Breeding abalone border crossing behavior identification method based on dynamic subtraction statistical characteristics

Abstract

The invention discloses a parent abalone border crossing behavior recognition method based on dynamic subtraction statistical characteristics, and belongs to the field of image recognition, and the method comprises the following steps: carrying out the blocking and graying processing of a wall-climbing video without parent abalone, and obtaining a video average gray value; based on a sampling rate, extracting a sampling frame of the abalone-free wall-climbing video, and calculating a difference value between a gray value of the sampling frame and the average gray value of the video to obtain a difference value sequence; performing pairwise comparison on the sampling frames to obtain an inter-frame difference matrix; and respectively drawing the difference value sequence and the inter-frame difference value matrix into difference value curves, and determining the abalone border crossing behavior based on the difference value curves. According to the method, challenges caused by background and illumination changes can be effectively coped with through combination of segmentation subtraction and statistical property analysis, and the detection precision and robustness are greatly improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of image recognition, and particularly relates to a method for identifying the overstepping behavior of breeding abalones based on the statistical characteristics of dynamic subtraction. Background Art

[0002] In the field of aquaculture, abalones, as aquatic products with high economic value, their reproductive and spawning behaviors have crucial impacts on aquaculture efficiency and economic benefits. Accurately monitoring their spawning behaviors is the key to achieving scientific aquaculture. Through early warning and monitoring of the overstepping behavior of breeding abalones during spawning, it is possible to effectively prevent abalones from spawning prematurely, thereby avoiding the decline in the quality of seedlings caused by incomplete gonadal maturity or unsuitable environmental conditions. This measure is crucial in the breeding of abalones because the spawning time and state of breeding abalones directly determine the development, survival rate, and subsequent growth performance of seedlings.

[0003] When modern object detection methods are applied to the field of aquaculture, although they have a relatively high level of intelligence, there are also various limitations. First of all, they are sensitive to environmental noise, such as interference from underwater suspended particles, bubbles, and light refraction, which easily lead to false detections or missed detections. In high-density aquaculture, the closeness or overlap of abalones also makes it difficult for the algorithm to distinguish individuals, affecting the accuracy of detection. In addition, these methods rely on large-scale labeled data and high-performance hardware, and the cost of training and updating models is relatively high. Especially in diverse environments, they need to be frequently adjusted to adapt to new scenarios. Deep learning models usually lack interpretability and have limited ability to identify rare abnormal behaviors, making it difficult to cope with dynamic and complex scenarios. More importantly, their high cost and technical complexity have limited applicability to small and medium-sized farms, and they also lack the ability to control fine boundary detection and multi-object separation. These drawbacks pose many challenges to the application of modern object detection methods in complex aquaculture scenarios.

[0004] To overcome the above problems, the present invention proposes a method for identifying the overstepping behavior of breeding abalones based on the statistical characteristics of dynamic subtraction. Summary of the Invention

[0005] To solve the above technical problems, the present invention proposes a method for identifying the overstepping behavior of breeding abalones based on the statistical characteristics of dynamic subtraction to solve the problems existing in the above prior art.

[0006] To achieve the above object, the present invention provides a method for identifying the overstepping behavior of breeding abalones based on the statistical characteristics of dynamic subtraction, including the following steps:

[0007] Perform block division and grayscale processing on the video without breeding abalones climbing on the wall to obtain the average grayscale value of the video;

[0008] Extract the sampled frames of the seedless abalone climbing wall video based on the sampling rate, and calculate the difference between the grayscale value of the sampled frame and the average grayscale value of the video to obtain a difference sequence;

[0009] Compare the sampled frames pairwise to obtain an inter-frame difference matrix;

[0010] Plot the difference sequence and the inter-frame difference matrix as difference curves respectively, and determine the overboundary behavior of the abalone based on the difference curves.

[0011] Optionally, the process of performing block division and grayscale processing on the seedless abalone climbing wall video to obtain the average grayscale value of the video includes:

[0012] Divide the video obtained from the seedless abalone climbing wall video into block videos;

[0013] Perform grayscale processing on the block videos and calculate the pixel grayscale values of each frame image in the block videos;

[0014] Calculate the average grayscale value of each frame image based on the pixel grayscale values of several said frame images;

[0015] Obtain the average grayscale value of the video based on the average grayscale values of several frame images.

[0016] Optionally, the calculation expression of the average grayscale value of the video is:

[0017]

[0018] In the formula, represents the average grayscale value of the video, n represents the total number of video frames, M×N represents the matrix obtained after grayscaling the current frame, k represents the kth frame of the video, and i and j respectively represent the row and column of the pixel in the image, Y ij represents the grayscale value of each pixel of each frame image.

[0019] Optionally, the process of comparing the sampled frames pairwise to obtain an inter-frame difference matrix includes:

[0020] Divide the sampled frames pairwise to obtain several pairs of video frames;

[0021] Calculate the grayscale difference between each pair of video frames to obtain the average grayscale difference;

[0022] Construct a symmetric matrix based on the average grayscale differences corresponding to several pairs of video frames to obtain an inter-frame difference matrix.

[0023] Optionally, the calculation expression of the average grayscale difference is:

[0024]

[0025] In the formula, Dij represents the average grayscale difference between two frames, F i (x, y) represents the grayscale value of frame i at position (x, y), F j (x, y) represents the grayscale value of frame j at position (x, y).

[0026] Optionally, the process of plotting the difference sequence as a difference curve to determine the out-of-bounds behavior of abalones includes:

[0027] Plot the difference sequence as a difference curve to obtain the first difference curve;

[0028] Calculate the first derivative of the first difference curve to obtain the curvature change rate;

[0029] If the curvature change rate at the current moment is higher than the average level, it is determined that an out-of-bounds behavior of abalones occurs.

[0030] Optionally, the process of plotting the inter-frame difference matrix as a difference curve to determine the out-of-bounds behavior of abalones includes:

[0031] Plot the inter-frame difference matrix as a difference curve to obtain the second difference curve;

[0032] Statistically calculate the mean and standard deviation of the second difference curve to set a dynamic threshold;

[0033] Traverse the second difference curve. When a point exceeding the dynamic threshold appears in the second difference curve, it is determined as an abnormal point;

[0034] Correspond the abnormal point with the time information of the boundary area. When the abnormal point occurs during the time period of the boundary area, it is determined that an out-of-bounds behavior of abalones occurs.

[0035] Compared with the prior art, the present invention has the following advantages and technical effects:

[0036] A method for identifying the out-of-bounds behavior of abalones based on the statistical characteristics of dynamic subtraction in the present invention realizes the efficient identification of the out-of-bounds behavior of abalones through precise grayscale processing and difference calculation. First, the video without abalones climbing on the wall is divided into blocks and grayscaled to obtain the average grayscale value of the video, providing a benchmark for subsequent analysis. Then, sampling frames are extracted based on the sampling rate, and the difference between their grayscale values and the average grayscale value of the video is calculated to form a difference sequence, which effectively captures the basic changes between video frames. Further, pairwise comparison of the sampling frames is performed to obtain an inter-frame difference matrix, deeply exploring the subtle dynamic differences between frames. Finally, the difference sequence and the inter-frame difference matrix are plotted as difference curves, and by analyzing the fluctuations and change trends of the curves, the out-of-bounds behavior of abalones is accurately determined. This method not only improves the accuracy of identification but also enhances the adaptability to complex backgrounds and lighting changes, providing strong technical support for the management of abalones in aquaculture and having important practical application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The drawings forming a part of this application are used to provide a further understanding of this application. The schematic embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation of this application. In the drawings:

[0038] Figure 1 It is a flowchart of a method and application for early warning of abalone spawning based on image recognition in an embodiment of the present invention;

[0039] Figure 2 It is the calculation steps of the difference sequence in an embodiment of the present invention;

[0040] Figure 3 It is a simulation diagram of inter-frame difference calculation in an embodiment of the present invention;

[0041] Figure 4 It is the grayscale difference between the sampling frame and the average grayscale value of the video in an embodiment of the present invention;

[0042] Figure 5 It is the image effect after dynamic subtraction processing in an embodiment of the present invention;

[0043] Figure 6 It is the average grayscale value difference between sampling frames in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0044] It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other. The following will refer to the drawings and combine the embodiments to detail this application.

[0045] It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. And although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0046] Embodiment 1

[0047] As Figure 1 shown, in this embodiment, a method for identifying the over-border behavior of abalones based on the statistical characteristics of dynamic subtraction is provided. This method performs excellently in dealing with complex backgrounds and dynamic environments. By combining segmentation subtraction and statistical characteristic analysis, this method can effectively cope with the challenges brought by background and illumination changes, and greatly improve the detection accuracy and robustness. It includes the following steps: dividing the video of abalones climbing on the wall into blocks and performing grayscale processing to obtain the average grayscale value of the video; extracting the sampling frames of the video of abalones climbing on the wall based on the sampling rate, and calculating the difference between the grayscale value of the sampling frame and the average grayscale value of the video to obtain a difference sequence; comparing the sampling frames pairwise to obtain an inter-frame difference matrix; plotting the difference sequence and the inter-frame difference matrix into difference curves respectively, and determining the over-border behavior of abalones based on the difference curves.

[0048] Step 1: First, after dividing the video of abalones climbing on the wall into blocks, take a part for grayscale processing to obtain several segmented videos, and calculate the average grayscale value of the entire video based on the several segmented videos. Take this average grayscale value as the standard, denoted as The calculation steps of the difference sequence are as Figure 2 shown.

[0049] ① Each frame of the video image consists of m×n pixels, and each pixel is represented by three color channels: red R, green G, and blue B. The grayscale value Y of each pixel in each frame of the image ij can be calculated by the following weighted average formula:

[0050] Y ij =0.299×R ij +0.587×G ij +0.114×B ij

[0051] where i and j respectively represent the row and column of the pixel in the image.

[0052] ② For a certain frame k of the video, its grayscaled frame is an M×N matrix. The average grayscale value Y of this frame k can be expressed as:

[0053]

[0054] where, Denotes the grayscale value at position (i, j) in the k-th frame.

[0055] ③ Suppose the video consists of n frames. Then the average grayscale value of the entire video can be expressed as Y:

[0056]

[0057] The average grayscale value of the entire video is the sum of the grayscale values of all pixels in all frames, divided by the total number of all pixels in the video. Each video frame is grayscaled, and the grayscale values of each of its pixels are accumulated, and finally the average value is taken.

[0058] Secondly, in this embodiment, specific frames are extracted from the video according to a set sampling rate, and the difference between the grayscale values of these frames and the above-mentioned average grayscale value Y of the video is calculated. This difference is used to analyze the changes between frames:

[0059]

[0060] where ΔY m is the absolute difference between the average grayscale value of the m-th sampled frame and the "standard".

[0061] The sampling interval is where FPS is the frame rate of the video and s is the number of sampled frames per second; thus obtaining a sequence of sampled differences, which lays the foundation for the statistical analysis of abalone behavior in step 3 below.

[0062] Step 2: Compare the sampled frames of the video pairwise, and calculate the grayscale difference between each pair of frames. Specifically, the difference between the two grayscale frames is quantified by calculating the absolute difference. These difference values are stored in a symmetric matrix, and finally the upper triangular part is extracted to analyze the changes between frames, as Figure 3 shown.

[0063] ① Suppose the grayscale images corresponding to the i-th frame and the j-th frame in the video are F i and Fj respectively, and each element D ij in the frame difference matrix represents the average grayscale difference between the two frames:

[0064]

[0065] where M and N are the width and height of the frame respectively; F i (x, y) represents the grayscale value of frame i at position (x, y).

[0066] ② The frame difference matrix D is a symmetric matrix, and only its upper triangular part needs to be extracted for statistical analysis.

[0067]

[0068] The matrix size is n×n, and the upper triangular part contains Elements are collected into a one-dimensional array by traversing the upper triangular matrix row by row and column by column.

[0069] Step 3: Quantify and describe the dynamic characteristics of the video content by performing statistical analysis on the grayscale differences between frames, thereby revealing the differences between different videos or the change patterns within a specific video.

[0070] ① The difference sequence ΔY obtained in step 1 above m Represents the grayscale difference of each sampling frame. In this embodiment, these differences are plotted into a difference curve to obtain a first difference curve. This curve will show the trend of grayscale changes between frames in the video, and can intuitively reflect the dynamic changes in the video; further statistical analysis can be performed to observe the rate of change of the curve by calculating the first-order derivative of the first difference curve. If the derivative value at a certain moment is significantly higher than the average level or the value of the surrounding points, it means that a more drastic change has occurred at this moment. That is, when this type of abalone cross-border behavior occurs, the first-order derivative at a certain moment will be significantly higher than the average level.

[0071] ② The one-dimensional array of inter-frame differences obtained in the above step 2 is plotted into a difference curve to obtain the second difference curve. This curve reflects the distribution of grayscale differences between different frames, and the trend of grayscale changes between frames can be intuitively seen. Statistical analysis is performed on the difference curve to represent the grayscale difference between frames. The dynamic threshold is set by statistically analyzing the overall mean and standard deviation of the difference curve, and the threshold is set based on the statistical results to determine the abalone wall. The dynamic threshold formula is: T = μ + k·σ, where k is the amplification factor; this threshold dynamically adapts to the overall level and fluctuation of the difference curve, and there is no need to set a fixed value. Traverse the difference curve and find the point that exceeds the dynamic threshold T, that is: ΔG i >T, it is determined as an abnormal point. These abnormal points are matched with the time information of the boundary area. If the abnormal point occurs in the time period of the boundary area, it can be determined as abalone crossing the boundary.

[0072] As a specific implementation of this embodiment, the present invention uses the common subtraction method, YOLOv8 and the abalone spawning warning method based on image recognition to experiment with abalone crossing the boundary behavior. The comparison results of the abalone crossing the boundary behavior recognition method based on dynamic subtraction statistical characteristics and the existing target detection performance are shown in Table 1.

[0073] Table 1

[0074]

[0075] Table 1 describes the comparison of the abalone over - boundary behavior recognition method based on the statistical characteristics of dynamic subtraction proposed in this embodiment with the ordinary subtraction method and YOLOv8. The results show that the abalone spawning early - warning method based on image recognition is superior to other methods in terms of accuracy, recall rate, F1 - score, and false - alarm rate in the detection of abalone wall - attaching spawning, and is more suitable for practical applications.

[0076] Through the statistical analysis of the difference curve by the method proposed in the present invention, first, the video without abalone climbing on the wall is segmented, and a part of the video frames is selected for grayscale conversion, and the average grayscale value of the entire video is calculated, which is 109.48. Then, the grayscale value of the sampled frames is analyzed for differences with the overall average grayscale value of the video to obtain the difference curve. As Figure 4 shown, it can be observed that at the critical moments when the abalone enters or leaves the detection area, significant changes occur in the difference curve. This change is mainly manifested as when the abalone enters the detection area (i.e., the over - boundary behavior occurs), the difference between the grayscale value of the sampled frame and the overall average grayscale value will increase sharply, and the interpolation is particularly obvious; when the abalone crawls away from the detection area, the difference will also show significant fluctuations.

[0077] The significance of this difference can be reflected by the morphological changes of the difference curve, thus realizing the accurate detection and judgment of the abalone over - boundary behavior. More specifically, the difference curve will show large fluctuations and drastic changes at the boundary of the detection area. Through the statistical analysis of the difference curve, the regular characteristics of the grayscale change in the video can be further quantified. At the same time, this method can reveal the entire dynamic process of the abalone from entering the detection area to crawling away from the detection area through the changes of each key point in the difference curve, providing a reliable theoretical basis and statistical support for judging the abalone over - boundary behavior.

[0078] As Figure 5 shown, it shows the image effects after dynamic subtraction processing of abalones in the case of over - boundary behavior and non - over - boundary behavior. These processed images intuitively reflect the different characteristics of abalone behavior. When the abalone has an over - boundary behavior, its dynamic subtraction effect is obvious, which can prominently display the movement trajectory and changes of the abalone in the detection area. When there is no over - boundary behavior, the dynamic subtraction image is relatively stable and shows almost no significant changes. This dynamic subtraction method effectively extracts the abalone movement characteristics and provides strong visual support for the rapid detection and differentiation of over - boundary behavior.

[0079] By pairwise comparing the sampled frames of the abalone climbing wall video, the gray - scale difference between each pair of frames can be calculated, thus enabling the precise detection and analysis of the abalone's out - of - bounds behavior. Specifically, this process is achieved by calculating the absolute difference of the corresponding pixel points of two gray - scale images, that is, by calculating the difference between two gray - scale frames pixel by pixel and taking the absolute value, a gray - scale difference matrix between frames can be obtained. These inter - frame difference values are stored in a symmetric matrix, where each element D ij represents the average gray - scale difference between the sampled frames i and j. Since the matrix is symmetric, only the upper triangular part needs to be extracted for subsequent statistical analysis.

[0080] In specific operations, this embodiment conducts dynamic analysis of the difference sequence for the extracted upper triangular part data. By comparing the changes of these differences over time, the regularity of the gray - scale differences between video frames can be observed, as Figure 6 shown. When the abalone exhibits out - of - bounds behavior, the inter - frame difference sequence will show significant changes, manifested as a sudden increase or violent fluctuation in the difference amplitude. This phenomenon is mainly attributed to the significant motion characteristics of the abalone when entering or leaving the detection area, resulting in a sudden increase in the change of inter - frame gray - scale values.

[0081] Furthermore, the analysis of the inter - frame gray - scale difference matrix can reflect the dynamic change trend of the video content, providing an efficient and reliable quantification method for detecting the abalone's out - of - bounds behavior. Through the structured storage and dynamic analysis of the difference matrix, the critical change points of the inter - frame gray - scale differences can be accurately captured, and these change points exactly correspond to the key behavioral moments when the abalone enters or leaves the detection area. This analysis method not only improves the accuracy of out - of - bounds behavior detection but also provides a highly versatile solution idea for capturing target behaviors in dynamic scenarios.

[0082] Based on the above experimental results, the accuracy rate of the abalone spawning early - warning method using the dynamic subtraction statistical characteristics proposed in the present invention reaches 87.23%. Through the comparative analysis of different gray - scale differences, it provides a reliable theoretical basis and practical support for the detection and recognition of the abalone's out - of - bounds spawning behavior. By analyzing the difference between the gray - scale value of the sampled frame and the overall average gray - scale value of the video, the key behavioral characteristics of the abalone entering or leaving the detection area out - of - bounds can be accurately captured; and through the dynamic statistical analysis of the inter - frame gray - scale difference matrix, the regularity of the inter - frame changes in the video can be further revealed, especially the significant fluctuations in the inter - frame gray - scale differences when the abalone's out - of - bounds spawning behavior occurs. The method of this patent provides an efficient, reliable and universal technical solution for abalone spawning early - warning.

[0083] The analysis method of the present invention based on inter - frame gray - scale differences does not rely on an object detection framework, but directly identifies the behavior of the abalone through the gray - scale changes between video frames. This makes the method simpler, more flexible, and has lower hardware requirements.

[0084] By calculating the mean and standard deviation of the difference curve and setting a dynamic threshold, the present invention can adaptively adjust the sensitivity to different video contents. In the case of complex backgrounds or changing lighting, the dynamic threshold can be automatically adjusted to adapt to the different dynamic changes of the video and reduce false positives. Compared with traditional methods based on background modeling or static thresholds, the dynamic threshold can better handle the gray-scale difference changes in different environments, improving the accuracy and robustness of detection.

[0085] The above are only the preferred specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed in the present application should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for identifying cross-border behavior of abalone based on dynamic subtraction statistical characteristics, characterized in that: The following steps are involved: The video of the abalone climbing the wall is divided into blocks and grayed to obtain the average gray value of the video; Extracting the sampling frames of the abalone wall climbing video based on the sampling rate, and calculating the difference between the grayscale value of the sampling frame and the average grayscale value of the video to obtain a difference sequence; Comparing the sampling frames in pairs to obtain an inter-frame difference matrix; The difference sequence and the inter-frame difference matrix are respectively plotted into difference curves, and the abalone crossing behavior is determined based on the difference curves.

2. The method for identifying abalone crossing-border behavior based on dynamic subtraction statistical characteristics according to claim 1 is characterized in that: The process of dividing and graying the video of the abalone climbing the wall to obtain the average gray value of the video includes: The video of the abalone climbing the wall is divided into blocks to obtain block videos; Gray-scaling the segmented video and calculating the pixel gray value of each frame image in the segmented video; Calculating an average grayscale value of each frame of image based on a plurality of pixel grayscale values ​​of each frame of image; The average grayscale value of the video is obtained based on the average grayscale values ​​of several frames of images.

3. The method for identifying abalone crossing-border behavior based on dynamic subtraction statistical characteristics according to claim 2 is characterized in that: The calculation expression of the average gray value of the video is: In the formula, represents the average grayscale value of the video, n represents the total number of video frames, M×N represents the matrix obtained by graying the current frame, k represents the kth frame of the video, i and j represent the rows and columns of pixels in the image respectively, and Y ij Represents the grayscale value of each pixel in each frame of the image.

4. The method for identifying abalone crossing-border behavior based on dynamic subtraction statistical characteristics according to claim 3 is characterized in that: The process of performing pairwise comparison on the sampling frames to obtain an inter-frame difference matrix includes: Dividing the sample frames into pairs to obtain a plurality of pairs of video frames; Calculate the grayscale difference between each pair of video frames to obtain the grayscale difference mean; A symmetric matrix is ​​constructed based on the mean grayscale differences corresponding to several pairs of video frames to obtain an inter-frame difference matrix.

5. The method for identifying abalone crossing-border behavior based on dynamic subtraction statistical characteristics according to claim 4 is characterized in that: The expression for calculating the mean value of the grayscale difference is: Where D ij Represents the mean grayscale difference between two frames, F i (x, y) represents the gray value of frame i at position (x, y), F j (x,y) represents the grayscale value of frame j at position (x,y).

6. The method for identifying abalone crossing-border behavior based on dynamic subtraction statistical characteristics according to claim 1 is characterized in that: The process of drawing the difference sequence into a difference curve to determine the abalone crossing-border behavior includes: Plotting the difference sequence into a difference curve to obtain a first difference curve; Calculating the first derivative of the first difference curve to obtain the curvature change rate; If the curvature change rate at the current moment is higher than the average level, it is determined that abalone crosses the boundary.

7. The method for identifying abalone crossing-border behavior based on dynamic subtraction statistical characteristics according to claim 6 is characterized in that: The process of drawing the inter-frame difference matrix into a difference curve to determine the abalone crossing-border behavior includes: Plotting the inter-frame difference matrix into a difference curve to obtain a second difference curve; Counting the mean and standard deviation of the second difference curve to set a dynamic threshold; Traversing the second difference curve, when a point exceeding the dynamic threshold appears in the second difference curve, determining it as an abnormal point; The abnormal point is matched with the time information of the boundary area, and when the abnormal point occurs in the time period of the boundary area, it is determined that abalone crosses the boundary.

Images