A high-precision precipitation particle phase identification method and system
By assigning unique identifiers to precipitation particles, constructing spatiotemporal trajectories, and performing two-dimensional verification, the problems of low accuracy and data loss in precipitation particle phase identification in existing technologies are solved, and high-precision precipitation particle phase identification is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING WEATHER MODIFICATION OFFICE
- Filing Date
- 2026-04-03
- Publication Date
- 2026-07-07
AI Technical Summary
Existing precipitation particle phase recognition technologies suffer from problems such as repeated particle identifiers, misaligned spatiotemporal trajectory construction, improper handling of misjudged points, and lack of time and space dual-dimensional verification, resulting in low recognition accuracy and data loss.
By assigning a unique identifier to each precipitation particle, performing continuous frame synchronous sampling, constructing a spatiotemporal trajectory and performing integrity verification, and combining temporal and spatial coherence verification, a source tracing sample set of suspected misjudgment points is constructed, and the true phase of the misjudgment points is repaired based on the particle's own data.
It achieves unique tracking of the entire precipitation particle sampling process, avoids duplicate identification and trajectory misalignment, improves identification accuracy, ensures the validity and continuity of data, overcomes the insufficient accuracy of single-dimensional verification, and avoids data loss and phase destruction.
Smart Images
Figure CN121978102B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of atmospheric detection technology, specifically to a high-precision method and system for identifying the phase state of precipitation particles. Background Technology
[0002] Accurate identification of precipitation particle phase is a core technology for meteorological monitoring and precipitation type prediction, and is of great significance for weather forecasting, disaster prevention and mitigation, and other work. However, existing precipitation particle phase identification technologies still have the following shortcomings in practical applications:
[0003] First, particle identification assignment mostly relies on single morphological features, which easily leads to the problem of duplicate identifications and makes it impossible to achieve unique tracking of the entire particle sampling process, directly resulting in misalignment and chaos in the construction of spatiotemporal trajectories;
[0004] Secondly, phase recognition is usually based on single-frame sampling data, without constructing a complete spatiotemporal trajectory for the particle motion process, and lacks verification of the consistency of phase changes;
[0005] Third, the handling of suspicious misjudgment points includes direct removal or blind replacement based on external preset rules, which not only causes the loss of effective data, but also easily disrupts the natural evolution continuity of particle phases, and the repair results deviate significantly from the true phase.
[0006] Fourth, although some technologies attempt trajectory verification, they do not combine time and space dimensions for collaborative verification, making it impossible to effectively identify abnormal sampling points in phase recognition;
[0007] Therefore, there is an urgent need for a high-precision method and system for identifying the phase state of precipitation particles. Summary of the Invention
[0008] To address the shortcomings of existing technologies, this invention provides a high-precision precipitation particle phase identification method and system, which solves the problems of difficulty in tracking precipitation particles, numerous misjudgments in phase identification, and lack of accurate basis for correcting misjudgment points.
[0009] To achieve the above objectives, the present invention provides the following technical solution: a high-precision precipitation particle phase identification method, comprising:
[0010] Step 1: Perform continuous frame synchronous sampling of precipitation particles in the detection area, assign a unique particle identifier to each collected precipitation particle, and synchronously collect the three-dimensional spatial coordinates, phase identification results and sampling timestamps of the particle in each sampling frame. For all continuous sampling points corresponding to the same particle identifier, construct the spatiotemporal trajectory of the particle according to the order of sampling timestamps, and perform integrity verification on the spatiotemporal trajectory. Mark the corresponding phase identification result sequence of the spatiotemporal trajectory that passes the integrity verification.
[0011] Step 2: Based on the spatiotemporal trajectory that has passed the integrity check, perform temporal coherence check and spatial coherence check respectively. Mark any sampling point that fails either of the above two types of checks as a suspected misjudgment point; otherwise, mark it as a valid sampling point, forming a dataset of suspected misjudgment points and a dataset of valid sampling points for each particle.
[0012] Step 3: For each suspected misjudgment point, determine the temporal position of the suspected misjudgment point in the spatiotemporal trajectory according to the unique identifier of the particle to which it belongs, construct the source sample set of the suspected misjudgment point, and repair the true phase of the suspected misjudgment point based on the phase changes of the effective sampling points in the source sample set.
[0013] As a further aspect of the present invention, assigning a unique particle identifier to each collected precipitation particle is specifically performed as follows:
[0014] For each particle image obtained by synchronous sampling of consecutive frames, extract the contour complexity C and particle size morphology comprehensive value F of each particle, where C and F are both ∈ [0,1].
[0015] Pre-determine a uniform interval division number S, and divide [0,1] into S consecutive and non-overlapping sub-intervals, namely [0,1 / S], [1 / S,2 / S], ..., [(S-1) / S,1], and assign fixed interval numbers 1, 2, 3...S to each sub-interval in order from left to right;
[0016] For the quantization features (C, F) of the current particle, they are matched with the above intervals respectively, and the interval number corresponding to C is Rc and the interval number corresponding to F is Rf.
[0017] A fixed integer N greater than the total number of intervals S is preset to generate a unique particle identifier ID, i.e., ID = Rc × N + Rf. The ID generated in the first frame of the particle is used as the permanent unique identifier, and subsequent frames will not recalculate it but will directly use it.
[0018] As a further aspect of the present invention, the specific rule for extracting the contour complexity C of each particle is as follows: traverse the edge pixels of the particle image, calculate the ratio of the total number of particle edge pixels to the total number of pixels of the particle's bounding rectangle, and denote it as C.
[0019] As a further aspect of the present invention, the specific rule for extracting the comprehensive particle size and morphology value F of each particle is as follows:
[0020] By scanning the three-dimensional contour of the particle, the ratio of the particle's major axis length to its minor axis length is calculated and denoted as M. At the same time, the ratio of the particle's volume to the volume of a sphere of the same diameter is calculated and denoted as K.
[0021] Taking the reciprocal of M yields the normalized elongation M' = 1 / M, and performing characteristic coupling calculations on M' and K yields the comprehensive particle size and morphology value F = M' × K.
[0022] As a further aspect of the present invention, the specific steps for verifying the integrity of the spatiotemporal trajectory are as follows:
[0023] Calculate the three-dimensional spatial straight-line distance between every two groups of adjacent sampling points in the spatiotemporal trajectory, denoted as Di, i∈[1,n-1], where n is the total number of sampling points in the spatiotemporal trajectory;
[0024] Based on the above three-dimensional straight-line distances, calculate the arithmetic mean Davg and variance SD of all three-dimensional straight-line distances;
[0025] Based on the characteristics of the spatiotemporal trajectory itself, the smoothness judgment threshold Ts=k×Davg is determined. If the method SD≤Ts of the spatiotemporal trajectory, it is judged as a smoothness qualified trajectory; otherwise, it is judged as a smoothness unqualified trajectory and marked as a candidate incomplete trajectory, which will not proceed to the next step of verification. Here, k is the smoothness ratio coefficient.
[0026] For a smooth trajectory, the Z-axis coordinate of the entrance boundary of the detection area is set as Zin and the Z-axis coordinate of the exit boundary is set as Zout. At the same time, the boundary quantization adaptation difference △Z is set, where Zin > Zout and △Z is an adjustable positive real number.
[0027] Extract the coordinates of the starting sampling point (X1,Y1,Z1) and the ending sampling point (Xn,Yn,Zn) of the smoothness qualified trajectory. Calculate the absolute value of the difference between the Z-axis coordinate of the starting sampling point and the Z-axis coordinate of the entrance boundary, |Z1-Zin|. If |Z1-Zin|≤△Z, the entrance is judged as qualified. Calculate the absolute value of the difference between the Z-axis coordinate of the ending sampling point and the Z-axis coordinate of the exit boundary, |Zn-Zout|. If |Zn-Zout|≤△Z, the exit is judged as qualified.
[0028] If a spatiotemporal trajectory simultaneously meets the requirements of smoothness, inlet fit, and outlet fit, it is considered to have passed the integrity check and will be retained; otherwise, it will be directly rejected.
[0029] As a further aspect of the present invention, the specific steps for performing time continuity verification are as follows:
[0030] For the spatiotemporal trajectory of a particle that has passed the integrity check, extract its corresponding phase identification result sequence, and for each sampling point to be checked in the sequence, extract the phase identification result P of its preceding adjacent sampling point and the phase identification result Q of its following adjacent sampling point, and denote the phase identification result of the current sampling point to be checked as R;
[0031] Compare the phase state identification result R of the current sampling point to be verified with the preceding phase state P and the following phase state Q: If R is a physically changeable intermediate state between P and Q, or is consistent with both P and Q, then the time continuity verification is deemed to have passed; otherwise, the time continuity verification is deemed to have failed.
[0032] The physically changeable intermediate state refers to the state in which the same precipitation particle exists between two different phases during a continuous and singular physical evolution process, and simultaneously possesses the core physical characteristics of both the initial and final phases.
[0033] As a further aspect of the present invention, the specific steps for performing spatial coherence verification are as follows:
[0034] Centered on the three-dimensional spatial coordinates of the current sampling point to be verified, a fixed radius A is set to delineate a three-dimensional spherical spatial neighborhood, and all particles in the neighborhood are traversed and included in the spatiotemporal trajectories that have passed the integrity verification to form a set of effective trajectories in the neighborhood.
[0035] For each complete spatiotemporal trajectory in the neighborhood effective trajectory set, extract its corresponding starting 3D coordinates and ending 3D coordinates, and calculate the starting coordinate distance and ending coordinate distance between any two spatiotemporal trajectories in the set. If the starting coordinate distance and ending coordinate distance are both within the distance determination interval, then they are classified into the same cluster set. Repeat this operation until all trajectories in the neighborhood effective trajectory set are classified into the corresponding cluster set. The distance determination interval is the inherent coordinate error range generated during the device sampling process.
[0036] The number of spatiotemporal trajectories contained in each cluster set is counted, and the cluster set with the most spatiotemporal trajectories is marked as the dominant cluster set. At the same time, the number of spatiotemporal trajectories corresponding to each phase state in the dominant cluster set is counted, and the phase state with the most trajectories is recorded as the dominant phase state of the neighborhood trajectory. The phase state of the current sampling point to be verified is compared with the dominant phase state of the neighborhood trajectory. If the two are consistent, the spatial coherence verification is determined to be passed; otherwise, the spatial coherence verification is determined to be failed.
[0037] As a further aspect of the present invention, the specific operation of constructing the source tracing sample set of suspected misjudgments is as follows:
[0038] Based on the particle identifier associated with the suspected misjudgment point, retrieve the spatiotemporal trajectory corresponding to the identifier that has passed the integrity check, and sort all the sampling points in the spatiotemporal trajectory in ascending order according to the sampling timestamp. At the same time, assign a unique time sequence number to each sampling point.
[0039] Centered on the time sequence number corresponding to the suspected misjudgment point, extend forward by B time sequence numbers and backward by B time sequence numbers to define it as the exclusive time sequence window for the suspected misjudgment point. Then, filter all sampling points within the exclusive time sequence window and retain the valid sampling points that have passed the time continuity check and spatial continuity check.
[0040] For all the selected valid sampling points, they are integrated in a unidirectional ascending order of their corresponding time sequence numbers to construct a dedicated source tracing sample set for the suspected misjudged point. At the same time, in the dedicated source tracing sample set, each valid sampling point is labeled with its corresponding time sequence number and phase identification result.
[0041] As a further aspect of the present invention, repairing the true phase state of suspected misjudgment points specifically includes:
[0042] Retrieve the dedicated source tracing sample set corresponding to the suspected misjudgment point, extract the time sequence number and phase identification result of all valid sampling points in the sample set, and organize them in a unidirectional ascending order of time sequence number to form an ordered sequence of time sequence number-phase state. At the same time, determine the phase state association type of the sample set based on this ordered sequence. The specific rules are as follows: if the phase states of all valid sampling points in the ordered sequence are completely the same, it is determined to be a phase state consistency association; if the phase states of valid sampling points in the ordered sequence increase in time sequence number, showing a continuous evolution from initial phase state → physically changeable intermediate state → terminal phase state, it is determined to be a phase state gradual change association.
[0043] Based on the determined phase association type, the true phase of the suspected misjudged point is determined. The specific rules are as follows: if it is a phase consistency association, the true phase of the suspected misjudged point is completely consistent with the phase of all valid sampling points in the dedicated traceability sample set; if it is a phase gradual change association, the true phase of the suspected misjudged point is the physically gradual intermediate state corresponding to the phase of the two valid sampling points adjacent to its temporal position.
[0044] The determined true phase state is bound to the suspected misjudgment point, and the corrected suspected misjudgment point is included in the effective sampling point dataset of the particle.
[0045] A high-precision precipitation particle phase identification system includes:
[0046] The trajectory construction module performs continuous frame synchronous sampling of precipitation particles in the detection area, assigns a unique particle identifier to each collected precipitation particle, and synchronously collects the three-dimensional spatial coordinates, phase identification results and sampling timestamps of the particle in each sampling frame. For all continuous sampling points corresponding to the same particle identifier, the spatiotemporal trajectory of the particle is constructed in the order of sampling timestamps, and the integrity of the spatiotemporal trajectory is verified. The spatiotemporal trajectory that passes the integrity verification is labeled with its corresponding phase identification result sequence.
[0047] The trajectory verification module performs temporal continuity verification and spatial continuity verification based on the spatiotemporal trajectory that has passed the integrity verification. It marks any sampling point that fails either of the two types of verification as a suspected misjudgment point, otherwise it marks it as a valid sampling point, thus forming a dataset of suspected misjudgment points and a dataset of valid sampling points for each particle.
[0048] The phase restoration module, for each suspected misjudgment point, determines the temporal position of the suspected misjudgment point in the spatiotemporal trajectory according to the unique identifier of the particle to which it belongs, constructs a source sample set for the suspected misjudgment point, and restores the true phase of the suspected misjudgment point based on the phase changes of the effective sampling points in the source sample set.
[0049] This invention provides a high-precision method and system for identifying the phase state of precipitation particles, which has the following advantages compared with the prior art:
[0050] (1) This invention generates a unique particle identifier by coupling dual features, constructs the spatiotemporal trajectory of the particle by continuous frame synchronous sampling and completes the integrity verification, realizes the unique tracking of the entire sampling process of precipitation particles, effectively avoids problems such as duplicate particle identifiers and misaligned trajectory construction, and eliminates incomplete trajectories by dual judgment of smoothness and boundary adaptation, thus ensuring the validity of trajectory data.
[0051] (2) This invention performs time and space coherence verification based on spatiotemporal trajectory. The time verification conforms to the physical gradual change law of particle phase state, and the space verification determines the dominant phase state in the neighborhood through trajectory clustering. The two complement each other and overcome the shortcomings of insufficient accuracy of traditional single-dimensional verification.
[0052] (3) This invention constructs a dedicated source tracing sample set for suspected misjudgment points, and derives the true phase state based on the phase state association type of the effective sampling points of the particle itself. Furthermore, the repair process relies solely on the particle's own data without any external preset rules. This avoids the loss of effective data caused by directly eliminating misjudgment points and solves the problem of the traditional blind replacement of phase state disrupting the continuity of particle evolution. Attached Figure Description
[0053] Figure 1 This is a flowchart of the steps of the present invention;
[0054] Figure 2 This is a flowchart illustrating the steps of assigning unique particle identifiers to precipitation particles in this invention.
[0055] Figure 3 This is the system principle block diagram of the present invention. Detailed Implementation
[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0057] like Figure 1 This invention provides a high-precision method for identifying the phase state of precipitation particles;
[0058] As an embodiment of this application, it includes:
[0059] Step 1: Perform continuous frame synchronous sampling of precipitation particles in the detection area, assign a unique particle identifier to each collected precipitation particle, and synchronously collect the three-dimensional spatial coordinates, phase identification results and sampling timestamps of the particle in each sampling frame. For all continuous sampling points corresponding to the same particle identifier, construct the spatiotemporal trajectory of the particle according to the order of sampling timestamps, and perform integrity verification on the spatiotemporal trajectory. Mark the corresponding phase identification result sequence of the spatiotemporal trajectory that passes the integrity verification.
[0060] Step 2: Based on the spatiotemporal trajectory that has passed the integrity check, perform temporal coherence check and spatial coherence check respectively. Mark any sampling point that fails either of the above two types of checks as a suspected misjudgment point; otherwise, mark it as a valid sampling point, forming a dataset of suspected misjudgment points and a dataset of valid sampling points for each particle.
[0061] Step 3: For each suspected misjudgment point, determine the temporal position of the suspected misjudgment point in the spatiotemporal trajectory according to the unique identifier of the particle to which it belongs, construct the source sample set of the suspected misjudgment point, and repair the true phase of the suspected misjudgment point based on the phase changes of the effective sampling points in the source sample set.
[0062] As a second embodiment of this application, it is implemented based on the first embodiment, except that this embodiment includes:
[0063] Step 1: Perform continuous frame-synchronous sampling of precipitation particles in the detection area and assign a unique particle identifier to each collected precipitation particle.
[0064] Since precipitation particles fall dynamically, single-frame sampling can only capture the state of the particles at a certain moment, but cannot reflect their trajectory and phase change process. Synchronous sampling can ensure that the time base of different sampling frames is consistent, avoiding spatiotemporal deviations caused by asynchronous sampling, and thus avoiding misalignment in subsequent construction of precipitation particle trajectories.
[0065] When performing continuous frame synchronous sampling, the required tools include a high-speed linear scan camera, a three-dimensional laser detection module, and a synchronous trigger controller. The high-speed linear scan camera is responsible for capturing the image features of the particles, the three-dimensional laser detection module is responsible for collecting the spatial position of the particles, and the synchronous trigger controller ensures that the sampling frequency and time reference of the camera and the laser module are completely consistent, so as to achieve synchronous correspondence in the three dimensions of frame-spatial position-time.
[0066] The three-dimensional spatial coordinates, phase identification results, and sampling timestamps of the particle are collected synchronously in each sampling frame:
[0067] The phase changes of precipitation particles may be accompanied by regular changes in spatial position, and subsequent spatial coherence verification needs to be based on the spatial position of the particles. If only two-dimensional coordinates are collected, the spatial distribution and movement trend of the particles cannot be accurately reflected, which will lead to deviations in spatial coherence verification.
[0068] The phase recognition result is the core object for subsequent verification and correction. If the initial phase recognition result is not collected, it is impossible to determine whether the sampling point is a misjudged point or to construct the phase change sequence.
[0069] The phase changes of precipitation particles are usually temporal. The collection timestamp can clarify the order of each sampling point and ensure the temporal correctness of the trajectory construction. Once there is a missing timestamp, it is impossible to determine the order of the sampling points and to construct a coherent spatiotemporal trajectory.
[0070] For all consecutive sampling points corresponding to the same particle identifier, the spatiotemporal trajectory of the particle is constructed in chronological order of sampling timestamps, and the integrity of the spatiotemporal trajectory is verified. The spatiotemporal trajectory that passes the integrity verification is labeled with its corresponding phase recognition result sequence.
[0071] The specific steps for verifying the integrity of the spatiotemporal trajectory are as follows:
[0072] Calculate the three-dimensional spatial straight-line distance between every two groups of adjacent sampling points in the spatiotemporal trajectory, denoted as Di, i∈[1,n-1], where n is the total number of sampling points in the spatiotemporal trajectory;
[0073] Based on the above three-dimensional straight-line distances, calculate the arithmetic mean Davg and variance SD of all three-dimensional straight-line distances;
[0074] Based on the characteristics of the spatiotemporal trajectory itself, the smoothness judgment threshold Ts=k×Davg is determined. If the variance SD of the spatiotemporal trajectory is less than or equal to Ts, it is judged as a smoothness qualified trajectory, indicating that the spatial distance fluctuation between adjacent sampling points is within a reasonable proportion of its own average distance, and there are no sampling breaks or spatial jumps. Otherwise, it is judged as a smoothness unqualified trajectory, indicating that there are spatial distance abrupt changes caused by missed or incorrect sampling, and it is marked as a candidate incomplete trajectory and will not proceed to the next step of verification. Here, k is the smoothness proportion coefficient.
[0075] For a smooth trajectory, the Z-axis coordinate of the entrance boundary of the detection area is set as Zin and the Z-axis coordinate of the exit boundary is set as Zout. At the same time, the boundary quantization adaptation difference ΔZ is set, where Zin > Zout (corresponding to the vertical height direction of the particle's natural fall). ΔZ is an adjustable positive real number used to quantify the degree of adaptation between the start and end sampling points and the boundary.
[0076] Extract the coordinates (X1, Y1, Z1) of the starting sampling point (i.e., the earliest timestamp) and the coordinates (Xn, Yn, Zn) of the ending sampling point (i.e., the latest timestamp) of the smoothness qualified trajectory;
[0077] Calculate the absolute value of the difference between the Z-axis coordinate of the initial sampling point and the Z-axis coordinate of the entrance boundary, |Z1-Zin|. If |Z1-Zin|≤△Z, then the entrance is deemed to be a qualified fit, indicating that the initial sampling point is the sampling point where the particle first enters the detection area.
[0078] Calculate the absolute value of the difference between the Z-axis coordinate of the termination sampling point and the Z-axis coordinate of the exit boundary, |Zn-Zout|. If |Zn-Zout|≤△Z, then the exit fit is qualified, indicating that the termination sampling point is the sampling point where the particle finally leaves the detection area.
[0079] If a spatiotemporal trajectory simultaneously meets the requirements of smoothness, inlet fit, and outlet fit, it is considered to have passed the integrity check and will be retained; otherwise, it will be directly rejected.
[0080] Step 2: Based on the spatiotemporal trajectory that has passed the integrity check, perform temporal consistency check and spatial consistency check respectively;
[0081] The specific steps for performing time continuity verification are as follows:
[0082] For the spatiotemporal trajectory of a particle that has passed the integrity check, extract its corresponding phase identification result sequence, and for each sampling point to be checked in the sequence, extract the phase identification result P of its preceding adjacent sampling point and the phase identification result Q of its following adjacent sampling point, and denote the phase identification result of the current sampling point to be checked as R;
[0083] The phase change of the same precipitation particle follows the law of physical gradual change. Its phase change must have a temporal relationship. Extracting adjacent phases is to build a clear temporal comparison benchmark and avoid misjudgment caused by comparison in a single direction.
[0084] Compare the phase state identification result R of the current sampling point to be verified with the preceding phase state P and the following phase state Q: If R is a physically changeable intermediate state between P and Q, or is consistent with both P and Q, then the time continuity verification is deemed to have passed; otherwise, the time continuity verification is deemed to have failed.
[0085] The phase changes of precipitation particles are all continuous physical gradual processes, such as snowflakes → semi-melted snow → raindrops, raindrops → freezing rain → ice particles. There are no instantaneous jumps without intermediate transitions. Since the physically changeable intermediate state is an inherent physical property of particle phase changes, using this as a judgment rule is in line with the natural physical laws of particles. Therefore, for any known preceding phase state P and subsequent phase state Q, based on the continuous and single physical evolution process type (such as heating and melting, cooling and freezing), those skilled in the art can uniquely determine the corresponding physically changeable intermediate state under the evolution path according to the well-known phase transition laws of precipitation particles. The example given above is a specific explanation of this general judgment logic.
[0086] The physically changeable intermediate state refers to the state between two different phases in the continuous and single physical evolution of the same precipitation particle, which simultaneously possesses the core physical characteristics of the initial phase and the final phase. It is the necessary phase state through which the particle naturally evolves from the initial phase to the final phase.
[0087] The specific steps for performing spatial coherence verification are as follows:
[0088] Centered on the three-dimensional spatial coordinates of the current sampling point to be verified, a fixed radius A is set to delineate a three-dimensional spherical spatial neighborhood, and all particles in the neighborhood are traversed and included in the spatiotemporal trajectories that have passed the integrity verification to form a set of effective trajectories in the neighborhood.
[0089] For each complete spatiotemporal trajectory in the neighborhood effective trajectory set, extract its corresponding starting 3D coordinates and ending 3D coordinates, and calculate the starting coordinate distance and ending coordinate distance between any two spatiotemporal trajectories in the set. If the starting coordinate distance and ending coordinate distance are both within the distance determination interval, then they are classified into the same cluster set. Repeat this operation until all trajectories in the neighborhood effective trajectory set are classified into the corresponding cluster set. The distance determination interval is the inherent coordinate error range generated during the device sampling process.
[0090] Within the same spatial neighborhood, for particles with the same phase, the distribution of the starting and ending coordinates of their complete trajectories has a natural consistency. By adopting start and end coordinate collaborative clustering, the spatiotemporal characteristics of the spatiotemporal trajectory can be fully utilized.
[0091] The number of spatiotemporal trajectories contained in each cluster set is counted, and the cluster set with the most spatiotemporal trajectories is marked as the dominant cluster set. At the same time, the number of spatiotemporal trajectories corresponding to each phase state in the dominant cluster set is counted, and the phase state with the most trajectories is recorded as the dominant phase state of the neighborhood trajectory. The phase state of the current sampling point to be verified is compared with the dominant phase state of the neighborhood trajectory. If the two are consistent, the spatial coherence verification is determined to be passed; otherwise, the spatial coherence verification is determined to be failed.
[0092] Mark any sampling point that fails either of the two types of verification as a suspected misjudgment point; otherwise, mark it as a valid sampling point. This will yield the dataset of suspected misjudgment points and the dataset of valid sampling points for each particle.
[0093] Step 3: For each suspected misjudgment point, based on the unique identifier of the particle to which it belongs, clarify the temporal position of the suspected misjudgment point in the spatiotemporal trajectory, and construct the source sample set of the suspected misjudgment points;
[0094] A suspected misjudged point is a single abnormal sampling point. It is impossible to determine its true phase state by relying solely on the point itself. However, directly removing it would result in the loss of effective data, while blindly replacing it would disrupt the continuity of the trajectory phase state. Therefore, the core purpose of constructing a source tracing sample set is to provide exclusive contextual and effective data support for suspected misjudged points, so as to avoid the subsequent repair process being without a basis.
[0095] The specific steps for constructing the source tracing sample set of suspected misjudgments are as follows:
[0096] Based on the unique particle identifier associated with the suspected misjudgment point, the spatiotemporal trajectory corresponding to the unique identifier that has passed the integrity check is retrieved, and all sampling points in the spatiotemporal trajectory are sorted in ascending order according to the sampling timestamp. At the same time, a unique time sequence number is assigned to each sampling point, and the time sequence number corresponding to the suspected misjudgment point is clarified so as to lock its unique time sequence position in the spatiotemporal trajectory.
[0097] Centered on the time sequence number corresponding to the suspected misjudgment point, extend forward by B time sequence numbers and backward by B time sequence numbers to define the exclusive time sequence window for that suspected misjudgment point. That is, the range of the exclusive time sequence window is strictly limited to "suspicious misjudgment point time sequence number - B" to "suspicious misjudgment point time sequence number + B". Sampling points outside this range are not included. All sampling points within the exclusive time sequence window are filtered to retain valid sampling points that have passed the time continuity check and spatial continuity check.
[0098] The dedicated timing window needs to cover sampling points within a fixed timing range before and after the suspected misjudgment point. This operation can effectively capture the instantaneous phase context of the suspected point, avoiding the introduction of irrelevant data due to an excessively large timing range or insufficient context due to an excessively small timing range. At the same time, only valid sampling points are retained because the phase recognition results of valid sampling points are real and reliable, and can be used as a reference for subsequent repair.
[0099] All valid sampling points selected within the dedicated time series window are integrated in a unidirectional ascending order of their corresponding time series numbers to construct a dedicated source tracing sample set for the suspected misjudged point. At the same time, in the dedicated source tracing sample set, each valid sampling point is labeled with its corresponding time series number and phase identification result to ensure that each data in the sample set can be traced back to the specific time series position of the spatiotemporal trajectory.
[0100] The actual phase state of suspected misjudged points is repaired based on the phase state changes of valid sampling points in the source sample set. The specific operation is as follows:
[0101] Retrieve the dedicated source tracing sample set corresponding to the suspected misjudgment point, extract the time sequence number and phase identification result of all valid sampling points in the sample set, and organize them in unidirectional ascending order of time sequence number to form an ordered sequence of time sequence number-phase state. At the same time, determine the phase state association type of the sample set based on this ordered sequence, with the specific rules as follows:
[0102] If all valid sampling points in an ordered sequence have the same phase state, it is determined to be a phase state consistency association;
[0103] If the phase states of valid sampling points in an ordered sequence increase in time sequence number, exhibiting a continuous evolution from initial phase state to physically changeable intermediate state to final phase state, then it is determined to be a phase-change type correlation.
[0104] For example, if the dedicated source sample set is {(2, snow), (4, snow)}, the sorted ordered sequence is [(2, snow) → (4, snow)]. Since all phases are the same, it is determined to be a phase consistency association. If the sample set is {(1, snow), (3, half-melted snow), (5, rain)}, the sorted ordered sequence is [(1, snow) → (3, half-melted snow) → (5, rain)]. It shows a continuous gradual change, so it is determined to be a phase gradual change association.
[0105] Based on the determined phase correlation type, the true phase of the suspected misjudged point is determined, and the specific rules are as follows:
[0106] If it is a phase consistency correlation, then the true phase of the suspected misjudged point is completely consistent with the phase of all valid sampling points in the dedicated traceability sample set;
[0107] If it is a phase-gradient correlation, the true phase of the suspected misjudged point is the physically gradient intermediate state corresponding to the phase of the two effective sampling points adjacent to its temporal position;
[0108] The determined true phase state is bound to the suspected misjudgment point. At the same time, the repair basis for the suspected misjudgment point is marked in the spatiotemporal trajectory of the particle, namely the time sequence number and phase state of the effective sampling point determined in the dedicated source tracing sample set. The phase state repair of the suspected misjudgment point is completed, and the repaired suspected misjudgment point is included in the effective sampling point dataset of the particle to replace the original misjudged phase state.
[0109] As a third embodiment of this application, this embodiment further discloses a method for assigning unique particle identifiers to precipitation particles based on embodiments one and two, such as... Figure 2 As shown, the specific content includes:
[0110] For each particle image obtained by synchronous sampling of consecutive frames, extract the contour complexity C and particle size morphology comprehensive value F of each particle;
[0111] The specific rule for extracting the contour complexity C of each particle is as follows: traverse the edge pixels of the particle image, calculate the ratio of the total number of particle edge pixels to the total number of pixels of the particle's bounding rectangle, and denote it as C. The larger the ratio, the more irregular the particle contour is, where C∈[0,1].
[0112] The total number of pixels at the particle edge can directly and uniquely characterize the tortuosity and irregularity of the particle contour. The more complex the contour, the more pixels at the edge. The total number of pixels in the bounding rectangle of the particle can characterize the size of the area occupied by the particle as a whole. Using this as a normalization benchmark can eliminate the influence of particle imaging size and distance on the quantization results.
[0113] The specific rules for extracting the comprehensive particle size and morphology value F for each particle are as follows:
[0114] By scanning the three-dimensional contour of the particle, the ratio of the particle's major axis length to its minor axis length is calculated and denoted as M. At the same time, the ratio of the particle's volume to the volume of a sphere of the same diameter is calculated and denoted as K.
[0115] The aforementioned continuous frame synchronous sampling acquires multiple two-dimensional particle images of the same particle from different viewpoints / spatial positions in a very short time. These images are the necessary inputs for particle three-dimensional contour scanning and three-dimensional point cloud / voxel reconstruction.
[0116] M is the only direct quantitative value of the elongation and flatness of the three-dimensional shape of the particle, which can characterize the difference in the geometric shape of the particle from spherical to needle-like / sheet-like / branch-like. K is the quantitative value of the particle's volume density and space filling, which directly reflects whether the particle's interior is loose, hollow, or porous. Even if two particles have the same M, they may belong to different phases due to their different internal density.
[0117] Taking the reciprocal of M, we obtain the normalized elongation M' = 1 / M. Then, we perform a feature coupling operation on M' and K to obtain the particle size and morphology comprehensive value F = M' × K. This value can simultaneously characterize the particle's morphological elongation and volume density, where F ∈ [0, 1].
[0118] Pre-determine a uniform interval division number S, and divide [0,1] into S consecutive and non-overlapping sub-intervals, namely [0,1 / S], [1 / S,2 / S], ..., [(S-1) / S,1], and assign fixed interval numbers 1, 2, 3...S to each sub-interval in order from left to right;
[0119] For the quantization features (C, F) of the current particle, they are matched with the above intervals respectively, and the interval number corresponding to C is Rc and the interval number corresponding to F is Rf.
[0120] Even if the same particle has slight fluctuations in (C,F) in different frames, it will fall into the same interval and get the exact same Rc and Rf. However, different particles with different shapes will fall into different intervals and get different numbers.
[0121] A fixed integer N greater than the total number of intervals S is preset to generate a unique particle identifier ID, i.e., ID = Rc × N + Rf. The ID generated in the first frame of the particle is used as the permanent unique identifier, and subsequent frames will not recalculate it but will directly use it.
[0122] like Figure 3 This invention provides a high-precision precipitation particle phase identification system;
[0123] As a fourth embodiment of this application, the specific steps include the following:
[0124] The trajectory construction module performs continuous frame synchronous sampling of precipitation particles in the detection area, assigns a unique particle identifier to each collected precipitation particle, and synchronously collects the three-dimensional spatial coordinates, phase identification results and sampling timestamps of the particle in each sampling frame. For all continuous sampling points corresponding to the same particle identifier, the spatiotemporal trajectory of the particle is constructed in the order of sampling timestamps, and the integrity of the spatiotemporal trajectory is verified. The spatiotemporal trajectory that passes the integrity verification is labeled with its corresponding phase identification result sequence.
[0125] The trajectory verification module performs temporal continuity verification and spatial continuity verification based on the spatiotemporal trajectory that has passed the integrity verification. It marks any sampling point that fails either of the two types of verification as a suspected misjudgment point, otherwise it marks it as a valid sampling point, thus forming a dataset of suspected misjudgment points and a dataset of valid sampling points for each particle.
[0126] The phase restoration module, for each suspected misjudgment point, determines the temporal position of the suspected misjudgment point in the spatiotemporal trajectory according to the unique identifier of the particle to which it belongs, constructs a source sample set for the suspected misjudgment point, and restores the true phase of the suspected misjudgment point based on the phase changes of the effective sampling points in the source sample set.
[0127] Some of the data in the above formulas are numerical calculations with dimensions removed, and the contents not described in detail in this specification are all prior art known to those skilled in the art.
[0128] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.
Claims
1. A high-precision method for identifying the phase state of precipitation particles, characterized in that, include: Step 1: Perform continuous frame synchronous sampling of precipitation particles in the detection area, assign a unique particle identifier to each collected precipitation particle, and synchronously collect the three-dimensional spatial coordinates, phase identification results and sampling timestamps of the particle in each sampling frame. For all continuous sampling points corresponding to the same particle identifier, construct the spatiotemporal trajectory of the particle according to the order of sampling timestamps, and perform integrity verification on the spatiotemporal trajectory. Mark the corresponding phase identification result sequence of the spatiotemporal trajectory that passes the integrity verification. Step 2: Based on the spatiotemporal trajectory that has passed the integrity check, perform temporal coherence check and spatial coherence check respectively. Mark any sampling point that fails either of the above two types of checks as a suspected misjudgment point; otherwise, mark it as a valid sampling point, forming a dataset of suspected misjudgment points and a dataset of valid sampling points for each particle. Step 3: For each suspected misjudgment point, determine the temporal position of the suspected misjudgment point in the spatiotemporal trajectory according to the unique identifier of the particle to which it belongs, construct the source sample set of the suspected misjudgment point, and repair the true phase of the suspected misjudgment point based on the phase changes of the effective sampling points in the source sample set.
2. The high-precision precipitation particle phase identification method according to claim 1, characterized in that, The specific operation of assigning a unique particle identifier to each collected precipitation particle is as follows: For each particle image obtained by synchronous sampling of consecutive frames, extract the contour complexity C and particle size morphology comprehensive value F of each particle, where C and F are both ∈ [0,1]. Pre-determine a uniform interval division number S, and divide [0,1] into S consecutive and non-overlapping sub-intervals, namely [0,1 / S], [1 / S,2 / S], ..., [(S-1) / S,1], and assign fixed interval numbers 1, 2, 3...S to each sub-interval in order from left to right; For the quantization features (C, F) of the current particle, they are matched with the above intervals respectively, and the interval number corresponding to C is Rc and the interval number corresponding to F is Rf. A fixed integer N greater than the total number of intervals S is preset to generate a unique particle identifier ID, i.e., ID = Rc × N + Rf. The ID generated in the first frame of the particle is used as the permanent unique identifier, and subsequent frames will not recalculate it but will directly use it.
3. The high-precision precipitation particle phase identification method according to claim 2, characterized in that, The specific rule for extracting the contour complexity C of each particle is as follows: traverse the edge pixels of the particle image, calculate the ratio of the total number of particle edge pixels to the total number of pixels of the particle's bounding rectangle, and denote it as C.
4. The high-precision precipitation particle phase identification method according to claim 2, characterized in that, The specific rules for extracting the comprehensive particle size and morphology value F for each particle are as follows: By scanning the three-dimensional contour of the particle, the ratio of the particle's major axis length to its minor axis length is calculated and denoted as M. At the same time, the ratio of the particle's volume to the volume of a sphere of the same diameter is calculated and denoted as K. Taking the reciprocal of M yields the normalized elongation M' = 1 / M, and performing characteristic coupling calculations on M' and K yields the comprehensive particle size and morphology value F = M' × K.
5. The high-precision precipitation particle phase identification method according to claim 1, characterized in that, The specific steps for verifying the integrity of spatiotemporal trajectories are as follows: Calculate the three-dimensional spatial straight-line distance between every two groups of adjacent sampling points in the spatiotemporal trajectory, denoted as Di, i∈[1,n-1], where n is the total number of sampling points in the spatiotemporal trajectory; Based on the above three-dimensional straight-line distances, calculate the arithmetic mean Davg and variance SD of all three-dimensional straight-line distances; Based on the characteristics of the spatiotemporal trajectory itself, the smoothness judgment threshold Ts=k×Davg is determined. If the variance SD of the spatiotemporal trajectory is less than or equal to Ts, it is judged as a smoothness qualified trajectory. Otherwise, it is judged as a smoothness unqualified trajectory and marked as a candidate incomplete trajectory, which will not proceed to the next step of verification. Here, k is the smoothness ratio coefficient. For a smooth trajectory, the Z-axis coordinate of the entrance boundary of the detection area is set as Zin and the Z-axis coordinate of the exit boundary is set as Zout. At the same time, the boundary quantization adaptation difference △Z is set, where Zin > Zout and △Z is an adjustable positive real number. Extract the coordinates of the starting sampling point (X1,Y1,Z1) and the ending sampling point (Xn,Yn,Zn) of the smoothness qualified trajectory. Calculate the absolute value of the difference between the Z-axis coordinate of the starting sampling point and the Z-axis coordinate of the entrance boundary, |Z1-Zin|. If |Z1-Zin|≤△Z, the entrance is judged as qualified. Calculate the absolute value of the difference between the Z-axis coordinate of the ending sampling point and the Z-axis coordinate of the exit boundary, |Zn-Zout|. If |Zn-Zout|≤△Z, the exit is judged as qualified. If a spatiotemporal trajectory simultaneously meets the requirements of smoothness, inlet fit, and outlet fit, it is considered to have passed the integrity check and will be retained; otherwise, it will be directly rejected.
6. The high-precision precipitation particle phase identification method according to claim 1, characterized in that, The specific steps for performing time consistency verification are as follows: For the spatiotemporal trajectory of a particle that has passed the integrity check, extract its corresponding phase identification result sequence, and for each sampling point to be checked in the sequence, extract the phase identification result P of its preceding adjacent sampling point and the phase identification result Q of its following adjacent sampling point, and denote the phase identification result of the current sampling point to be checked as R; Compare the phase state identification result R of the current sampling point to be verified with the preceding phase state P and the following phase state Q: If R is a physically changeable intermediate state between P and Q, or is consistent with both P and Q, then the time continuity verification is deemed to have passed; otherwise, the time continuity verification is deemed to have failed. The physically changeable intermediate state refers to the state in which the same precipitation particle exists between two different phases during a continuous and singular physical evolution process, and simultaneously possesses the core physical characteristics of both the initial and final phases.
7. The high-precision precipitation particle phase identification method according to claim 1, characterized in that, The specific steps for performing spatial coherence verification are as follows: Centered on the three-dimensional spatial coordinates of the current sampling point to be verified, a fixed radius A is set to delineate a three-dimensional spherical spatial neighborhood, and all particles in the neighborhood are traversed and included in the spatiotemporal trajectories that have passed the integrity verification to form a set of effective trajectories in the neighborhood. For each complete spatiotemporal trajectory in the neighborhood effective trajectory set, extract its corresponding starting 3D coordinates and ending 3D coordinates, and calculate the starting coordinate distance and ending coordinate distance between any two spatiotemporal trajectories in the set. If the starting coordinate distance and ending coordinate distance are both within the distance determination interval, then they are classified into the same cluster set. Repeat this operation until all trajectories in the neighborhood effective trajectory set are classified into the corresponding cluster set. The distance determination interval is the inherent coordinate error range generated during the device sampling process. The number of spatiotemporal trajectories contained in each cluster set is counted, and the cluster set with the most spatiotemporal trajectories is marked as the dominant cluster set. At the same time, the number of spatiotemporal trajectories corresponding to each phase state in the dominant cluster set is counted, and the phase state with the most trajectories is recorded as the dominant phase state of the neighborhood trajectory. The phase state of the current sampling point to be verified is compared with the dominant phase state of the neighborhood trajectory. If the two are consistent, the spatial coherence verification is determined to be passed; otherwise, the spatial coherence verification is determined to be failed.
8. The high-precision precipitation particle phase identification method according to claim 1, characterized in that, The specific steps for constructing the source tracing sample set for suspected misjudgments are as follows: Based on the particle identifier associated with the suspected misjudgment point, retrieve the spatiotemporal trajectory corresponding to the identifier that has passed the integrity check, and sort all the sampling points in the spatiotemporal trajectory in ascending order according to the sampling timestamp. At the same time, assign a unique time sequence number to each sampling point. Centered on the time sequence number corresponding to the suspected misjudgment point, extend forward by B time sequence numbers and backward by B time sequence numbers to define it as the exclusive time sequence window for the suspected misjudgment point. Then, filter all sampling points within the exclusive time sequence window and retain the valid sampling points that have passed the time continuity check and spatial continuity check. For all the selected valid sampling points, they are integrated in a unidirectional ascending order of their corresponding time sequence numbers to construct a dedicated source tracing sample set for the suspected misjudged point. At the same time, in the dedicated source tracing sample set, each valid sampling point is labeled with its corresponding time sequence number and phase identification result.
9. The high-precision precipitation particle phase identification method according to claim 1, characterized in that, The specific steps to correct the true phase of suspicious misjudgment points include: Retrieve the dedicated source tracing sample set corresponding to the suspected misjudgment point, extract the time sequence number and phase identification result of all valid sampling points in the sample set, and organize them in a unidirectional ascending order of time sequence number to form an ordered sequence of time sequence number-phase state. At the same time, determine the phase state association type of the sample set based on this ordered sequence. The specific rules are as follows: if the phase states of all valid sampling points in the ordered sequence are completely the same, it is determined to be a phase state consistency association; if the phase states of valid sampling points in the ordered sequence increase in time sequence number, showing a continuous evolution from initial phase state → physically changeable intermediate state → terminal phase state, it is determined to be a phase state gradual change association. Based on the determined phase association type, the true phase of the suspected misjudged point is determined. The specific rules are as follows: if it is a phase consistency association, the true phase of the suspected misjudged point is completely consistent with the phase of all valid sampling points in the dedicated traceability sample set; if it is a phase gradual change association, the true phase of the suspected misjudged point is the physically gradual intermediate state corresponding to the phase of the two valid sampling points adjacent to its temporal position. The determined true phase state is bound to the suspected misjudgment point, and the corrected suspected misjudgment point is included in the effective sampling point dataset of the particle.
10. A high-precision precipitation particle phase identification system, used to execute the high-precision precipitation particle phase identification method according to any one of claims 1-9, characterized in that, include: The trajectory construction module performs continuous frame synchronous sampling of precipitation particles in the detection area, assigns a unique particle identifier to each collected precipitation particle, and synchronously collects the three-dimensional spatial coordinates, phase identification results and sampling timestamps of the particle in each sampling frame. For all continuous sampling points corresponding to the same particle identifier, the spatiotemporal trajectory of the particle is constructed in the order of sampling timestamps, and the integrity of the spatiotemporal trajectory is verified. The spatiotemporal trajectory that passes the integrity verification is labeled with its corresponding phase identification result sequence. The trajectory verification module performs temporal continuity verification and spatial continuity verification based on the spatiotemporal trajectory that has passed the integrity verification. It marks any sampling point that fails either of the two types of verification as a suspected misjudgment point, otherwise it marks it as a valid sampling point, thus forming a dataset of suspected misjudgment points and a dataset of valid sampling points for each particle. The phase restoration module, for each suspected misjudgment point, determines the temporal position of the suspected misjudgment point in the spatiotemporal trajectory according to the unique identifier of the particle to which it belongs, constructs a source sample set for the suspected misjudgment point, and restores the true phase of the suspected misjudgment point based on the phase changes of the effective sampling points in the source sample set.