Rat damage identification method based on dynamic detection

Abstract

The invention discloses a rat damage identification method based on dynamic detection, and the method comprises the steps: employing an optical flow field to calculate and capture a displacement vector of a mouse between continuous frames, thereby obtaining the movement direction and speed of the mouse; in combination with the image gradient and the optical flow field, the features of the mouse activity area are enhanced through a weighted fusion matrix, and background noise is suppressed; the method comprises the following steps: constructing a background model, optimizing a background by using an optical flow field and a weighted fusion matrix, ensuring that interference of a dynamic object (such as rats) on background updating is minimized, introducing spatial-temporal conjoint analysis of a multi-frame image sequence, performing nonlinear reasoning on a rat activity area, and enhancing the significance of a boundary by using a dynamic boundary diagram. And the mouse activity area can be accurately identified. According to the method, real-time, efficient and accurate mouse recognition service is provided, the features of mice are enhanced through the dynamic optical flow and nonlinear image fusion technology, and the recognition precision and robustness are improved through the self-adaptive background optimization and multi-level region detection technology.

Description

Technical Field

[0001] The present invention relates to the technical field of computer vision and image processing, and in particular to a rodent damage identification method based on dynamic detection. Background Art

[0002] With accelerating urbanization and increasingly complex agricultural production environments, rodent infestations pose a persistent threat to public health, food security, and facility safety. Rodents not only damage crops and supplies but also spread a variety of diseases, causing significant economic losses and health risks. To achieve early detection and rapid response to rodent activity, an efficient, reliable, and intelligent rodent identification technology system is urgently needed.

[0003] Traditional manual patrols are time-consuming, labor-intensive, and rely heavily on experience, making them difficult to cover large areas or monitor at night. Information technology, particularly image processing and dynamic detection, can enable real-time perception and accurate identification of rodent activity without human intervention, becoming a key development direction in rodent control. Consequently, dynamic behavior modeling and multimodal fusion recognition technologies based on image sequences are becoming a hot topic and a key breakthrough in research and engineering applications.

[0004] There are at least the following technical problems in the existing technology: the existing technology is difficult to cope with complex dynamic backgrounds and environmental changes, and is prone to false alarms or missed alarms. The background modeling does not have the ability to adaptively update, and it is difficult to maintain a stable recognition effect under conditions such as lighting changes, occlusion interference, etc. The traditional method has insufficient accuracy in capturing the movement trajectory of mice and cannot extract micro-scale disturbance features, resulting in low recognition accuracy. There is a lack of joint modeling methods for image structural features and motion information, making it difficult to effectively enhance the characteristics of the target area. The overall robustness and real-time performance are poor, making it difficult to meet actual application needs. Summary of the Invention

[0005] The present invention provides a rodent pest identification method based on dynamic detection to solve the problems that the existing technology is difficult to cope with complex dynamic backgrounds and environmental changes, is prone to false alarms or missed alarms, and the background modeling does not have the ability to adaptively update. It is difficult to maintain a stable recognition effect under conditions of lighting changes, occlusion interference, etc. The traditional method has insufficient accuracy in capturing the movement trajectory of rodents and cannot extract micro-scale disturbance features, resulting in low recognition accuracy. It lacks a means to jointly model image structural features and motion information, making it difficult to effectively enhance the characteristics of the target area. The overall robustness and real-time performance are poor, making it difficult to meet actual application needs.

[0006] A method for identifying rodent damage based on dynamic detection of the present invention specifically includes the following technical solutions: A method for identifying rodent damage based on dynamic detection includes the following steps:

[0007] S1. Real-time capture of rodent activity data. Optical flow calculations are used to capture the rodent's displacement vector between consecutive frames, thereby obtaining the rodent's movement direction and speed. Combining image gradients with optical flow, a weighted fusion matrix is ​​used to enhance the characteristics of the rodent's activity area, suppress background noise, and improve the accuracy of rodent recognition.

[0008] S2. After image fusion feature enhancement, a background model is constructed and the background is optimized using the optical flow field and weighted fusion matrix to ensure that the interference of dynamic objects (such as mice) on background updates is minimized. By introducing a spatiotemporal joint analysis of multi-frame image sequences, nonlinear inference is performed on the mouse activity area, and the dynamic boundary map is used to enhance the saliency of the boundary, so as to accurately identify the mouse activity area.

[0009] Preferably, the S1 specifically includes:

[0010] To obtain information on the movement of rodents within the monitoring area, the optical flow field is calculated to capture pixel motion between different time frames. The optical flow field describes the displacement vector of each pixel in the image between consecutive time frames. By analyzing the changes between consecutive frame images, the movement direction and speed of each pixel can be obtained, thereby reflecting the dynamic behavior of moving objects, especially the activity trajectory of rodents.

[0011] Preferably, the S1 specifically includes:

[0012] After acquiring the optical flow field, the image gradient and the optical flow field need to be jointly processed to highlight the characteristics of the rodent activity area. In dynamic scenes, the image gradient represents the local rate of change of the image in the spatial dimension, while the optical flow field reflects the direction and amplitude of the movement of pixels in consecutive frames. To enhance the visual saliency of rodents, a weighted fusion matrix is ​​introduced to characterize the contribution of each image pixel to the recognition of moving targets at a specific moment. Taking into account multiple factors such as the intensity and directionality of the image gradient and the similarity between the optical flow field and the reference optical flow vector, the fusion is performed using exponential decay and nonlinear weighting. The weight of each pixel in the fusion matrix reflects its responsiveness to motion and structural changes.

[0013] Preferably, the S2 specifically includes:

[0014] After completing the image fusion feature enhancement, it is necessary to build a background model that can dynamically adapt to environmental changes in order to effectively separate the difference between static background and dynamic targets (such as mice) in a continuous image sequence; introduce the optical flow field and weighted fusion matrix to finely control the background modeling process to ensure that the background update will not be disturbed by the short and rapid movement of mice.

[0015] Preferably, the S2 specifically includes:

[0016] After background modeling optimization, in order to accurately extract the boundaries of areas where rodents may be active from the image, a spatial-temporal joint analysis method of multi-frame image sequences is introduced to perform nonlinear inference on regional motion features, and finally obtain an accurate dynamic boundary map; by establishing a recursive regional boundary enhancement function to integrate the image disturbances and local texture changes caused by rodent activities in multiple time frames, and combining it with the image weighted fusion results to further enhance boundary contrast and coherence.

[0017] The beneficial effects of the technical solution of the present invention are:

[0018] 1. By calculating dynamic optical flow fields and weighted fusion of image gradients, this invention can accurately capture the movement trajectories and details of rodent activity. This not only improves the visualization of rodent behavior but also effectively eliminates the influence of environmental interference and non-target moving objects. The application of the weighted fusion matrix in the image enhances the visual saliency of rodents, ensuring that rodent activity areas can be distinguished from the background, significantly improving recognition accuracy.

[0019] 2. Through dynamic background modeling and an adaptive update mechanism, the present invention can adjust the background image update process in real time under changing environmental conditions. It can effectively cope with interference such as lighting changes, environmental clutter, and object movement, ensuring the system's continued and efficient operation in dynamic environments.

[0020] 3. By introducing the combined processing and weighted fusion of optical flow and image gradients, the system enhances the recognition of rodent activity areas. Background noise, non-target motion, and irregular changes are effectively suppressed, reducing the system's false positives and false negatives. In particular, the combination of dynamic boundary detection and temporal features makes the boundaries of rodent activity clearer, enabling accurate identification of rodent behavior patterns in complex environments.

[0021] 4. The present invention uses a spatial-temporal joint analysis method for multi-frame image sequences to extract regional features of rodent activity in multiple time frames. Through a recursive regional boundary enhancement function combined with a time weighting function, the dynamic boundaries of the activity area are optimized frame by frame, thereby improving the detection accuracy of the target area, enabling precise positioning of rodent activity, and timely response. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a flow chart of a rodent damage identification method based on dynamic detection according to the present invention. DETAILED DESCRIPTION

[0023] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0024] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0025] The following describes in detail a specific solution of a rodent damage identification method based on dynamic detection provided by the present invention with reference to the accompanying drawings.

[0026] Refer to the attached Figure 1 , which shows a flow chart of a rodent damage identification method based on dynamic detection provided by an embodiment of the present invention, the method comprising the following steps:

[0027] S1. Real-time capture of rodent activity data. Optical flow calculations are used to capture the rodent's displacement vector between consecutive frames, thereby obtaining the rodent's movement direction and speed. Combining image gradients with optical flow, a weighted fusion matrix is ​​used to enhance the characteristics of the rodent's activity area, suppress background noise, and improve the accuracy of rodent recognition.

[0028] Multiple infrared sensors and cameras are deployed within the target monitoring area to capture real-time data on rodent activity. The infrared sensors detect changes in heat within the area, sensitively sensing rodent movement. The sensors transmit this motion data to a central processing platform via a wireless network. The cameras capture video streams at a high frame rate (e.g., 30 frames per second), encoded in YUV420 format.

[0029] To obtain information on the movement of rodents within the monitoring area, the optical flow field is calculated to capture pixel motion between different time frames. The optical flow field describes the displacement vector of each pixel in the image between consecutive time frames. By analyzing the changes between consecutive frame images, the movement direction and speed of each pixel can be obtained, thereby reflecting the dynamic behavior of moving objects, especially the activity trajectory of rodents.

[0030] In order to accurately calculate the optical flow field, we construct an optical flow minimization problem so that the calculated optical flow field best matches the actual changes in the image. We also solve a least squares problem to balance the gradient change and time gradient of the image. The gradient of the image reflects the spatial change of each point in the image, while the time gradient reflects the change of the image over time. The purpose of optical flow is to obtain accurate motion information by minimizing the error between the two, thereby helping to identify the activity patterns of mice. The optical flow field is a quantitative expression of the motion information between image frames, and its calculation formula is as follows:

[0031]

[0032] in, It means the image coordinate position at time t The optical flow field at the location is the motion vector of the current pixel in unit time, which consists of two components Composition, by minimizing the error term in the optical flow field calculation, find the appropriate velocity vector , respectively represent the horizontal and vertical movement speeds of the image; The input image is Position, grayscale value or brightness value at time t, Is the spatial gradient vector of the image, indicating the image and local changes in direction; It is the derivative of image brightness in the time dimension, that is, the rate at which pixel brightness changes over time; Represents the dot product of the gradient vector and the optical flow vector, which measures the movement of the pixel in the gradient direction; is the regularization coefficient, which is used to balance the weight between the data term (optical flow constraint) and the regular term (smoothness term); Represents a vector The square of the second norm is used to control the smoothness of the solution and prevent overfitting. Solve and calculate the movement direction and speed of each pixel .

[0033] After acquiring the optical flow field, the image gradient and the optical flow field need to be jointly processed to highlight the characteristics of the rodent activity area. In dynamic scenes, the image gradient represents the local rate of change of the image in the spatial dimension, while the optical flow field reflects the direction and amplitude of the movement of pixels in consecutive frames. In order to enhance the visual saliency of rodents, a weighted fusion matrix is ​​introduced to characterize the contribution of each pixel in the image to the recognition of moving targets at a specific moment. Taking into account multi-dimensional factors such as the intensity and directionality of the image gradient, the similarity between the optical flow field and the reference optical flow vector, exponential decay and nonlinear weighting are used for fusion. The weight of each pixel in the fusion matrix reflects its response to motion and structural changes. By constructing and calculating the weighted fusion matrix, the background noise area can be effectively suppressed, while the rodent activity area with target characteristics can be enhanced. The formula of the weighted fusion matrix is ​​as follows:

[0034]

[0035] in, is the image coordinate point at time t in the weighted fusion matrix The fusion strength value at ; is the direction number, representing the main direction gradient or 8-neighborhood direction; Representing an image The gradient information in the k-th direction measures the intensity of the pixel change in the k-th direction. represents the spatial variable component in the kth direction; is the weight of the kth direction, indicating its influence on the final fusion value; It is the nonlinear enhancement index of the k-th direction gradient, which determines the degree of influence of local changes in the image; is the reference optical flow vector in the kth direction, and the weighted average of the optical flow vector in the current direction is taken; It is the optical flow variance parameter in the kth direction, which is used to control the scale of the Gaussian kernel and reflects the consistency of the motion.

[0036] By weightedly fusing image gradients and optical flow fields, the visual features of rodent activity areas are made more prominent, thereby improving the accuracy of rodent identification in subsequent steps. At the same time, background noise and irrelevant motion are suppressed, reducing the possibility of false positives.

[0037] S2. After image fusion feature enhancement, a background model is constructed and the background is optimized using the optical flow field and weighted fusion matrix to ensure that the interference of dynamic objects (such as mice) on background updates is minimized. By introducing a spatiotemporal joint analysis of multi-frame image sequences, nonlinear inference is performed on the mouse activity area, and the dynamic boundary map is used to enhance the saliency of the boundary, so as to accurately identify the mouse activity area.

[0038] After completing the image fusion feature enhancement, it is necessary to build a background model that can dynamically adapt to environmental changes in order to effectively separate the difference between static background and dynamic targets (such as mice) in a continuous image sequence; introduce the optical flow field and weighted fusion matrix to finely control the background modeling process to ensure that the background update will not be disturbed by the short and rapid movement of mice.

[0039] Specifically, the background image is the result of weighted regression calculation of the previous T frames. The contribution of each frame is determined by its corresponding time weight, local dynamic adjustment coefficient, and weighted fusion matrix. These three factors work together to act on the difference between the input image and the optical flow field, thereby dynamically optimizing the background model. Its mathematical expression is:

[0040]

[0041] in, At time t, the image coordinate point The background image pixel value at ; is the time window length, which indicates the range of frames to look back in the background modeling process; It is the historical frame time index used when background is updated; is the time weighting coefficient of the historical frame, reflecting the impact of the historical frame of the moment on the current context; is the adaptive motion compensation coefficient, which is used to suppress the interference of moving areas on background estimation and prevent dynamic objects from excessively affecting background updates; $T$ is the time window, which determines the number of historical frames for background updates; and $M(x, y, t')$ is the image fusion matrix calculated in the previous steps, which can further enhance the background modeling effect of moving areas.

[0042] By continuously updating the background model to adapt to environmental changes in different time periods (such as lighting changes, object movement, etc.), and using a weighted fusion matrix to reduce the interference of dynamic objects on background modeling, the accuracy and timeliness of background modeling are ensured.

[0043] After optimizing the background model, a joint spatial-temporal analysis method for multiple image sequences was introduced to accurately extract the boundaries of areas where rodents may be active. Nonlinear inference was performed on regional motion features, ultimately yielding an accurate dynamic boundary map. A recursive regional boundary enhancement function was established to integrate image perturbations and local texture changes caused by rodent activity across multiple time frames. This function was then combined with the weighted image fusion results to further enhance boundary contrast and coherence. Considering the typical temporal continuity and spatial micro-scale perturbation characteristics exhibited by rodents during their activities, a dynamic boundary map was constructed to achieve a dynamic enhanced representation of the boundaries of continuous motion. The formula is as follows:

[0044]

[0045] in, Indicates that at time t, the position The dynamic boundary intensity value of the pixel at reflects the boundary changes of the rodent activity area; Represents the first-order gradient operator, which is used to calculate the intensity of the gray value change of the pixel; Indicates that the second-order Laplace gradient operation is performed on the image to highlight edge changes; is a time-weighted function, for each historical moment Weighted by the degree of contribution; is the minimum boundary strength threshold, which is used to control the saliency of dynamic boundaries and avoid false detection due to image blur.

[0046] The background image in the history frame and weighted fusion matrix Perform frame-by-frame multiplication to form an enhanced spatiotemporal image response, and then pass it through the time weighting function Aggregation reflects the boundary changes of the region over time, and extracts the boundary significance through spatial gradient operation, and finally the maximum function and threshold are combined to form the boundary Controlling the robustness and sensitivity of the output results ensures that the boundary response can accurately lock the mouse activity area. It has high robustness and real-time response capabilities, and perfectly connects with the aforementioned background optimization and image fusion stages.

[0047] The rodent identification method of the present invention can effectively cope with various changes in dynamic environments and provide real-time, efficient and accurate rodent identification services. It not only enhances the characteristics of rodents through dynamic optical flow and nonlinear image fusion technology, but also improves the accuracy and robustness of identification through adaptive background optimization and multi-level area detection technology.

[0048] In summary, a rodent damage identification method based on dynamic detection was completed.

[0049] The order in which the embodiments of the invention are presented is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0050] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

[0051] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. A rodent damage identification method based on dynamic detection, characterized in that: Including steps: S1. Real-time capture of rodent activity data. Optical flow calculations are used to capture the rodent's displacement vector between consecutive frames, thereby obtaining the rodent's movement direction and speed. Combining image gradients with optical flow, a weighted fusion matrix is ​​used to enhance the characteristics of the rodent's activity area, suppress background noise, and improve the accuracy of rodent recognition. S2. After image fusion feature enhancement, a background model is constructed and the background is optimized using the optical flow field and weighted fusion matrix to ensure that the interference of dynamic objects such as rodents on background updates is minimized. By introducing a spatiotemporal joint analysis of multi-frame image sequences, nonlinear inference is performed on the rodent activity area, and the dynamic boundary map is used to enhance the saliency of the boundary, so as to accurately identify the rodent activity area.

2. The rodent damage identification method based on dynamic detection according to claim 1, characterized in that: In S1, in order to accurately calculate the optical flow field, an optical flow field minimization problem is constructed so that the calculated optical flow field best conforms to the actual changes of the image. The optical flow field is a quantitative expression of the motion information between image frames, and its calculation formula is as follows: ； in, It means the image coordinate position at time t The optical flow field at the location is the motion vector of the current pixel in unit time, which consists of two components Composition, by minimizing the error term in the optical flow field calculation, find the appropriate velocity vector , respectively represent the horizontal and vertical movement speeds of the image; The input image is Position, grayscale value or brightness value at time t, Is the spatial gradient vector of the image, indicating the image and local changes in direction; It is the derivative of image brightness in the time dimension, that is, the rate at which pixel brightness changes over time; Represents the dot product of the gradient vector and the optical flow vector, which measures the movement of the pixel in the gradient direction; is the regularization coefficient, which is used to balance the weight between the data term optical flow constraint and the regular term smoothing term; Represents a vector The square of the second norm is used to control the smoothness of the solution and prevent overfitting; the standard least squares method is used to Solve and calculate the movement direction and speed of each pixel .

3. The rodent damage identification method based on dynamic detection according to claim 1, characterized in that: In S1, the image gradient and the optical flow field are jointly processed to highlight the characteristics of the rodent activity area; a weighted fusion matrix is ​​introduced, and the formula is as follows: ； in, is the image coordinate point at time t in the weighted fusion matrix The fusion strength value at ; is the number of directions, representing the main direction gradient or 8-neighborhood direction; Representing an image The gradient information in the k-th direction measures the intensity of the pixel change in the k-th direction. represents the spatial variable component in the kth direction; is the weight of the kth direction, indicating its influence on the final fusion value; It is the nonlinear enhancement index of the k-th direction gradient, which determines the degree of influence of local changes in the image; is the reference optical flow vector in the kth direction, and the weighted average of the optical flow vector in the current direction is taken; It is the optical flow variance parameter in the kth direction, which is used to control the scale of the Gaussian kernel and reflects the consistency of the motion.

4. The rodent damage identification method based on dynamic detection according to claim 1, characterized in that: In S2, a background model that can dynamically adapt to environmental changes is constructed, and the optical flow field and weighted fusion matrix are introduced to finely control the background modeling process to ensure that the background update will not be disturbed by short and rapid mouse movements. Its mathematical expression is: ； in, At time t, the image coordinate point The background image pixel value at ; is the time window length, which indicates the range of frames to look back in the background modeling process; It is the historical frame time index used when background is updated; is the time weighting coefficient of the historical frame, reflecting the impact of the historical frame of the moment on the current context; is the adaptive motion compensation coefficient, which is used to suppress the interference of moving areas on background estimation and prevent dynamic objects from excessively affecting the background update; $T$ is the time window, which determines the number of historical frames for background update; and $M(x,y,t')$ is the image fusion matrix calculated in the previous steps, which can further enhance the background modeling effect of the moving area.

5. The rodent damage identification method based on dynamic detection according to claim 1, characterized in that: In S2, a spatial-temporal joint analysis method of multi-frame image sequences is introduced to perform nonlinear reasoning on regional motion features, and finally obtain an accurate dynamic boundary map. The formula is as follows: ； in, Indicates that at time t, the position The dynamic boundary intensity value of the pixel at reflects the boundary changes of the rodent activity area; Represents the first-order gradient operator, which is used to calculate the intensity of the gray value change of the pixel; Indicates that the second-order Laplace gradient operation is performed on the image to highlight edge changes; is a time-weighted function, for each historical moment Weighted by the degree of contribution; is the minimum boundary strength threshold, which is used to control the saliency of dynamic boundaries and avoid false detection due to image blur.

Images