Non-contact river flow measurement method and system based on target tracking algorithm

Abstract

The invention provides a non-contact river flow measurement method and system based on a target tracking algorithm, and relates to the technical field of hydrological survey, and the method comprises the steps: when a monitoring period is triggered, recognizing and determining a plurality of target floaters in a target river region through a plurality of river cameras, and carrying out the video shooting of the plurality of target floaters, a plurality of camera videos are obtained; transmitting the plurality of camera videos to a processor, and tracking and sorting the plurality of camera videos to obtain a plurality of target floating object videos; analyzing the flow velocity of each target floater according to the plurality of target floater videos to obtain a plurality of floater flow velocities, and obtaining the current river flow of the target river region; the early-stage river flow of the last monitoring period is obtained, the river flow increment is obtained through comparison, and when the river flow increment is larger than an increment threshold value, flow abnormity early warning is conducted. The technical problem that in the prior art, a contact type measurement method is inaccurate and unreliable in river flow measurement is solved.

Description

Technical Field

[0001] The present invention relates to the field of hydrological survey, and in particular to a non-contact river flow measurement method and system based on a target tracking algorithm. Background Art

[0002] In river flow measurement, traditional contact measurement methods (such as the current meter method and the buoy method) require the equipment to be placed directly into the water body and rely on manual operation. Since the equipment needs to be immersed in the water flow for data collection, it is extremely susceptible to sediment wear, floating objects entanglement and water flow impact. In addition, due to its reliance on manual operation, measurement data distortion is very likely to occur, which in turn leads to inaccurate and unreliable river flow measurement. Summary of the Invention

[0003] The present invention aims to solve the technical problem that the contact measurement method in the prior art has inaccurate and unreliable river flow measurement, and provides a non-contact river flow measurement method and system based on a target tracking algorithm.

[0004] The technical solution of the present invention to solve the above technical problems is as follows: In a first aspect, the present invention provides a non-contact river flow measurement method based on a target tracking algorithm, comprising: When a monitoring cycle is triggered, a plurality of target floating objects are identified and determined in a target river area by the plurality of river cameras, and videos of the plurality of target floating objects are captured to obtain a plurality of camera videos; transmitting the plurality of camera videos to the processor, tracking and arranging the plurality of camera videos according to a plurality of target floating objects, and obtaining a plurality of target floating object videos; Analyzing the flow velocity of each of the target floating objects according to the videos of the multiple target floating objects to obtain multiple floating flow velocities, and obtaining a current river flow in the target river area based on the multiple floating flow velocities; The previous river flow of the previous monitoring period is obtained, and the previous river flow is compared with the current river flow to obtain a river flow increment. When the river flow increment is greater than an increment threshold, a flow abnormality warning is issued.

[0005] In a second aspect, the present invention provides a non-contact river flow measurement system based on a target tracking algorithm, comprising: a data acquisition module, configured to, when a monitoring cycle is triggered, identify and determine a plurality of target floating objects in a target river area through the plurality of river cameras, and capture videos of the plurality of target floating objects to obtain a plurality of camera videos; a data collating module, configured to transmit the plurality of camera videos to the processor, track and collate the plurality of camera videos according to a plurality of target floating objects, and obtain a plurality of target floating object videos; a flow analysis module, configured to analyze the flow velocity of each of the target floating objects according to the videos of the target floating objects, obtain multiple floating flow velocities, and obtain the current river flow of the target river area based on the multiple floating flow velocities; The abnormal warning module is used to obtain the previous river flow of the previous monitoring period, compare the previous river flow with the current river flow to obtain the river flow increment, and issue a flow abnormality warning when the river flow increment is greater than the increment threshold.

[0006] The beneficial effects of the present invention are: Compared to the existing technology, this application first uses multiple river cameras to identify and determine multiple target floating objects in the target river area when the monitoring cycle is triggered, and videos of the multiple target floating objects are taken to obtain multiple camera videos. Multiple river cameras set on one side of the target river area collect multiple camera videos, and floating object identification is performed to determine multiple target floating objects, providing the necessary data basis for subsequent flow calculations. Secondly, the multiple camera videos are transmitted to the processor, and the multiple camera videos are tracked and organized according to the multiple target floating objects to obtain multiple target floating object videos. The multiple target floating objects are filtered out from the mixed videos collected by the multiple cameras according to the acquisition time sequence, and an independent video of each floating object is separated to provide accurate data support for subsequent flow rate calculations. Next, the velocity of each target floating object is analyzed based on multiple target floating object videos to obtain multiple floating flow velocities. Based on these multiple floating flow velocities, the current river flow in the target river area is obtained. By analyzing the spatial position changes of the floating objects at different time points in the video, the movement velocity of each floating object is calculated. The average cross-sectional flow velocity is then obtained by weighting the confidence of the multiple floating flow velocities. Finally, the current river flow is calculated based on the cross-sectional area of ​​the river, achieving non-contact and accurate calculation of river flow, which improves the accuracy and reliability of flow monitoring. Finally, the previous river flow of the previous monitoring cycle is obtained, and the previous river flow is compared with the current river flow to obtain the river flow increment. When the river flow increment is greater than the increment threshold, a flow anomaly warning is issued. Based on the calculated current river flow and combined with historical data, a flow anomaly warning can be achieved.

[0007] Through the above technical solution, this application collects multiple camera videos of multiple target floats according to the triggered monitoring cycle, then extracts individual videos of each target float. The flow velocities of the multiple target floats are calculated using a target tracking algorithm. The cross-sectional average velocity is then calculated by weighting the confidence of the multiple float flow velocities. Finally, the current river flow is calculated based on the cross-sectional area of ​​the river, and the flow increments between adjacent monitoring cycles are compared to provide an early warning of flow anomalies. In this way, the target tracking algorithm enables non-contact river flow measurement, improving the accuracy and reliability of river flow measurement. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 A flow chart of a non-contact river flow measurement method based on a target tracking algorithm provided by the present invention; Figure 2 This is a structural schematic diagram of the non-contact river flow measurement system based on the target tracking algorithm provided by the present invention.

[0009] In the accompanying drawings, the components represented by the reference numerals are as follows: Data collection module 11, data sorting module 12, traffic analysis module 13, abnormal warning module 14. DETAILED DESCRIPTION

[0010] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 those skilled in the art without making any creative efforts shall fall within the scope of protection of the present invention.

[0011] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of the specified features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0012] In the description of the present invention, the term "for example" is used to mean "used as an example, illustration or illustration". Any embodiment of the present invention described as "for example" is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any person skilled in the art to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes are not elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed herein.

[0013] Example 1, as Figure 1 As shown, an embodiment of the present invention provides a non-contact river flow measurement method based on a target tracking algorithm, including: S10: When a monitoring cycle is triggered, a plurality of target floating objects are identified and determined in a target river area by the plurality of river cameras, and videos of the plurality of target floating objects are captured to obtain a plurality of camera videos; Traditional river flow measurement methods, such as current meters and buoys, suffer from issues such as fragile equipment, high maintenance costs, and difficulty adapting to complex environments. Therefore, by collecting video data from the target river area, a floating object recognition algorithm can be used to accurately locate multiple floating objects. Combined with the timing information captured by the video, the object's movement can be dynamically tracked, and cross-sectional flow velocity and flow can be inferred using a fluid dynamics model.

[0014] In response to the above problems, when the monitoring cycle is triggered, the present application uses the multiple river cameras to identify and determine multiple target floating objects in the target river area, and shoots videos of the multiple target floating objects to obtain multiple camera videos.

[0015] Specifically, step S10 in the method includes: When a monitoring cycle is triggered, the target river area is photographed by the floating object recognition camera group to obtain a front-facing camera video; Within a preset time period, a plurality of target floating objects are sequentially determined through the front camera video; Synchronously starting the floating object tracking camera group to shoot the target river area to obtain a post-camera video; The front camera video and the rear camera video are aggregated to obtain a plurality of camera videos.

[0016] In an embodiment of the present application, multiple river cameras are set along one side of the target river area, including a floating object recognition camera group and a floating object tracking camera group. Multiple river cameras are used to identify and determine multiple target floating objects in the target river area, and video of the multiple target floating objects is captured to obtain multiple camera videos. Among them, there is a certain distance between the floating object recognition camera group and the floating object tracking camera group. Usually, the floating object recognition camera group is set in a certain area upstream of the floating object tracking camera group. For example, the floating object recognition camera group is set 50 meters upstream of the floating object tracking camera group. This is because the floating object recognition camera group first performs a preliminary identification of floating objects in the upstream area of ​​the river and determines the target floating objects. The floating object tracking camera group is located a certain distance downstream. When the floating objects move to this area with the water flow, the relay is used to continuously track and shoot the identified targets. This layout avoids cross-interference between the functions of the two groups of cameras in the same area, ensures efficient connection between the identification and tracking links, and can make the entire monitoring process smoother. For example, the floating object identification camera group is set 50 meters upstream of the floating object tracking camera group. The floating object identification camera group can use a wide-angle lens, such as a 120° field of view, to capture and identify target floating objects by scanning the target river area over a large range. The floating object tracking camera group can use a telephoto lens, such as a 30x optical zoom, to achieve continuous tracking of floating objects.

[0017] Furthermore, when a monitoring cycle is triggered, the target river channel area is photographed by the floating object recognition camera group to obtain a front-facing camera video. The monitoring cycle is a pre-set monitoring frequency, typically 1-10 minutes / time. For example, if the pre-set monitoring cycle is 5 minutes / time, the river channel camera is activated every five minutes to take pictures. Those skilled in the art can dynamically adjust this based on the actual basin conditions. For example, during flood seasons or when flow rates change dramatically, the monitoring cycle can be shortened to 1-2 minutes / time to capture instantaneous flow fluctuations. During dry seasons or when flow rates are stable, the monitoring cycle can be extended to 8-10 minutes / time to save monitoring costs. For example, the pre-set monitoring cycle is 5 minutes / time. When the time reaches the monitoring cycle, the floating object recognition camera group set in the target river channel area is activated, and each floating object recognition camera independently captures a front-facing camera video of 10-30 seconds, obtaining multiple front-facing camera videos.

[0018] Secondly, within a preset time period, multiple target floating objects are sequentially identified using the front-facing camera video. The preset time period is a pre-set target floating object identification time threshold, such as 5 minutes, 10 minutes, etc., which can be dynamically adjusted by those skilled in the art based on actual conditions. Target floating objects are floating objects on the water surface that can be captured, identified, and tracked by the camera crew, such as leaves, branches, bottle caps, and tracer balls. Specifically, the front-facing camera video is transmitted to a processor, frame extraction is performed on the front-facing camera video to obtain a front-facing camera image frame set, and then a floating object identifier for the target river area is called. The floating object identifier sequentially identifies floating objects in each front-facing camera image frame in the front-facing camera image frame set to determine multiple target floating objects. Furthermore, if the target floating object cannot be identified through the front-facing camera video within a preset time period or the number of identified target floating objects is too small, a certain number of naturally degradable tracer balls can be released as target floating objects. For example, if only one target floating object is identified through the front-facing camera video within a preset 10 minutes, 10 naturally degradable tracer balls can be released for the accuracy of flow monitoring.

[0019] Next, the floating object tracking camera group is synchronously activated to capture the target river area and obtain a post-camera video. For example, at the same time as the floating object identification camera group is activated, the floating object tracking camera group is also synchronously activated to capture the target river area and obtain multiple post-camera videos, wherein the shooting time is consistent with that of the floating object identification camera group, such as setting the shooting time to 30 seconds to ensure that the timestamps of the two sets of videos are completely aligned. Synchronously activating the floating object tracking camera group and capturing videos of the same length can avoid the problem of floating object tracks being lost due to perspective switching in traditional single cameras, thereby improving the accuracy of subsequent target identification and trajectory analysis.

[0020] Finally, the front camera video and the rear camera video are combined to obtain multiple camera videos. For example, the multiple front camera videos and the multiple rear camera videos can be combined according to the timestamps to form multiple camera videos with multiple perspectives.

[0021] Specifically, the “determining multiple target floating objects sequentially through the front camera video” includes: Transmitting the front camera video to the processor, performing frame extraction on the front camera video to obtain a front camera image frame set; A floating object identifier of the target river area is retrieved, and floating objects are sequentially identified in each front camera image frame in the front camera image frame set by the floating object identifier to determine a plurality of target floating objects.

[0022] In an embodiment of the present application, the front camera video is first transmitted to the processor, and frame extraction is performed on the front camera video to obtain a front camera image frame set. For example, the processor can perform frame extraction on the front camera video at a fixed frame rate, such as 25-30 frames per second, to obtain a certain number of image sequences as the front camera image frame set. This is to ensure data accuracy while reducing the amount of data. Those skilled in the art can dynamically adjust the frame rate of the front camera video frame extraction based on the actual front camera video duration, camera parameters, etc. For example, if the front camera video duration is 30 seconds, frame extraction can be performed at a fixed frame rate of 25-30 frames per second to extract 750-900 frames of images to obtain the front camera image frame set.

[0023] Next, a floating object identifier for the target river channel area is retrieved, and the floating object identifier sequentially identifies floating objects in each of the front-facing camera image frames in the front-facing camera image frame set to determine a plurality of target floating objects. Specifically, a pre-trained floating object identifier is retrieved, each of the front-facing camera image frames in the front-facing camera image frame set is input, and floating object identification results are output.

[0024] Furthermore, the steps of constructing the "floating object identifier" include: Collecting historical river video data of the target river area and obtaining a preset floating object information database of the target river area; Extracting frames from the historical river channel video data to construct a sample river channel image set; Perform floating object identification on the sample river channel image set based on the preset floating object information library to obtain a sample image floating object label set; The floating object identifier is obtained by training using the sample river image set as input and the sample image floating object label set as a supervision signal.

[0025] In an embodiment of the present application, historical river video data of the target river area is first collected, and a pre-set floating object information database for the target river area is obtained. For example, the historical river video data can be collected by collecting historical video data from river cameras or by using drone oblique photography. For example, the pre-set floating object information database is a feature database constructed based on historical monitoring data of the target river area. It contains information on various types of floating objects that have appeared in the target river area at different times, including multi-dimensional attributes such as the name, color characteristics, texture characteristics, shape characteristics, size range, and movement patterns of each type of floating object. For example, the pre-set floating object information database for a specific target river area may include descriptions of floating plastic bottles, including common colors (e.g., transparent, blue, white, etc.), surface reflective texture, cylindrical profile, and common sizes (e.g., diameter 5-10 cm, height 15-30 cm, etc.). By structured integration of this historical data, the pre-set floating object information database provides accurate feature references for identifying and labeling floating objects.

[0026] Secondly, the historical river channel video data is subjected to frame extraction to construct a sample river channel image set. For example, the historical river channel video data can be subjected to frame extraction at 25-30 frames per second to obtain a certain number of image sequences as the sample river channel image set.

[0027] Next, floating objects are identified in the sample river channel image set based on the preset floating object information library to obtain a sample image floating object label set. For example, the sample river channel image set can be manually identified, and the sample river channel images with floating objects can be labeled using the LabelImg tool based on the preset floating object information library of the target river channel area, with labels including the names of the floating objects, to serve as the sample image floating object label set.

[0028] Finally, the sample river image set is used as input, and the sample image floating object label set is used as a supervisory signal to train and obtain the floating object identifier. For example, the floating object identifier can be constructed based on a deep convolutional neural network, which is mainly composed of a data receiving layer, a model layer, and an output layer. The data layer is responsible for receiving river images, and the model layer includes two parts: a backbone feature extraction network and a multi-scale detection head. The backbone network can use CSPDarknet or ResNet, which extracts visual features from the input river images step by step through multi-layer convolution and downsampling operations. The shallow layer captures details such as edges and textures, and the deep layer abstracts the semantic information of floating objects. In response to the challenge of large scale differences of floating objects in river scenes, the model integrates feature pyramid structures, such as FPN or PANet, through automatic The top-down and bottom-up feature fusion mechanisms combine deep semantic features with shallow detail features, significantly improving the ability to simultaneously identify small floating objects (such as distant foam) and large objects (such as nearby tree trunks). The detection head deploys an anchor box prediction mechanism on the fused multi-scale feature map. Prior anchor boxes with multiple aspect ratios are preset for each spatial location. The convolutional layer outputs bounding box coordinate offsets, target confidence, and category probability in parallel. The post-processing stage uses confidence threshold filtering (to eliminate false detections with a probability below 0.5) and non-maximum suppression (to eliminate overlapping boxes). Finally, the results with category labels are output based on the preset floating object information database. For example, the training process of the floating object identifier can be achieved through the following technical paths: 1. Data preparation: The sample river image set and the sample image floating object label set are divided into training, validation, and test sets in a ratio of 7:1.5:1.5, which serve as training data for model training. 2. Model training: Input the training set data into the floating object identifier, use the sample image floating object label set as the supervision signal, and adopt CrossEntropyLoss as the classification loss function to measure the difference between the recognition result and the label, such as the probability of misclassifying a tree branch as a plastic bottle. When the recognition accuracy of the model on the independent test set reaches above 95%, it is considered converged and the trained floating object identifier is obtained.

[0029] In summary, compared to the prior art, when a monitoring cycle is triggered, the present application uses the multiple river cameras to identify and determine multiple target floating objects in the target river area, and then captures video of the multiple target floating objects to obtain multiple video clips. In this way, by using multiple river cameras set up on one side of the target river area to capture multiple video clips, identify floating objects, and determine multiple target floating objects, this provides the necessary data foundation for subsequent flow calculations.

[0030] S20: transmitting the plurality of camera videos to the processor, tracking and arranging the plurality of camera videos according to a plurality of target floating objects, and obtaining a plurality of target floating object videos; In the above steps, the multiple video clips captured by multiple cameras may contain multiple floating objects. The movement trajectories of these floating objects are intertwined. If they are not distinguished, the subsequent flow velocity calculation will result in large errors due to data mixing.

[0031] In response to the above problem, the present application transmits the multiple camera videos to the processor, tracks and organizes the multiple camera videos according to multiple target floating objects, and obtains multiple target floating object videos.

[0032] Specifically, step S20 in the method includes: Extracting image frames from the plurality of camera videos to obtain a plurality of camera image sequences and a corresponding plurality of acquisition time sequences; Time-aligning and combining the multiple camera image sequences according to multiple acquisition time sequences to obtain a multi-position image sequence; determining a first target floating object from the plurality of target floating objects; Tracking and identifying the movement trajectory of the first target floating object in the multi-position image sequence, obtaining a set of image frames containing the first target floating object, and forming a video of the first target floating object; The remaining target floating object videos are acquired in the same manner as that of acquiring the first target floating object video, to obtain a plurality of target floating object videos.

[0033] In an embodiment of the present application, image frames are first extracted from the multiple camera videos to obtain multiple camera image sequences and corresponding multiple acquisition time sequences. For example, image frames can be extracted from the camera videos captured by each camera at a fixed frame rate, such as 30 frames per second, and the acquisition time corresponding to each frame is recorded to form an acquisition time sequence accurate to the millisecond level, thereby obtaining multiple camera image sequences and corresponding multiple acquisition time sequences.

[0034] Secondly, the multiple camera image sequences are temporally aligned and combined according to the multiple acquisition time sequences to obtain a multi-position image sequence. For example, the multiple camera image sequences are temporally aligned according to the timestamps of the acquisition time sequences to obtain a multi-position image sequence, thereby achieving fusion of camera images taken at the same time point but from different perspectives.

[0035] Next, a first target floating object is determined from the plurality of target floating objects. For example, a floating object is selected from the plurality of target floating objects as the first target floating object, and an object with a larger area and a clear motion trajectory is usually preferred to facilitate tracking of the motion trajectory.

[0036] Furthermore, the trajectory of the first target floating object is tracked and identified in the multi-position image sequence, and a set of image frames containing the first target floating object is obtained to form a video of the first target floating object. For example, a target tracking algorithm such as DeepSORT can be used, combined with visual features, to continuously track and identify the trajectory of the floating object in the multi-position image sequence, filter out all image frames containing the first target floating object, and form a video of the first target floating object. For example, the DeepSORT algorithm is used to track and identify the trajectory of the first target floating object: first, an improved YOLOv5 detector is used to locate the target floating object in the first frame image, and 128-dimensional deep features are extracted to construct a target template. In subsequent frame processing, the target position can be predicted using a Kalman filter, and the motion similarity (IoU) and appearance similarity (cosine distance) between the detection box and the predicted box are calculated. The Hungarian algorithm is used to achieve optimal matching. The final output target trajectory video fully records the movement process of the floating object, and the displacement measurement accuracy reaches ±2%, providing reliable data support for subsequent flow calculations.

[0037] Finally, the remaining target floating object videos are acquired in the same manner as the first target floating object video, resulting in multiple target floating object videos. For example, by repeating the above process, further target floating objects are selected and their trajectory is tracked, resulting in multiple target floating object videos. This process effectively avoids mutual interference among multiple floating objects, extracting the motion trajectory of a single floating object from the complex mixed video stream, and providing reliable data support for subsequent accurate calculation of river flow based on floating object motion analysis.

[0038] In summary, compared to the prior art, this application transmits the multiple camera videos to the processor, tracks and organizes them according to multiple target floating objects, and obtains multiple target floating object videos. This allows multiple target floating objects to be filtered out from the mixed video captured by multiple cameras based on the acquisition time sequence, separating out independent videos of each floating object, providing accurate data support for subsequent flow velocity calculations.

[0039] S30: Analyzing the flow velocity of each of the target floating objects according to the videos of the multiple target floating objects to obtain multiple floating flow velocities, and obtaining a current river flow in the target river area based on the multiple floating flow velocities; The above steps obtain independent videos of multiple target floating objects. These videos fully record the movement trajectories of the floating objects in the target river area. Therefore, by analyzing the spatial position changes of the floating objects at different time points in the video, the movement flow rate of each floating object can be calculated, and the current river flow in the target river area can be obtained based on the flow rate.

[0040] To address the above problem, the present application analyzes the flow velocity of each target floating object according to the multiple target floating object videos to obtain multiple floating flow velocities, and obtains the current river flow of the target river area based on the multiple floating flow velocities.

[0041] Specifically, step S30 in the method includes: extracting a first target floating object video from the plurality of target floating object videos; determining a first movement position sequence and a first movement time sequence of the first target floating object in a target river channel area based on the first target floating object video; obtaining a first floating flow velocity of the first target floating object according to the first motion position sequence and the first motion time sequence; The first float flow rate is added to the plurality of float flow rates.

[0042] In the embodiment of the present application, a first target floating object video is first extracted from the plurality of target floating object videos. For example, an independent video of the first target floating object is selected from the multi-target video set, including a continuous motion trajectory of the first target floating object under the perspectives of multiple cameras.

[0043] Secondly, based on the first target floating object video, determine the first motion position sequence and the first motion time sequence of the first target floating object in the target river area. For example, first obtain the positioning center point of the first target floating object in the first target floating object video frame to obtain the pixel coordinates (x, y), and then use the camera calibration parameters to convert the pixel coordinates of the first target floating object in the video frame into the physical coordinates (X, Y) in the world coordinate system to obtain the first motion position sequence of the first target floating object in the target river area, and at the same time record the acquisition time of each frame in the first target floating object video, strictly correspond the physical coordinates of each frame with the acquisition timestamp, and obtain the first motion time sequence accurate to the millisecond level, for example, obtain the first target floating object in the first target floating object video frame. The pixel coordinates of the positioning center point of each frame in the video are then converted into physical coordinates using the camera calibration parameters, such as [(1000.000, 500.000), (1000.009, 501.000), (1000.017, 501.001), (1000.025, 501.001), …], as the first motion position sequence. At the same time, the timestamps of each frame image in the corresponding first target floating object video are collected, such as [30.123 seconds, 30.126 seconds, 30.129 seconds, 30.132 seconds, …], as the first motion time sequence.

[0044] Next, the first floating velocity of the first target floating object is obtained based on the first motion position sequence and the first motion time sequence. For example, based on the first motion position sequence and the first motion time sequence, the displacement and time difference between adjacent frames can be calculated to obtain the instantaneous velocity, and then the arithmetic average of all instantaneous velocity can be taken to obtain the first floating velocity of the first target floating object. For example, the displacement between adjacent frames can be calculated first, ,in,( , ) is the physical coordinate information of the i-th frame image, ( , ) is the physical coordinate information of the i+1th frame image, is the displacement between the i-th frame image and the i+1-th frame image, and then the time difference between the i-th frame image and the i+1-th frame image is calculated according to the first motion time series: , and then according to the displacement of the i-th frame image and the i+1-th frame image The time difference between the i-th frame image and the i+1-th frame image , calculate the instantaneous flow velocity between the i-th frame image and the i+1-th frame image: For example, the coordinate information of the i-th frame image and the i+1-th frame image are 1000.000, 500.000), (1000.009, 501.000), and then the timestamps corresponding to the i-th frame image and the i+1-th frame image are obtained according to the first motion time sequence, which are 30.123 seconds and 30.126 seconds respectively. =0.009, Seconds = 0.003 seconds, then the instantaneous flow velocity between the i-th frame image and the i+1-th frame image is: , calculate multiple instantaneous flow velocities using the same method, and then take the arithmetic average of all instantaneous flow velocities to obtain the first floating flow velocity of the first target floating object. For example, the calculated first floating flow velocity is 3.1 .

[0045] Finally, the first float flow rate is added to the plurality of float flow rates. For example, the first float flow rate 3.1 Add to multiple float flow rates.

[0046] Furthermore, the “obtaining a current river flow in a target river area based on a plurality of floating flow rates” includes: Obtaining a flow velocity confidence of each target floating object, wherein the flow velocity confidence is determined by a historical tracking success rate of a floating object type corresponding to the target floating object; Performing weighted averaging processing on the plurality of floating objects based on the confidence level of the flow velocity of each target floating object to obtain the cross-sectional average flow velocity of the target river area; The cross-sectional area of ​​the target river channel region is obtained, and the cross-sectional average flow velocity is multiplied by the cross-sectional area to obtain the current river channel flow of the target river channel region.

[0047] In the embodiments of this application, different types of floating objects, such as plastic bottles, branches, and foam, have natural differences in shape, color, and reflective properties during video tracking. For example, a regularly shaped plastic bottle has more stable features in the image and a higher tracking success rate, while a branch has complex edges and is easily confused with the background, resulting in a greater tracking error. Therefore, the flow velocity confidence of different floating objects can be obtained based on historical data. The weighted average of the flow velocities of multiple floating objects can be used to obtain the cross-sectional average flow velocity of the target river area. The cross-sectional area is then combined to calculate the river flow rate. Specifically: First, the velocity confidence level of each target floating object is obtained. This velocity confidence level is determined by the historical tracking success rate of the floating object type corresponding to the target floating object. For example, based on historical data, the historical tracking success rates of different floating object types are obtained. For example, if the historical tracking success rate of a tree branch is 85% and the historical tracking success rate of a plastic bottle is 90%, then 85% and 90% are used as the velocity confidence levels for the tree branch and the plastic bottle, respectively. The higher the velocity confidence level, the more reliable the tracking of that floating object type.

[0048] Secondly, based on the velocity confidence of each target floating object, a weighted average process is performed on the flow velocities of the plurality of floating objects to obtain the cross-sectional average flow velocity of the target river area, wherein: , is the confidence level of the velocity of the i-th floating object type, is the velocity of the i-th type of floating object, and n is the total number of floating object types. For example, the velocity calculated by tracking the branches is 3.1 , tracking the plastic bottle and calculating the flow rate to be 2.8 , the confidence levels of the flow rates of branches and plastic bottles are 85% and 90% respectively. Substituting them into the above cross-section average flow rate calculation formula, we can obtain =2.58 m / s. In this way, by considering the flow velocity confidence of different target floating objects and performing weighted average calculation on the flow velocities of multiple floating objects, the influence of floating object types with low flow velocity confidence in the fusion calculation is reduced, and a more realistic and reliable cross-sectional average flow velocity of the target river area is obtained.

[0049] Finally, the cross-sectional area of ​​the target river channel region is obtained, and the cross-sectional average flow velocity is multiplied by the cross-sectional area to obtain the current river channel flow of the target river channel region, where the current river channel flow = cross-sectional average flow velocity * cross-sectional area. For example, the cross-sectional area can be obtained by manually measuring the cross-sectional geometric dimensions of the target river channel region, or by scanning the cross-sectional area of ​​the target river channel region with a UAV laser radar. For example, the cross-sectional area of ​​the target river channel region obtained by manual measurement is 70m 2 , the average flow velocity of the section is 2.58 m / s, then the current river flow = 70*2.58=180.6m 3 / Second.

[0050] In summary, compared to existing technologies, this application analyzes the flow velocity of each target floating object based on the video of the target floating objects, obtains multiple floating flow velocities, and then uses these multiple floating flow velocities to determine the current river flow in the target river area. Thus, by analyzing the spatial position changes of floating objects at different time points in the video, the movement velocity of each floating object is calculated. The confidence-weighted calculation of the multiple floating flow velocities is then used to determine the average cross-sectional flow velocity. Finally, the current river flow is calculated based on the cross-sectional area of ​​the river, achieving non-contact, precise calculation of river flow and improving the accuracy and reliability of flow monitoring.

[0051] S40: Obtain the previous river flow of the previous monitoring period, compare the previous river flow with the current river flow to obtain a river flow increment, and issue a flow abnormality warning when the river flow increment is greater than an increment threshold.

[0052] In the embodiment of the present application, by obtaining the previous river flow of the previous monitoring cycle and comparing and analyzing it with the current river flow, an abnormal flow warning can be achieved. Specifically, the river flow is continuously calculated according to the preset monitoring cycle (such as 5 minutes / time), and the historical data is stored in the time series database to form a complete flow time series. After the current cycle flow calculation is completed, the flow data of the previous cycle is automatically extracted, and the river flow increment is obtained by calculating the difference between the two. The river flow increment = |previous cycle flow - current river flow|. For example, the monitoring cycle is 5 minutes / time, and the flow of the previous cycle is 170.6m 3 / second, the current river flow is 180.6m 3 / second, then the river flow increment = |170.6m 3 / second-180.6m 3 / second|=10m 3 / Second.

[0053] Furthermore, the increment threshold is pre-set, and those skilled in the art can dynamically adjust the increment threshold according to the river type, historical flow data and hydrological characteristics. For example, the threshold for flood rivers in mountainous areas can be set to a flow increment of 500m 3 / s, and the flow increment for urban drainage rivers can be set to 50m 3 / s, once the river flow increment exceeds the increment threshold, it is immediately determined to be a flow anomaly, and a multi-level early warning mechanism is triggered. Through SMS, GIS map alarms, on-site sound and light prompts, etc., early warnings are issued to relevant departments and personnel. At the same time, the downstream gates are linked to pre-discharge floods, and LED screens along the coast display early warning information, which buys valuable time for flood control scheduling, water resources management and other decision-making, and effectively improves the initiative and effectiveness of river safety management. For example, the increment threshold is 50m 3 / s, the river flow increment is 10m 3 / second, if the incremental threshold is not reached, no traffic anomaly warning will be issued.

[0054] In summary, compared to existing technologies, this application obtains the previous river flow rate from the previous monitoring cycle, compares this previous river flow rate with the current river flow rate to obtain a flow rate increment, and issues a flow anomaly warning when this flow increment exceeds a threshold. In this way, based on the calculated current river flow rate and combined with historical data, a flow anomaly warning can be implemented.

[0055] In summary, the embodiments of the present application have at least the following technical effects: Compared to the existing technology, this application first uses the multiple river cameras to identify and determine multiple target floating objects in the target river area when a monitoring cycle is triggered, and then captures video of the multiple target floating objects to obtain multiple video clips. In this way, multiple river cameras set on one side of the target river area collect multiple video clips, perform floating object identification, and determine multiple target floating objects, providing the necessary data foundation for subsequent flow calculations.

[0056] Next, the present application transmits the multiple camera videos to the processor, tracks and organizes the multiple camera videos according to the multiple target floating objects, and obtains multiple target floating object videos. This allows the multiple target floating objects to be filtered out from the mixed video captured by multiple cameras based on the acquisition time sequence, separating out independent videos of each floating object, providing accurate data support for subsequent flow velocity calculations.

[0057] Furthermore, this application analyzes the flow velocity of each target floating object based on the video of the multiple target floating objects, obtaining multiple floating flow velocities. Based on these multiple floating flow velocities, the current river flow in the target river area is determined. Thus, by analyzing the spatial position changes of floating objects at different time points in the video, the movement velocity of each floating object is calculated. The confidence-weighted calculation of these multiple floating flow velocities is then used to determine the average cross-sectional flow velocity. Finally, the current river flow is calculated based on the cross-sectional area of ​​the river, achieving non-contact, precise river flow calculation and improving the accuracy and reliability of flow monitoring.

[0058] Finally, the application obtains the previous river flow rate from the previous monitoring cycle, compares the previous river flow rate with the current river flow rate to obtain the river flow increment, and issues a flow anomaly warning when the flow increment exceeds the increment threshold. In this way, based on the calculated current river flow rate and combined with historical data, a flow anomaly warning can be implemented.

[0059] Through the above technical solution, this application collects multiple camera videos of multiple target floats according to the triggered monitoring cycle, then extracts individual videos of each target float. The flow velocities of the multiple target floats are calculated using a target tracking algorithm. The cross-sectional average velocity is then calculated by weighting the confidence of the multiple float flow velocities. Finally, the current river flow is calculated based on the cross-sectional area of ​​the river, and the flow increments between adjacent monitoring cycles are compared to provide an early warning of flow anomalies. In this way, the target tracking algorithm enables non-contact river flow measurement, improving the accuracy and reliability of river flow measurement.

[0060] Example 2, as Figure 2 As shown, based on the same inventive concept as the non-contact river flow measurement method based on the target tracking algorithm provided in Example 1, an embodiment of the present invention also provides a non-contact river flow measurement system based on the target tracking algorithm, including: The data acquisition module 11 is configured to, when a monitoring cycle is triggered, identify and determine a plurality of target floating objects in a target river area using the plurality of river cameras, and capture videos of the plurality of target floating objects to obtain a plurality of captured videos; a data collating module 12, configured to transmit the plurality of camera videos to the processor, track and collate the plurality of camera videos according to a plurality of target floating objects, and obtain a plurality of target floating object videos; a flow analysis module 13 for analyzing the flow velocity of each of the target floating objects according to the videos of the target floating objects to obtain a plurality of floating flow velocities, and obtaining a current river flow in the target river area based on the plurality of floating flow velocities; The abnormal warning module 14 is used to obtain the previous river flow of the previous monitoring period, compare the previous river flow with the current river flow to obtain the river flow increment, and issue a flow abnormality warning when the river flow increment is greater than the increment threshold.

[0061] The data acquisition module 11 is specifically used for: When a monitoring cycle is triggered, the target river area is photographed by the floating object recognition camera group to obtain a front-facing camera video; Within a preset time period, a plurality of target floating objects are sequentially determined through the front camera video; Synchronously starting the floating object tracking camera group to shoot the target river area to obtain a post-camera video; The front camera video and the rear camera video are aggregated to obtain a plurality of camera videos.

[0062] Furthermore, the “determining multiple target floating objects sequentially through the front camera video” includes: Transmitting the front camera video to the processor, performing frame extraction on the front camera video to obtain a front camera image frame set; A floating object identifier of the target river area is retrieved, and floating objects are sequentially identified in each front camera image frame in the front camera image frame set by the floating object identifier to determine a plurality of target floating objects.

[0063] Specifically, the steps of constructing the "floating object identifier" include: Collecting historical river video data of the target river area and obtaining a preset floating object information database of the target river area; Extracting frames from the historical river channel video data to construct a sample river channel image set; Perform floating object identification on the sample river channel image set based on the preset floating object information library to obtain a sample image floating object label set; The floating object identifier is obtained by training using the sample river image set as input and the sample image floating object label set as a supervision signal.

[0064] The data sorting module 12 is specifically configured to: Extracting image frames from the plurality of camera videos to obtain a plurality of camera image sequences and a corresponding plurality of acquisition time sequences; Time-aligning and combining the multiple camera image sequences according to multiple acquisition time sequences to obtain a multi-position image sequence; determining a first target floating object from the plurality of target floating objects; Tracking and identifying the movement trajectory of the first target floating object in the multi-position image sequence, obtaining a set of image frames containing the first target floating object, and forming a video of the first target floating object; The remaining target floating object videos are acquired in the same manner as that of acquiring the first target floating object video, to obtain a plurality of target floating object videos.

[0065] The traffic analysis module 13 is specifically configured to: extracting a first target floating object video from the plurality of target floating object videos; determining a first movement position sequence and a first movement time sequence of the first target floating object in a target river channel area based on the first target floating object video; obtaining a first floating flow velocity of the first target floating object according to the first motion position sequence and the first motion time sequence; The first float flow rate is added to the plurality of float flow rates.

[0066] Furthermore, the “obtaining a current river flow in a target river area based on a plurality of floating flow rates” includes: Obtaining a flow velocity confidence of each target floating object, wherein the flow velocity confidence is determined by a historical tracking success rate of a floating object type corresponding to the target floating object; Performing weighted averaging processing on the plurality of floating objects based on the confidence level of the flow velocity of each target floating object to obtain the cross-sectional average flow velocity of the target river area; The cross-sectional area of ​​the target river channel region is obtained, and the cross-sectional average flow velocity is multiplied by the cross-sectional area to obtain the current river channel flow of the target river channel region.

[0067] The abnormal warning module 14 is specifically used to: The previous river flow of the previous monitoring period is obtained, and the previous river flow is compared with the current river flow to obtain a river flow increment. When the river flow increment is greater than an increment threshold, a flow abnormality warning is issued.

[0068] In summary, the embodiments of the present application have at least the following technical effects: Compared to the existing technology, this application first uses a data acquisition module. When the monitoring cycle is triggered, multiple river cameras are used to identify and determine multiple target floating objects in the target river area, and video of the multiple target floating objects is captured to obtain multiple camera videos. Multiple river cameras set on one side of the target river area collect multiple camera videos, and floating object identification is performed to determine multiple target floating objects, providing the necessary data basis for subsequent flow calculations. Secondly, through the data sorting module, the multiple camera videos are transmitted to the processor, and the multiple camera videos are tracked and sorted according to the multiple target floating objects to obtain multiple target floating object videos. The multiple target floating objects are screened out from the mixed video collected by the multiple cameras according to the acquisition time sequence, and an independent video of each floating object is separated to provide accurate data support for subsequent flow rate calculations. Next, through the flow analysis module, the flow velocity of each target floating object is analyzed based on multiple target floating object videos to obtain multiple floating flow velocities. Based on these multiple floating flow velocities, the current river flow in the target river area is obtained. By analyzing the spatial position changes of the floating objects at different time points in the video, the movement velocity of each floating object is calculated. The cross-sectional average flow velocity is then obtained by weighting the confidence of the multiple floating flow velocities. Finally, combined with the cross-sectional area of ​​the river, the current river flow is calculated, achieving non-contact and accurate calculation of river flow, improving the accuracy and reliability of flow monitoring. Finally, through the anomaly warning module, the previous river flow of the previous monitoring cycle is obtained, and the previous river flow is compared with the current river flow to obtain the river flow increment. When the river flow increment exceeds the increment threshold, a flow anomaly warning is issued. Based on the calculated current river flow and combined with historical data, a flow anomaly warning can be implemented. In this way, non-contact river flow measurement is achieved through the target tracking algorithm, improving the accuracy and reliability of river flow measurement.

[0069] It should be noted that, in the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0070] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0071] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0072] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0073] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0074] Although preferred embodiments of the present invention have been described, additional changes and modifications to these embodiments may occur to those skilled in the art once the basic inventive concepts become known.

[0075] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the present invention and its equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A non-contact river flow measurement method based on a target tracking algorithm, characterized in that: Applied to river flow monitoring equipment, the river flow monitoring equipment includes multiple river cameras and a processor; the method includes: When a monitoring cycle is triggered, a plurality of target floating objects are identified and determined in a target river area by the plurality of river cameras, and videos of the plurality of target floating objects are captured to obtain a plurality of camera videos; transmitting the plurality of camera videos to the processor, tracking and arranging the plurality of camera videos according to a plurality of target floating objects, and obtaining a plurality of target floating object videos; Analyzing the flow velocity of each of the target floating objects according to the videos of the multiple target floating objects to obtain multiple floating flow velocities, and obtaining a current river flow in the target river area based on the multiple floating flow velocities; The previous river flow of the previous monitoring period is obtained, and the previous river flow is compared with the current river flow to obtain a river flow increment. When the river flow increment is greater than an increment threshold, a flow abnormality warning is issued.

2. The method according to claim 1, characterized in that The multiple river cameras are arranged along one side of the target river area, including a floating object identification camera group and a floating object tracking camera group; the multiple river cameras are used to identify and determine multiple target floating objects in the target river area, and video of the multiple target floating objects is captured to obtain multiple video recordings, including: When a monitoring cycle is triggered, the target river area is photographed by the floating object recognition camera group to obtain a front-facing camera video; Within a preset time period, a plurality of target floating objects are sequentially determined through the front camera video; Synchronously starting the floating object tracking camera group to shoot the target river area to obtain a post-camera video; The front camera video and the rear camera video are aggregated to obtain a plurality of camera videos.

3. The method according to claim 2, characterized in that The plurality of target floating objects are sequentially determined by the front camera video, including: Transmitting the front camera video to the processor, performing frame extraction on the front camera video to obtain a front camera image frame set; A floating object identifier of the target river area is retrieved, and floating objects are sequentially identified in each front camera image frame in the front camera image frame set by the floating object identifier to determine a plurality of target floating objects.

4. The method according to claim 3, characterized in that The steps of constructing the floating object identifier include: Collecting historical river video data of the target river area and obtaining a preset floating object information database of the target river area; Extracting frames from the historical river channel video data to construct a sample river channel image set; Perform floating object identification on the sample river channel image set based on the preset floating object information library to obtain a sample image floating object label set; The floating object identifier is obtained by training using the sample river image set as input and the sample image floating object label set as a supervision signal.

5. The method according to claim 1, wherein The plurality of camera videos are transmitted to the processor, and the plurality of camera videos are tracked and sorted according to a plurality of target floating objects to obtain a plurality of target floating object videos, including: Extracting image frames from the plurality of camera videos to obtain a plurality of camera image sequences and a corresponding plurality of acquisition time sequences; Time-aligning and combining the multiple camera image sequences according to multiple acquisition time sequences to obtain a multi-position image sequence; determining a first target floating object from the plurality of target floating objects; Tracking and identifying the movement trajectory of the first target floating object in the multi-position image sequence, obtaining a set of image frames containing the first target floating object, and forming a video of the first target floating object; The remaining target floating object videos are acquired in the same manner as that of acquiring the first target floating object video, to obtain a plurality of target floating object videos.

6. The method according to claim 5, characterized in that Analyzing the flow velocity of each target floating object according to the plurality of target floating object videos to obtain a plurality of floating flow velocities includes: extracting a first target floating object video from the plurality of target floating object videos; determining a first movement position sequence and a first movement time sequence of the first target floating object in a target river channel area based on the first target floating object video; obtaining a first floating flow velocity of the first target floating object according to the first motion position sequence and the first motion time sequence; The first float flow rate is added to the plurality of float flow rates.

7. The method according to claim 6, characterized in that The current river flow in the target river area is obtained based on multiple floating flow rates, including: Obtaining a flow velocity confidence of each target floating object, wherein the flow velocity confidence is determined by a historical tracking success rate of a floating object type corresponding to the target floating object; Performing weighted averaging processing on the flow velocities of the plurality of floating objects based on the flow velocity confidence of each target floating object to obtain a cross-sectional average flow velocity of the target river channel area; The cross-sectional area of ​​the target river channel region is obtained, and the cross-sectional average flow velocity is multiplied by the cross-sectional area to obtain the current river channel flow of the target river channel region.

8. The non-contact river flow measurement system based on target tracking algorithm is characterized by: Used to perform the method according to any one of claims 1 to 7, comprising: a data acquisition module, configured to, when a monitoring cycle is triggered, identify and determine a plurality of target floating objects in a target river area through the plurality of river cameras, and capture videos of the plurality of target floating objects to obtain a plurality of camera videos; a data collating module, configured to transmit the plurality of camera videos to the processor, track and collate the plurality of camera videos according to a plurality of target floating objects, and obtain a plurality of target floating object videos; a flow analysis module, configured to analyze the flow velocity of each of the target floating objects according to the videos of the target floating objects, obtain multiple floating flow velocities, and obtain the current river flow of the target river area based on the multiple floating flow velocities; The abnormal warning module is used to obtain the previous river flow of the previous monitoring period, compare the previous river flow with the current river flow to obtain the river flow increment, and issue a flow abnormality warning when the river flow increment is greater than the increment threshold.

Images