A method for monitoring Himalayan marmots based on surveillance video images

Abstract

This invention discloses a method for monitoring Himalayan marmots based on surveillance video images, comprising: acquiring video surveillance of the area to be monitored; performing target detection on the video surveillance images to obtain multiple marmot targets; inputting each marmot target into a trajectory tracker to obtain multiple tracking trajectories; excluding misidentified non-marmot trajectories from the multiple tracking trajectories based on speed constraints; splicing the remaining trajectories belonging to the same trajectory based on time and distance constraints to obtain multiple marmot trajectories; calculating the number and speed of the multiple marmot trajectories in different time periods to analyze the activity quantity and speed characteristics of the marmots; and clustering and contour extraction of the multiple marmot trajectories to analyze the activity range characteristics of the marmots. This embodiment can fully detect and mine the activity characteristics of marmots.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a method for monitoring Himalayan marmots based on surveillance video images. Background Technology

[0002] Himalayan marmots are the natural host of plague. By monitoring Himalayan marmots, we can understand their living environment and activity patterns, and prevent the risk of plague.

[0003] In the prior art, patent application CN110443561A discloses a marmot information collection system, which integrates multiple modules such as data collection, data editing, data transmission, and analysis. It analyzes the relationship between marmot distribution and vegetation type, vegetation cover, soil type, and burrow type based on location data, attribute data, and image data. Subsequently, it conducted detection and data collection during plague surveillance in the natural plague focus area of ​​Himalayan marmots in Qinghai Province. Combined with relevant marmot ecological data, it inverted the macroscopic distribution characteristics of Himalayan marmots in Qinghai Province and drew a remote sensing distribution prediction map. This method can predict the distribution of marmots in different regions macroscopically, but it is difficult to focus on changes in the activity characteristics of Himalayan marmots in smaller research areas and to explore marmot movement characteristics.

[0004] Patent application CN114841948A discloses a method for surveying marmot density based on orthophotos from unmanned aerial vehicles (UAVs). First, UAVs collect images of marmots in different survey areas. Then, the images are deduplicated and combined into a single global image. Finally, intelligent recognition technology is used to statistically analyze the number, density, and distribution of marmots in the images, thereby obtaining the number of marmots active in the study area. However, this method has limitations. UAV photography is susceptible to adverse weather conditions and requires on-site operation or monitoring, making it difficult to achieve 24 / 7 monitoring. It also has shortcomings in actively and in real-time monitoring changes in marmot activity characteristics, and there are certain errors in the calculation of marmot density.

[0005] Therefore, there is an urgent need to propose a monitoring method for Himalayan marmots to fully detect and explore their activity characteristics without human intervention. Summary of the Invention

[0006] This invention provides a method for monitoring Himalayan marmots based on surveillance video images to solve the aforementioned technical problems.

[0007] In a first aspect, embodiments of the present invention provide a method for monitoring Himalayan marmots based on surveillance video images, including:

[0008] Acquire video surveillance footage of the area to be monitored;

[0009] Target detection is performed on the video surveillance images to obtain multiple marmot targets; each marmot target is input into a trajectory tracker to obtain multiple tracking trajectories;

[0010] Based on the speed constraint, misidentified non-marmot tracks are excluded from the multiple tracking tracks; based on the time and distance constraints, the tracking tracks belonging to the same track in the remaining tracks are spliced ​​together to obtain multiple marmot tracks;

[0011] The number and speed of the multiple marmot tracks were calculated in different time periods to analyze the activity quantity and speed characteristics of the marmots.

[0012] Clustering and contour extraction were performed on the multiple marmot trajectories to analyze the activity range characteristics of the marmots.

[0013] In a second aspect, embodiments of the present invention provide an electronic device, the electronic device comprising:

[0014] One or more processors;

[0015] Memory, used to store one or more programs.

[0016] When the one or more programs are executed by the one or more processors, the one or more processors implement the Himalayan marmot monitoring method based on surveillance video images as described in any embodiment.

[0017] Thirdly, embodiments of the present invention also provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the Himalayan marmot monitoring method based on surveillance video images as described in any embodiment.

[0018] In summary, this invention provides a method for monitoring Himalayan marmots based on surveillance video images. By using video images of Himalayan marmots in plague foci captured by surveillance probes, and employing target detection models and tracking algorithms, real-time identification and tracking of Himalayan marmots are achieved without human intervention. High-temporal-resolution marmot activity trajectories are extracted from these images, which are then applied to all-weather monitoring and rapid analysis of the number, speed, and range characteristics of Himalayan marmot activity. This method overcomes the shortcomings in accuracy and real-time performance of monitoring marmot activity characteristics using methods such as drones or manual surveys, thus improving the efficiency of investigating and monitoring natural plague foci in Himalayan marmots. Attached Figure Description

[0019] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 This is a flowchart of a Himalayan marmot monitoring method based on surveillance video images provided in an embodiment of the present invention;

[0021] Figure 2 This is a flowchart of another Himalayan marmot detection method based on surveillance video images provided in an embodiment of the present invention;

[0022] Figure 3 This is a schematic diagram of a video image marmot recognition result provided by an embodiment of the present invention;

[0023] Figure 4 This is a schematic diagram of a video image marmot tracking result provided in an embodiment of the present invention;

[0024] Figure 5 This is a schematic diagram illustrating the activity and quantity characteristics of marmots according to an embodiment of the present invention;

[0025] Figure 6 This is a schematic diagram illustrating the activity speed characteristics of a marmot according to an embodiment of the present invention;

[0026] Figure 7 This is a schematic diagram illustrating the maximum gathering range of marmots according to an embodiment of the present invention;

[0027] Figure 8 A schematic structural diagram of an electronic device provided by an embodiment of the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0029] In the description of the present invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate and simplify the description of the present invention. They are not intended to indicate or imply that the devices or components referred to must have, be constructed, or operate in a specific orientation, and therefore should not be construed as limitations on the present invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0030] In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0031] Figure 1 This is a flowchart illustrating a method for monitoring Himalayan marmots based on surveillance video images, provided in an embodiment of the present invention. The method is executed by an electronic device, such as... Figure 1 As shown, the specific steps include the following:

[0032] S110. Construct a marmot image dataset.

[0033] In this embodiment, videos of Himalayan marmots are captured by surveillance cameras in epidemic areas to form a video dataset; and a marmot image dataset is created based on this video dataset as a dataset for building a marmot detection model.

[0034] In one specific implementation, a candidate set of videos with obvious marmot objects is first selected through manual rough identification. Based on this, OpenCV (Open Source Computer Vision Library) is used to obtain the video frame rate. Combining the exposure degree of the target in the video with the target's motion state, the video is sampled frame by frame at fixed time intervals to obtain a preliminary Himalayan marmot target dataset.

[0035] Next, an image dataset for training the deep learning model is constructed. Optionally, the LabelImg tool is used to annotate the initial Himalayan marmot target dataset, including the target category and detection boxes (i.e., bounding boxes), both stored in a format conforming to the YOLO model. For example, the marmot category can be named "marmot," and the storage format can be as follows:

[0036]

[0037] in, This indicates the id (identifier) ​​value for the current category. These represent the x-coordinate and y-coordinate values ​​of the center point of the normalized detection box, as well as the height and width of the detection box, respectively.

[0038] Finally, the image dataset and the labeled dataset are divided according to a certain ratio to generate training and test sets. The training set is used to train a suitable neural network model that can extract and fuse image target features to detect marmot targets. The test set is used to verify the model's detection performance.

[0039] S120. Construct a marmot detection model using the marmot image dataset.

[0040] This step, based on the aforementioned training and testing sets, constructs a marmot detection model through model training, model evaluation, and inference testing.

[0041] In one specific implementation, the version numbers of the software and hardware used for model training are first determined. In this embodiment, the system environment is Windows 11, the CPU (Central Processing Unit) is a 14-core Intel(R) Core(TM) Ultra 5 125H, the memory size is 32GB, the integrated compilation environment is PyCharmCommunity 2024.3, the Python version is 3.12, and the PyTorch version is 2.5.1.

[0042] The YOLO11s model was used for training, and the core parameters involved included image size (imgsz), number of training epochs, batch size, data augmentation strategy (close_mosaic), training device, and number of workers. Under these basic parameters, each marmot image was input into the YOLO11s model, causing the model output to continuously approximate the labeled detection boxes in the input images.

[0043] The trained model needs to be evaluated for its performance. Metrics such as mean average precision (mAP), precision, and recall can be used to evaluate the model. If the model can achieve a high level of precision, it can be used as the final marmot detection model (also known as a detector).

[0044] S130. Acquire video surveillance footage of the area to be monitored; perform target detection on the video surveillance images to obtain multiple marmot targets; input each marmot target into a trajectory tracker to obtain multiple tracking trajectories.

[0045] The area to be monitored here can be an epidemic source area, which may be the same as or different from the epidemic source area in S110. This step combines a detector and a tracker to track Himalayan marmots in video images in real time. Optionally, the tracker can use the DeepSORT algorithm, and the core tracking parameters involved include the maximum number of survival frames of the trajectory (max_age), the minimum number of initialization frames (n_init), the maximum allowable overlap ratio during non-maximum suppression (nms_max_overlap), the cosine distance threshold for appearance matching (max_cosine_distance), the number of nearest neighbor matches (nn_budget), and a ReID (Re-identification) network (embedder) for extracting target appearance features, etc. Each parameter can be determined according to the target's motion state and the complexity of the background in the video image.

[0046] In one specific implementation, after acquiring video surveillance footage of the area to be monitored, the video images are first detected using a detector, and a confidence threshold is set. Targets with a confidence level higher than this threshold are input into the tracker. The DeepSORT algorithm in the tracker first models the target motion state using Kalman filtering and constructs a target motion feature matrix using Mahalanobis distance. Then, it models the target appearance state using a ReID network and constructs a target appearance feature matrix using cosine similarity. Finally, the appearance feature matrix and motion feature matrix are fused to obtain a comprehensive cost matrix. Based on this matrix, a bipartite graph of the detection box sets in adjacent frames is constructed. Then, the Hungarian algorithm is used for data association to find the optimal matching result between each detection box, thereby obtaining the tracking trajectory.

[0047] While ensuring real-time performance, other methods can also be used to detect and track Himalayan marmots. For example, a variant based on YOLO11 can be used to detect marmots, and algorithms such as ByteTrack, BotSORT, and StrongSORT can be used to track the target. This embodiment does not impose any specific limitations.

[0048] S140. Based on the speed constraint, exclude misidentified non-marmot tracks from the multiple tracking tracks; and based on the time constraint and distance constraint, stitch together the remaining tracks that belong to the same track to obtain multiple marmot tracks.

[0049] This step filters and optimizes the multiple tracking trajectories obtained in S130, eliminates misidentified non-marmot trajectories, and splices together the severed marmot trajectories to obtain multiple continuous marmot trajectories.

[0050] Optionally, since the camera is stationary, the Himalayan marmot detection model itself has inherent fixed errors. Furthermore, it is affected by changes in lighting and background, and when using the YOLO11s model for target detection, it may incorrectly identify some background objects similar to marmots (such as rocks) as marmots. Therefore, this embodiment uses trajectory velocity constraints to eliminate most misidentified objects. Simultaneously, due to short-term occlusion of the target, the trajectory of the same target may switch frequently. Therefore, this embodiment uses time and distance constraints to fuse different ID trajectories of the same object, improving the accuracy of trajectory count.

[0051] In one specific implementation, firstly, the average activity speed of each tracking trajectory is calculated based on the positions and timestamps of adjacent trajectory points. Specifically, the position of a trajectory point can be the position of the center point of a detection box, and the timestamp of a trajectory point can be the timestamp of a detection box. The activity speed between two adjacent detection boxes is obtained by dividing the position difference between the center points of any two adjacent detection boxes in a single trajectory by the difference in their timestamps. Assuming there are n detection boxes in a single trajectory (n is a natural number greater than 2), n-1 activity speeds can be obtained. The average activity speed of a single trajectory is obtained by calculating the arithmetic mean of these n-1 activity speeds.

[0052] Next, a maximum and minimum threshold for the marmot's activity speed are set, and tracking trajectories with an average activity speed below the minimum threshold or above the maximum threshold are discarded. The maximum and minimum thresholds can be obtained through a combination of multiple experiments and manual judgment; this embodiment does not impose specific limitations.

[0053] For the remaining tracking trajectories, the following operations can be performed on each pair of trajectories: Obtain the position (x1, y1) and time t1 (i.e., timestamp) of the end point of the first trajectory, Track1, and the position (x2, y2) and time t2 of the start point of the second trajectory, Track2. If the time difference between the two positions is within a set duration threshold, and the displacement value of the center point of the detection box at the two positions in the x-direction or y-direction is less than a set distance threshold, then the two trajectories can be considered to belong to the same complete trajectory. Finally, all segmented trajectories belonging to the same trajectory are stitched together to form a complete continuous trajectory. For ease of distinction and description, this embodiment refers to each stitched continuous trajectory, as well as the trajectory without occlusion segments, as a marmot trajectory.

[0054] S150. Calculate the number and speed of the multiple marmot tracks in different time periods to analyze the activity quantity and speed characteristics of the marmots; perform clustering and contour extraction on the multiple marmot tracks to analyze the activity range characteristics of the marmots.

[0055] This step involves extracting and analyzing the activity characteristics of marmots based on their trajectories. Optionally, changes in the number, speed, and range of marmots can be monitored to further analyze their habitat and activity patterns, thereby mitigating the risk of plague.

[0056] First, we introduce the activity and quantity characteristics of marmots. Optionally, the activity and quantity characteristics can be statistically analyzed on a daily and hourly scale, representing the number of marmot tracks detected in video images each day and hour, respectively. This indicator is the core reference for changes in the activity characteristics of Himalayan marmots, and periods of sudden increase in the number of marmot tracks can be identified as periods of abnormal activity.

[0057] Next, we introduce the activity speed characteristics of marmots. Optionally, based on the average activity speed obtained in S140, the average activity speed of all marmot tracks in each time period can be calculated. Periods with sudden increases or decreases in average activity speed represent periods of abnormal activity speed. Activity speed characteristics are closely related to the behavioral patterns of Himalayan marmots and changes in their surrounding environment. When the overall movement speed of marmots suddenly increases or decreases in a short period of time, it indicates that their movement state has become abnormal.

[0058] Finally, the activity range characteristics of marmots are introduced. These characteristics manifest as the areas marmots prefer to inhabit in video images, i.e., the spatial characteristics of their activity. In one specific implementation, the extraction and analysis of activity range may include the following steps:

[0059] Step 1: Mask the video images. Video images captured by the same camera, besides the area where Himalayan marmots frequently roam, also include other irrelevant areas (such as background areas like the sky). Masking can remove these areas. Optionally, you can use ArcGIS Pro software to create a polygon shapefile of the irrelevant areas, and then mask those areas. The remaining areas are the correct areas to be analyzed.

[0060] Step 2: Perform trajectory clustering within the correct region to be analyzed. Marmot activity trajectory data is inherently chaotic; therefore, a clustering algorithm can be used to cluster all trajectory points, resulting in a cluster corresponding to a marmot activity concentration area. Optionally, the DBSCAN algorithm can be used for clustering. This algorithm defines a "cluster" as a region with sufficient density and performs well in non-convex datasets. This algorithm includes two core parameters: minimum number of points (min_samples) and scan radius (eps). Its basic process can be described as follows: first, randomly select a point in the region as the starting point; then, search for points smaller than eps; continue searching for points smaller than eps based on this new point until no new points are generated. If the number of these points is greater than min_samples, then it can be defined as a cluster.

[0061] Step 3: Trajectory Contour Extraction Based on Clusters. In this embodiment, concave hull contour extraction is performed on each cluster with the highest number of trajectory points to obtain at least one clustered region of marmot activity. Marmot trajectory data is essentially a series of points. After trajectory clustering, clusters of marmot activity can be obtained. The contours of these clusters are uneven, and commonly used convex hull algorithms are difficult to extract the contour range. Therefore, this embodiment uses the Alpha Shapes algorithm to extract the external contour of the clusters. The principle of the two-dimensional Alpha Shapes algorithm is to draw a circle with radius alpha through any two trajectory points. If the distance between the two points is less than twice alpha, there will be two such circles. If there are no other points inside any circle, the line connecting the two points can be considered as one of the boundary lines, thus gradually obtaining the concave hull contour of the cluster group. This contour can preserve the concave region in the area, resulting in a more accurate contour range.

[0062] Step 4: Activity Area Feature Extraction. Taking any clustered region as an example, the texture, brightness, contrast, and edge features inside and outside the clustered region are extracted to analyze the activity range characteristics of marmots. This step extracts image features inside and outside the clustered region and analyzes the characteristics of the marmot activity clustered region based on these features. Optionally, the main image features involved include texture features, brightness, contrast, and edge features. Texture features reflect the visual homogeneity in the image, manifested as the organized and structured arrangement attributes of the object surface. The gray-level co-occurrence matrix can be used to obtain the contrast, correlation, energy, and homogeneity of the image texture as texture features. Image edge features reflect the distribution of high-frequency signals in the image. The Canny algorithm can be used to calculate the gradient distribution in different directions as image edge features. Brightness and contrast can be obtained by solving for the average and standard deviation of gray values ​​in different regions. The texture, brightness, contrast, and edge features inside and outside the clustered region can be used to analyze the activity characteristics of marmots. For example, the activity range of marmots in a new area can be predicted based on the texture features inside and outside the clustered region.

[0063] Figure 2 This is a flowchart of another Himalayan marmot detection method based on surveillance video images provided by an embodiment of the present invention. Step 1, Step 2, and Step 3 in the diagram represent three main steps, and Track represents a trajectory. The entire process of the above method can also be combined with... Figure 2 I understand.

[0064] To demonstrate the monitoring results of the above method, this embodiment applied the method to video observation data from August 22 to September 14, 2024, at a Himalayan marmot plague foci in Lhasa. Each video in the observation data was approximately 30 minutes long. The YOLO11s model and DeepSORT algorithm were used to detect and track Himalayan marmots. The values ​​of the monitoring device parameters, the training parameters of the YOLO11s model, and the tracking parameters of DeepSORT are shown in Table 1.

[0065] Table 1. Values ​​of various experimental parameters

[0066]

[0067] The YOLO11s model stopped training at round 245, showing convergence. The model performed best at round 164, with mAP50 of 0.938, mAP50-95 of 0.470, Precision of 0.947, and Recall of 0.882. Here, mAP50 is the average of the average precision calculated with an IoU threshold of 0.5, and mAP50-95 is the average of the average precision calculated with an IoU threshold of [0.5, 0.95]. IoU represents the intersection-over-union ratio between the predicted detection box and the actual bounding box. Furthermore, the loss changes between the training and test sets are relatively consistent, demonstrating that the marmot detection model possesses a certain degree of generalization ability. Figure 3 This is a schematic diagram of the marmot recognition results in a video image. The values ​​in the diagram represent the confidence scores of the marmot detection bounding boxes.

[0068] Based on this marmot detection model, a confidence threshold of 0.7 was used as the detector's input confidence level. Targets exceeding this threshold were input into the tracker, and the DeepSORT algorithm was used for target tracking. The processing speed reached 19 fps (frames per second), higher than the video frame rate, and can be considered real-time processing. Figure 4 This is a schematic diagram illustrating the tracking results of marmots in video images. The above operation was performed on all surveillance videos, ultimately extracting 765 tracking trajectories from 602 videos.

[0069] Then, based on the characteristics of marmot activity in the images, the minimum threshold (min_v) and maximum threshold (max_v) for marmot activity speed were set to 2 pixels per second and 100 pixels per second, respectively, with a time constraint of 60 seconds and a distance constraint of 50 pixels. Finally, 121 marmot tracks were selected and merged from 765 tracking tracks.

[0070] Based on 121 marmot tracks, the activity quantity, speed, and range characteristics of marmots can be further monitored, and their activity patterns can be analyzed. The final analysis results show that Himalayan marmots move slowly overall, their activity ranges exhibit some clustering, their daytime activity frequency is generally low but locally high, and their hourly activity frequency shows concentrated peaks and stable troughs. Regarding the activity quantity characteristics, frequent marmot activity was detected on only 6 days out of 24 days, with the peak activity occurring at 10:00 AM and the lowest at 7:00 PM. In terms of activity speed characteristics, the average movement speed of marmots in the images is 14.42 pixels per second, with the average speed below 10 pixels per second for most of the time, indicating that the marmots are currently feeding. Figure 6This represents the variation in activity speed characteristics of marmots. In the activity range characteristics, eps in DBSCAN clustering was set to 15, min_samples to 10, and alpha in the Alpha Shape algorithm was set to 0.03. It was found that marmots prefer to be active in low-contrast and low-brightness areas that are mainly grasslands, but also have some mounds, rocks, and low shrubs as occlusions. Figure 5 , Figure 6 These respectively reflect the characteristics of changes in the number and speed of marmot activity. Figure 7 The area represented by the red line represents the largest gathering area of ​​marmot activity. Table 2 reflects the texture features, edge features, contrast and brightness calculation results of the images within and outside the gathering area.

[0071] Table 2. Experimental Results Values

[0072]

[0073] In summary, this embodiment uses video observation data from a certain epidemic source as an example. It employs a target detection and tracking method based on YOLO11s and DeepSORT to detect and track Himalayan marmots in video images in real time. Based on the marmot's trajectory, feature extraction is performed to monitor changes in the marmot's movement status from three aspects: quantity, speed, and activity range. The solution is feasible and the conclusions are reliable.

[0074] In summary, this invention provides a method for monitoring Himalayan marmots based on surveillance video images to improve the investigation results and efficiency of natural plague foci. First, the video images are processed using the YOLO11 model and the DeepSORT algorithm to detect and track marmot targets in real time. Then, the activity characteristics of marmots in natural environments are mined and quantified based on their activity trajectories to monitor changes in their activity status. This method can overcome the shortcomings of using drones or traditional manual surveys and statistics to monitor changes in marmot activity characteristics in terms of initiative and real-time performance, enabling all-weather monitoring and rapid analysis of Himalayan marmot activity characteristics without human intervention. Furthermore, this embodiment tracks the trajectories of targets with confidence levels greater than a set threshold, automatically filling in some missed trajectory points and compensating for missing marmot locations in target detection, providing more trajectory point information for clustering algorithms.

[0075] More specifically, using object detection and trajectory tracking to extract the activity range of marmots has the following advantages: If object detection is performed on each frame of the image without trajectory tracking, it may cause the target to be missed. For example, if the target is briefly occluded (by grass, stones, etc.) or its movement state changes, it will be difficult to detect the target in the image using the detector. However, by using the "object detection + trajectory tracking" method, the same target can be tracked. Even if the target is briefly occluded, the range of the target in the image can be roughly determined by the difference, thus obtaining a more accurate activity range of the marmot.

[0076] Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention, such as... Figure 8 As shown, the device includes a processor 60, a memory 61, an input device 62, and an output device 63; the number of processors 60 in the device can be one or more. Figure 8 Taking a processor 60 as an example; the processor 60, memory 61, input device 62, and output device 63 in the device can be connected via a bus or other means. Figure 8 Taking the example of a connection between China and Israel via a bus.

[0077] The memory 61, as a computer-readable storage medium, can be used to store software programs, computer-executable programs, and modules, such as the program instructions / modules corresponding to the Himalayan marmot monitoring method based on surveillance video images in this embodiment of the invention. The processor 60 executes various functional applications and data processing of the device by running the software programs, instructions, and modules stored in the memory 61, thereby realizing the aforementioned Himalayan marmot monitoring method based on surveillance video images.

[0078] The memory 61 may primarily include a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a given function; the data storage area may store data created based on terminal usage. Furthermore, the memory 61 may include high-speed random access memory and non-volatile memory, such as at least one disk storage device, flash memory, or other non-volatile solid-state storage device. In some instances, the memory 61 may further include memory remotely located relative to the processor 60, which can be connected to the device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0079] Input device 62 can be used to receive input digital or character information, and to generate key signal inputs related to user settings and function control of the device. Output device 63 may include display devices such as a display screen.

[0080] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the Himalayan marmot monitoring method based on surveillance video images according to any embodiment.

[0081] The computer storage medium of this invention can be any combination of one or more computer-readable media. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0082] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0083] Program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0084] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages—such as Java, Smalltalk, and C++—as well as conventional procedural programming languages—such as C or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0085] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A method for monitoring Himalayan marmots based on surveillance video images, characterized in that, include: Acquire video surveillance footage of the area to be monitored; Target detection was performed on the video surveillance images to obtain multiple marmot targets; Input each marmot target into the trajectory tracker to obtain multiple tracking trajectories; Based on the activity patterns of marmots, determine the highest and lowest thresholds for marmot activity speed; The average activity speed of each tracking track is calculated based on the position and timestamp of adjacent track points in each tracking track; Tracking trajectories with an average activity speed higher than the highest threshold or lower than the lowest threshold are identified as misidentified non-marmot trajectories and excluded. Traverse each remaining trajectory; if the difference between the timestamps of the endpoint of one remaining trajectory and the starting point of another remaining trajectory is less than a set duration threshold, and the distance between the endpoint and the starting point along any spatial coordinate axis is less than a set distance threshold, then determine the one remaining trajectory and the other remaining trajectory as tracking trajectories belonging to the same trajectory; then stitch together all tracking trajectories belonging to the same trajectory into a single marmot trajectory. Calculate the number of marmot tracks in each time period, and identify periods of sudden increase in the number of tracks as periods of abnormal activity. Calculate the average activity speed of the marmot's trajectory within each time period, and identify periods of sudden increase or decrease in average activity speed as periods of abnormal activity speed. Monitor environmental changes during abnormal periods to analyze the influencing factors of marmot activity; Multiple clusters were obtained by clustering the trajectory points in multiple marmot trajectories. Concave hull contour extraction is performed on the clusters with the highest number of trajectory points to obtain the activity areas of marmots; texture features, brightness, contrast and edge features inside and outside the clusters are extracted to analyze the activity range features of marmots; based on the texture features inside and outside the clusters, the activity range of marmots in the new area is predicted.

2. The method according to claim 1, wherein performing target detection on the video surveillance image to obtain multiple marmot targets includes: The YOLO model is trained using labeled marmot images, enabling the trained model to display marmots in the images as bounding boxes. Using the trained YOLO model, target detection is performed on the video surveillance images to obtain multiple detection boxes and their confidence scores; Multiple detection boxes with confidence levels higher than a set threshold are identified as multiple marmot targets.

3. The method according to claim 1, characterized in that, The extraction of concave contours from clusters with the highest number of trajectory points includes: The Alpha Shapes algorithm is used to extract the concave hull contours of clusters with the highest number of trajectory points.

4. An electronic device, characterized in that, include: One or more processors; Memory, used to store one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the Himalayan marmot monitoring method based on surveillance video images as described in any one of claims 1-3.

5. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the Himalayan marmot monitoring method based on surveillance video images as described in any one of claims 1-3.

Images