflective region before correction, and γ is the gamma correction index. Y is the brightness value of the i-th pixel point of the non-reflective region before correction.

[0023] Preferably, the water surface image is preprocessed to obtain an HSI image, including:

[0024] The water surface image is corrected by using a transmission transformation method to obtain a preliminary correction image; a median filter is used to filter noise of the preliminary correction image, and the filtered preliminary correction image is converted to an HSI color space to obtain an HSI image.

[0025] Preferably, the brightness value corresponding to each pixel point of the water surface region is divided to obtain an optimal brightness value segmentation threshold, including:

[0026] Based on the brightness value of the brightness channel in the HSI color space, the brightness value corresponding to each pixel point of the water surface region is divided by using the Otsu threshold method to obtain an optimal brightness value segmentation threshold.

[0027] Preferably, the pixel point with a brightness value greater than the optimal brightness value segmentation threshold constitutes a reflective region, including:

[0028] The pixel points with a brightness value greater than the optimal brightness value segmentation threshold are classified, and the corresponding pixel points in each class constitute a reflective region.

[0029] Preferably, the local uniformity and the clarity of the non-reflective region are:

[0030] The inverse difference matrix corresponding to the gray scale co-occurrence matrix of the non-reflective region is taken as the local uniformity; and the contrast corresponding to the gray scale co-occurrence matrix is taken as the definition.

[0031] Preferably, after the brightness value of the non-reflective region is corrected based on the updated gamma correction index, the method further comprises:

[0032] The brightness average value corresponding to each pixel point of the corrected non-reflective region is taken as the brightness value of all pixel points in the reflective region.

[0033] The embodiment of the present application has at least the following beneficial effects:

[0034] The embodiment of the present application relates to an intelligent monitoring scene of aquaculture. The system comprises a visual perception module, a light intensity factor acquisition module, a region division module, a region correction module and a water body abnormal region acquisition module. The visual perception module is used for collecting a water surface image and acquiring a corresponding water surface region; the light intensity factor acquisition module is used for calculating a light intensity factor based on the difference between the minimum standard voltage, the maximum standard voltage and the real-time standard voltage under outdoor light intensity; the region division module is used for analyzing the brightness value of the water surface region to obtain a reflective region and a non-reflective region; the region correction module is used for constructing a correction energy function based on the light intensity factor, the local uniformity of the non-reflective region, the definition, the image information entropy and the brightness difference before and after correction; updating the gamma correction index based on the correction energy function, and correcting the brightness value of the non-reflective region based on the gamma correction index; and the water body abnormal region acquisition module is used for acquiring a water body abnormal region in the corrected water surface region. The system takes into account that the reflective region on the water surface causes different regions of the same image to have different brightness values and that the light changes within a day cause different frames to present different brightness values, adopts an adaptive image enhancement algorithm, solves the problem of brightness imbalance between different frames, and solves the problem that a single parameter image enhancement algorithm for the same image causes the reflective region to be stretched, so that the contrast with the normal region is increased. Meanwhile, after image enhancement, the brightness value of the reflective region is compensated, so that the brightness value of the reflective region is close to the brightness value of the normal region, the precision of subsequent detection of the water surface region of aquaculture is improved, and the interference of dynamic objects is excluded. The embodiment of the present application analyzes the aspect that light changes cause different frames to present different brightness values, and then corrects the water surface region in the water surface image, improves the precision of detection of the water surface region of aquaculture, and excludes the interference of dynamic objects. BRIEF DESCRIPTION OF DRAWINGS

[0035] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0036] Figure 1 The system block diagram of the intelligent monitoring system for aquaculture provided by an embodiment of the present application. DETAILED DESCRIPTION

[0037] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined purpose, the specific implementation, structure, features and effects of the intelligent monitoring system for aquaculture according to the present application will be described in detail below in combination with the drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0038] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0039] The embodiment of the present application provides a specific implementation method of an intelligent monitoring system for aquaculture. The system is suitable for an intelligent monitoring scene of aquaculture. In the scene, each module communicates with each other through wireless communication technology and a comprehensive management and control platform, and accesses a local area network. The wireless communication module is based on Zigbee technology and is suitable for occasions with short communication distance, high power consumption requirement and low transmission data, and is used to control the transmission of information between the sending node and the receiving node of the wireless communication implementation module between the visual perception module, the light intensity factor acquisition module, the region division module, the region correction module, the water body abnormal region acquisition module and the comprehensive management and control platform. The comprehensive management and control platform integrates the water surface information and time information of aquaculture, and the real-time detection information of the aquaculture site collected by the camera, and performs water body anomaly detection and analysis. In order to solve the problem that the accuracy of direct anomaly detection and analysis of the water surface image is low due to the interference of dynamic objects on the water surface caused by changes in light. The embodiment of the present application analyzes the different brightness values of different frames caused by changes in light, and then corrects the water surface area in the water surface image, improves the precision of water surface area anomaly detection of aquaculture, and also eliminates the interference of dynamic objects.

[0040] The specific scheme of the intelligent monitoring system for aquaculture provided by the present application will be specifically described below in combination with the drawings.

[0041] Please refer to Figure 1 Fig. 1 shows a system block diagram of an intelligent monitoring system for aquaculture according to an embodiment of the present application, which comprises the following modules:

[0042] A visual perception module 10 is configured to collect water surface images, pre-process the water surface images to obtain HSI images, and segment the water surface region in the HSI images.

[0043] A wide-angle camera is used to cover the water surface region of the aquaculture to be monitored. For the video stream obtained by the wide-angle camera, the water surface image corresponding to each frame is obtained. It should be noted that the water surface image is an RGB image. Since the magnification of the edge part and the center part of the wide-angle camera is inconsistent, the obtained water surface image of the aquaculture will be distorted. Therefore, the water surface image is further pre-processed.

[0044] The pre-processed water surface image is obtained to obtain the corresponding HSI image. Specifically, a perspective transformation method is used to correct the water surface image to obtain a preliminary corrected image. Perspective transformation requires obtaining the coordinates of a set of four points of the uncorrected water surface image and a target image, respectively. The perspective transformation matrix is calculated by the two sets of coordinate points. After the perspective transformation matrix is calculated, the transformation matrix is executed on the entire water surface image to achieve image correction. The target image is an image obtained by photographing a standard image template. It should be noted that the perspective transformation matrix is confirmed by four pairs of points, which is a known technology in the art. The median filter is used to filter the noise of the preliminary corrected image, and the filtered preliminary corrected image is converted to the HSI color space to obtain the HSI image. The H, S, and I channels are hue, saturation, and brightness, respectively. The values of the filtered preliminary corrected image in the R, G, and B channels are converted to the values of the H, S, and I channels of the HSI image.

[0045] I = (R + G + B) / 3

[0046]

[0047]

[0048]

[0049] where min(R, G, B) represents the minimum value in the R, G, and B channels. The obtained HSI image is input to the comprehensive management and control platform.

[0050] Since the water body anomaly occurs in the water surface of aquaculture, including abnormal water quality and water color, it is necessary to extract the region of interest from the obtained HSI image to obtain the region range of the water surface of aquaculture, and the region where the grass, road and building on the river bank are located is not analyzed for anomaly. Since the pose of the wide-angle camera adopted is fixed, i.e. there is a fixed vision and sight distance, the region range of the water surface of aquaculture can be manually calibrated on the man-machine interaction page of the comprehensive management and control platform, i.e. for each frame of HSI image obtained, the region in the fixed region range is extracted as the water surface region.

[0051] The light intensity factor acquisition module 20 is configured to acquire the minimum standard voltage and the maximum standard voltage corresponding to the outdoor light intensity, and calculate the light intensity factor based on the difference between the minimum standard voltage, the maximum standard voltage and the real-time standard voltage corresponding to the real-time outdoor light intensity.

[0052] The light intensity of the environment can seriously affect the imaging quality of the water surface image and the subsequent water body anomaly analysis. Therefore, the light intensity in the environment can be perceived by a photosensitive resistor in a photosensitive sensor, and the perception result can be input to the comprehensive management and control platform to adjust the water surface image. The core of the photosensitive sensor is the photosensitive resistor, and its working principle is the internal photoelectric effect. Any light impact on it will cause it to release electrons. When the light intensity is higher, the current flowing through the photosensitive resistor is larger, and the voltage across the photosensitive resistor is larger. The output voltage is a standard voltage, and the value range of the standard voltage is 0-5V. The real-time standard voltage corresponding to the real-time environmental light is sent to the comprehensive management and control platform.

[0053] The real-time HSI image and the corresponding real-time standard voltage are received by the comprehensive management and control platform.

[0054] Since the change interval of the outdoor light intensity only occupies a part of the photosensitive resistor, in order to improve the characterization ability of the photosensitive resistor to the change of the outdoor light intensity, the minimum standard voltage and the maximum standard voltage corresponding to the outdoor light intensity are acquired, and the light intensity factor is calculated based on the difference between the minimum standard voltage, the maximum standard voltage and the real-time standard voltage corresponding to the real-time outdoor light intensity. Specifically, the difference between the real-time standard voltage and the minimum standard voltage is calculated as a first difference, the difference between the maximum standard voltage and the minimum standard voltage is calculated as a second difference, and the ratio of the first difference to the second difference is taken as a voltage adjustment coefficient. The ratio of the voltage adjustment coefficient to the preset voltage threshold value is the light intensity factor. In the embodiment of the present application, the value of the preset voltage threshold value is 5, and in other embodiments, the implementer can adjust the value according to the actual situation.

[0055] The calculation formula of the light intensity factor q is:

[0056] q = (Volt-Volt min ) / α(Volt max -Volt min )

[0057] wherein Volt is a real-time standard voltage; Volt min is a minimum standard voltage; Volt max is a maximum standard voltage; a is a preset voltage threshold; Volt-Volt min is a first difference; Volt max -Volt min is a second difference; (Volt-Volt min ) / (Volt max -Volt min ) is a voltage adjustment coefficient.

[0058] wherein the light intensity factor ranges from 0 to 1, and the greater the outdoor light intensity, the closer the light intensity factor is to 1; otherwise, the smaller the outdoor light intensity, the closer the light intensity factor is to 0.

[0059] The area division module 30 is configured to divide the luminance values of the pixel points in the water surface area to obtain an optimal luminance value segmentation threshold, and the reflective region is formed by the pixel points with luminance values greater than the optimal luminance value segmentation threshold; the non-reflective region is obtained by subtracting the reflective region from the water surface area.

[0060] The water surface area in the HSI image is adaptively gamma corrected based on the obtained light intensity factor, so as to enhance the image. Single gamma correction can ignore the change of light intensity with time in a day, that is, different frames of HSI images are under different environmental light intensities, and have different brightness characteristics, and are suitable for different gamma parameters. Meanwhile, considering that the brightness of the same water surface area can be uneven, for example, the reflection of the water surface can cause some regions to present a reflective phenomenon, that is, the luminance value of the reflective region presenting the reflective phenomenon is much higher than that of other regions, at this time, single gamma correction can stretch the reflective region, and the contrast with other regions is increased, and thus the HSI image enhancement effect is poor.

[0061] Therefore, the luminance values of the pixel points in the water surface area are divided to obtain an optimal luminance value segmentation threshold. Specifically, based on the luminance values of the luminance channel in the HSI color space, the luminance values of the pixel points in the water surface area are divided by using the Otsu threshold method to obtain the optimal luminance value segmentation threshold.

[0062] Pixels with a luminance value greater than the optimal luminance value segmentation threshold are classified, and corresponding pixels in each category form a reflective light region. Specifically, DBSCAN algorithm is used to cluster the reflective light region, a neighborhood radius Eps=3 is set, and another number threshold MinPts=5 is set, to obtain Q independent non-intersecting reflective light regions and a number of isolated reflective light points. For the isolated reflective light points, the average of the HSI three channels of the eight-neighborhood pixels of the isolated reflective light points is taken as the value of the HSI three channels of the isolated reflective light points. For the Q independent non-intersecting reflective light regions, the luminance values of the pixels in the reflective light regions are not corrected. The water surface region is subtracted from the reflective light region to obtain a non-reflective light region.

[0063] The region correction module 40 is configured to construct a correction energy function based on the light intensity factor, the local uniformity of the non-reflective light region, the definition of the non-reflective light region, the image information entropy of the non-reflective light region, and the luminance difference of the non-reflective light region before and after correction; update the gamma correction index based on the correction energy function, and correct the luminance value of the non-reflective light region based on the updated gamma correction index.

[0064] The correction energy function is affected by the light intensity factor q of the outdoor light intensity, the luminance difference V before and after correction, the image information entropy E, the local uniformity C, and the definition θ factor.

[0065] In the formula, the partial features of the gray level co-occurrence matrix of the non-reflective light region are taken as the local uniformity and definition of the non-reflective light region, the gray level co-occurrence matrix is used to extract the texture features of the water surface region, it is assumed that the gray level L=16, and the inverse difference moment IDM and contrast CON features of the gray level co-occurrence matrix Gr are taken as the local uniformity C and definition θ of the water surface region. The inverse difference moment IDM reflects the uniformity and homogeneity of the image texture. The smaller the inverse difference moment, the greater the local texture change of the image, the more uneven the local uniformity, and the smaller the value of the local uniformity C. The contrast CON reflects the definition of the image, and the greater the contrast, the clearer the image and the greater the θ.

[0066]

[0067]

[0068] In the formula, Gr(i,j) represents the element in the i-th row and the j-th column of the gray level co-occurrence matrix.

[0069] When the light intensity of the environment is greater, that is, the light intensity factor q is greater, the correction energy function value is greater, the image enhancement effect is better, and the difference between the luminance mean values of the non-reflective light region before and after correction is smaller, indicating that the image enhancement effect is better. The richer the information contained in the image, and the greater the corresponding information entropy E, the better the image enhancement effect. When the local texture change is greater, the contrast is greater, the definition of the image is higher, and the image enhancement effect is better.

[0070] The correction energy function J is constructed based on the local uniformity, the definition, the difference of brightness before and after correction, and the light intensity factor of outdoor light intensity.

[0071]

[0072] Wherein, q is the light intensity factor; V is the difference of brightness of non-reflective area before and after correction; I in ′ is the average brightness of non-reflective area before correction; θ is the definition; C is the local uniformity; E is the image information entropy.

[0073] In the correction energy function, the difference of brightness before and after correction is V = |I in ′-I out ′|, wherein I in ′ is the average brightness of non-reflective area before correction, and I out ′ is the average brightness of non-reflective area after correction; the image information entropy is Wherein, p(i) represents the proportion of the number of pixels with the gray level i to the total number of pixels, since the gamma correction is not a linear transformation, the gray level may not be an integer.

[0074] Further, the gamma correction index is updated based on the correction energy function. Specifically, two golden section points between a preset first end point and a preset second end point of the gamma correction index are obtained by using the golden section method, and the golden section points include a first golden section point and a second golden section point; wherein the first golden section point is smaller than the second golden section point. In the embodiment of the present application, the value of the preset first end point is 0.2, and the value of the preset second end point is 0.6, which can be adjusted according to actual conditions in other embodiments. The first correction energy function value and the second correction energy function value corresponding to the first golden section point and the second golden section point are calculated respectively; when the first correction energy function value is greater than the second correction energy function value, the first end point remains unchanged, and the second end point is updated to the second golden section point; when the second correction energy function value is greater than the first correction energy function value, the second end point remains unchanged, and the first end point is updated to the first golden section point; the first end point and the second end point are repeatedly updated until the distance between the updated first end point and the second end point is less than or equal to the preset index error accuracy, and the gamma correction index is updated to the average of the first end point and the second end point. In the embodiment of the present application, the value of the preset index error accuracy is 10 -3 , which can be adjusted according to actual conditions in other embodiments.

[0075] That is, since the more appropriate gamma correction index value is between 0.2 and 0.6, the golden section algorithm is used to make the correction energy function take the maximum value, and the accuracy of the index error is 10 -3, the first end point of the initial gamma correction index is set as a1=0.2 and the second end point is set as a2=0.6, the first correction energy function value J1 and the second correction energy function value J2 of two golden section points w1 and w2 between the two end points are obtained by using the golden section algorithm, when the first correction energy function value J1 is greater than the second correction energy function value J2, the first end point a1 remains unchanged, and the second end point a2 is updated as the second golden section point; when the second correction energy function value J2 is greater than the first correction energy function value J1, the second end point a2 remains unchanged, and the first end point a1 is updated as the first golden section point. Repeat the step until the distance between the first end point and the second end point is less than or equal to the precision of the index error, that is, |a2-a1|≤10 -3 It should be noted that, since the first golden section point and the second golden section point are different, the first correction energy function value and the second correction energy function value obtained are also different.

[0076] Further, based on the updated gamma correction index, the brightness value of the non-reflective region is corrected as follows: the average brightness value of the non-reflective region is obtained, and the corrected brightness value of the non-reflective region is obtained.

[0077] The calculation formula of the corrected brightness value of the non-reflective region is:

[0078]

[0079] wherein, I out is the corrected brightness value of the i-th pixel point of the non-reflective region; γ is the gamma correction index, I ini is the brightness value of the i-th pixel point of the non-reflective region before correction.

[0080] That is, the brightness value of the non-reflective region is corrected by using the optimal gamma index, the input is the brightness value of each pixel point of the non-reflective region, and the output is the brightness value of each pixel point of the non-reflective region after image enhancement, that is, the brightness value of each pixel point of the non-reflective region after correction, so as to realize the brightness value adjustment of the non-reflective region. Then the corrected brightness average value of each pixel point of the non-reflective region is taken as the brightness value of all pixel points in the reflective region.

[0081] The corrected reflective region and the non-reflective region constitute the corrected water surface region.

[0082] The water body abnormal region acquisition module 50 is used for acquiring the similarity of each pixel point of the corrected water surface region and each pixel point of the standard water surface region, and the pixel point with a similarity less than a preset similarity threshold is taken as an initial growth point. The initial growth point is regionally grown to obtain a water body abnormal region. ​

[0083] For the corrected water surface area, abnormality detection is performed, which can be performed according to the values of the three channels of the standard water surface area as the HSI parameter vector hsi N of three element values to determine whether the water body is abnormal. The HSI parameter vector hsi of each pixel point in the corrected water surface area is calculated.

[0084] Based on the HSI parameter vector hsi of each pixel point in the corrected water surface area and the standard water body parameter vector hsi N , the structural similarity sim is calculated. Since the water body anomaly mainly reflects abnormal water quality and water color, that is, the anomaly mainly reflects the difference in the value of the hue channel H, and further, the difference in the first element of the HSI parameter vector hsi, different weights are respectively assigned to the hue channel, the saturation channel and the brightness channel. The hue channel is taken as the first channel in the HSI color space, the saturation channel is taken as the second channel in the HSI color space, and the brightness channel is taken as the third channel in the HSI color space.

[0085] Wherein, the similarity sim of the i-th pixel point in the corrected water surface area is calculated according to the following formula: i

[0086]

[0087] Wherein, hsi(1,p) is the value of the p-th channel in the HSI color space of the i-th pixel point in the corrected water surface area; a p is the weight of the p-th channel; hsi N (1,p) is the value of the p-th channel in the HSI color space of the i-th pixel point in the standard water surface area; and p is the number of channels in the HSI color space.

[0088] In the embodiment of the present application, the weight a1 of the first channel is 0.8, the weight a2 of the second channel is 0.1, and the weight a3 of the third channel is 0.1. In other embodiments, the implementer can adjust the values according to the actual situation.

[0089] ​The pixel points with similarity less than a preset similarity threshold value are taken as initial growth points of the candidate water body abnormal area, growth rules are set, and region growing is performed on the initial growth points to obtain the water body abnormal area.The growth rules are specifically: if the similarity between a growth point and the pixel points in its eight adjacent domains is greater than 0.8, it is considered that the growth rules are met, the pixel point is added to the selected water body abnormal area, and the newly added pixel point is taken as a new growth point to determine whether there is a new pixel point in the eight adjacent domains of the new growth point that meets the growth rules.After the growth is completed, if the number of pixel points contained in the growth region is greater than a preset abnormal threshold value, the regions corresponding to the multiple pixel points are marked as water body abnormal regions; otherwise, when the number of pixel points contained in the growth region is less than or equal to the preset abnormal threshold value, the region is taken as a noise region and is not processed.In the embodiment of the present application, the preset similarity threshold value is 0.75, and the preset abnormal threshold value is 20, and in other embodiments, the implementer can adjust the value according to the actual situation.

[0090] The clock module of the comprehensive management and control platform performs timing operation on the detected water body abnormal area, and if the similarity between the water body abnormal area and each frame after it within five minutes is less than 0.9, it is considered that the water body abnormal area is a dynamic object region, such as a water bird appearing on the water surface, and is not processed. The final obtained water body abnormal area is warned by the comprehensive management and control platform, so that subsequent professional personnel can process the water body abnormal area, and intelligent monitoring of aquaculture is realized.

[0091] In summary, the embodiment of the present application relates to an intelligent monitoring scene of aquaculture. The system includes a visual perception module, a light intensity factor acquisition module, a region division module, a region correction module and a water body abnormal area acquisition module. The visual perception module is used to collect water surface images and obtain corresponding water surface regions; the light intensity factor acquisition module is used to calculate the light intensity factor based on the difference between the minimum standard voltage, the maximum standard voltage and the real-time standard voltage under outdoor light intensity; the region division module is used to analyze the brightness values of the water surface regions to obtain reflective regions and non-reflective regions; the region correction module is used to construct a correction energy function based on the light intensity factor, the local uniformity of the non-reflective region, the definition, the image information entropy and the brightness difference before and after correction; update the gamma correction index based on the correction energy function, and correct the brightness values of the non-reflective region based on the gamma correction index; the water body abnormal area acquisition module is used to obtain the water body abnormal area in the corrected water surface region. The embodiment of the present application improves the precision of water surface region abnormality detection of aquaculture by correcting the water surface regions in the water surface images, and also eliminates the interference of dynamic objects.

[0092] It is to be noted that the above-mentioned embodiments illustrate rather than limit the application, and that those skilled in the art will be able to design many alternative embodiments without departing from the scope of the appended claims. The word "comprising" does not exclude the presence of elements or steps other than those listed in a claim. In a claim, the word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. It is further noted that claims may

[0093] The various embodiments in the specification are described in progressive manner, and the same or similar parts between the embodiments can be mutually referred to, and each embodiment focuses on the difference from other embodiments.

[0094] The above descriptions are merely possible embodiments of the present application, and are not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. An intelligent monitoring system for aquaculture, characterized in that, The system includes the following modules: The visual perception module is used to acquire water surface images, preprocess the water surface images to obtain HSI images, and segment the water surface region in the HSI images; The light intensity factor acquisition module is used to acquire the minimum standard voltage and the maximum standard voltage corresponding to the outdoor light intensity, and to calculate the light intensity factor based on the difference between the minimum standard voltage, the maximum standard voltage and the real-time standard voltage corresponding to the real-time outdoor light intensity. The region segmentation module is used to divide the brightness values ​​corresponding to each pixel in the water surface region to obtain an optimal brightness value segmentation threshold. Pixels with brightness values ​​greater than the optimal brightness value segmentation threshold constitute a reflective region. The non-reflective region is obtained by subtracting the reflective region from the water surface region. The region correction module is used to construct a correction energy function based on the light intensity factor, the local uniformity of the non-reflective region, the sharpness of the non-reflective region, the image information entropy of the non-reflective region, and the brightness difference of the non-reflective region before and after correction; update the gamma correction index based on the correction energy function; and correct the brightness value of the non-reflective region based on the updated gamma correction index. The water body anomaly region acquisition module is used to acquire the similarity between each pixel in the corrected water surface region and each pixel in the standard water surface region. Pixels with similarity less than a preset similarity threshold are used as initial growth points. Region growth is performed on the initial growth points to obtain the water body anomaly region. The method of constructing a correction energy function based on the light intensity factor, the local uniformity of the non-reflective region, the sharpness of the non-reflective region, the image information entropy of the non-reflective region, and the brightness difference of the non-reflective region before and after correction includes: The correction energy function is: , Wherein, J is the correction energy function; q is the light intensity factor; V is the brightness difference of the non-reflective area before and after correction; I in ' represents the average brightness of the non-reflective area before correction; ϑ represents the sharpness; C represents the local uniformity; E represents the image information entropy; The updating of the gamma correction exponent based on the correction energy function includes: The golden section method is used to obtain two golden section points between a preset first endpoint and a preset second endpoint of the gamma correction index. The golden section points include a first golden section point and a second golden section point; wherein, the first golden section point is smaller than the second golden section point. Calculate the first and second corrected energy function values ​​corresponding to the first and second golden section points, respectively. When the first corrected energy function value is greater than the second corrected energy function value, the first endpoint remains unchanged, and the second endpoint is updated to the second golden section point. When the second corrected energy function value is greater than the first corrected energy function value, the second endpoint remains unchanged, and the first endpoint is updated to the first golden section point. Repeat the updating of the first and second endpoints until the distance between the updated first and second endpoints is less than or equal to the preset exponential error accuracy, then update the gamma correction exponent to the mean of the first and second endpoints.

2. The intelligent monitoring system for aquaculture according to claim 1, characterized in that, The calculation of the light intensity factor based on the difference between the minimum standard voltage, the maximum standard voltage, and the real-time standard voltage corresponding to the real-time outdoor light intensity includes: The difference between the real-time standard voltage and the minimum standard voltage is calculated as a first difference; the difference between the maximum standard voltage and the minimum standard voltage is calculated as a second difference; the ratio of the first difference and the second difference is used as a voltage adjustment coefficient, and the ratio of the voltage adjustment coefficient to a preset voltage threshold is used as the light intensity factor.

3. The intelligent monitoring system for aquaculture according to claim 1, characterized in that, The correction of the brightness value of the non-reflective area based on the updated gamma correction index includes: The formula for calculating the brightness value of the corrected non-reflective area is: , Among them, I outi γ is the brightness value of the i-th pixel in the corrected non-reflective area; γ is the gamma correction index, I ini This is the brightness value of the i-th pixel in the non-reflective area before correction.

4. The intelligent monitoring system for aquaculture according to claim 1, characterized in that, The preprocessing of the water surface image to obtain the HSI image includes: The water surface image is corrected using the transmission transformation method to obtain a preliminary corrected image; noise is filtered out from the preliminary corrected image using a median filter, and the filtered preliminary corrected image is converted to the HSI color space to obtain an HSI image.

5. The intelligent monitoring system for aquaculture according to claim 1, characterized in that, The step of dividing the brightness values ​​corresponding to each pixel in the water surface area to obtain the optimal brightness value segmentation threshold includes: Based on the brightness values ​​of the brightness channel in the HSI color space, the brightness values ​​corresponding to each pixel in the water surface area are divided using the Otsu thresholding method to obtain the optimal brightness value segmentation threshold.

6. The intelligent monitoring system for aquaculture according to claim 1, characterized in that, The reflective area is composed of pixels whose brightness values ​​are greater than the optimal brightness value segmentation threshold, including: Pixels with brightness values ​​greater than the optimal brightness value segmentation threshold are classified, and the corresponding pixels in each category constitute a reflective area.

7. The intelligent monitoring system for aquaculture according to claim 1, characterized in that, The local uniformity and sharpness of the non-reflective areas are as follows: The inverse difference moment corresponding to the gray-level co-occurrence matrix of the non-reflective region is taken as the local uniformity; the contrast corresponding to the gray-level co-occurrence matrix is ​​taken as the sharpness.

8. The intelligent monitoring system for aquaculture according to claim 1, characterized in that, After correcting the brightness value of the non-reflective area based on the updated gamma correction index, the method further includes: The average brightness of each pixel in the corrected non-reflective area is used as the brightness value of all pixels in the reflective area.

Images