CN-122020215-A - Method for detecting weak signals of ship network based on density clustering

CN122020215ACN 122020215 ACN122020215 ACN 122020215ACN-122020215-A

Abstract

The invention discloses a ship network weak signal detection method based on density clustering, which comprises the following steps of obtaining a task difficulty coefficient through a task difficulty model based on signal-to-noise ratio, signal duty ratio and time-frequency diagram background uniformity, obtaining a characteristic quality coefficient through a characteristic quality model based on total number of characteristic points, characteristic point signal-to-noise ratio distribution and characteristic space dimension, obtaining a process state coefficient through a process state model based on core point proportion, iteration number and final clustering quantity, obtaining a clustering adaptation degree through a clustering adaptation degree model based on the characteristic quality coefficient and noise point proportion and clustering quality index under the process state coefficient, and obtaining a target neighborhood radius through a neighborhood radius optimization model based on the clustering adaptation degree, the task difficulty coefficient and the neighborhood radius. The invention provides a more reliable and efficient weak signal detection solution in complex and changeable marine environments.

Inventors

YU XIAOXIA
YAO YELI

Assignees

浙江广厦建设职业技术大学
北京高校邦数字科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260120

Claims (10)

1. A ship network weak signal detection method based on density clustering is characterized by comprising the following steps: s1, acquiring a task difficulty coefficient through a task difficulty model based on signal-to-noise ratio, signal duty ratio and time-frequency diagram background uniformity; S2, acquiring a characteristic quality coefficient through a characteristic quality model based on the total number of the characteristic points, the signal-to-noise ratio distribution of the characteristic points and the characteristic space dimension; s3, acquiring a process state coefficient through a process state model based on the proportion of core points, the iteration times and the final clustering quantity; s4, acquiring a clustering adaptation degree through a clustering adaptation degree model based on the characteristic quality coefficient and the noise point proportion and the clustering quality index under the process state coefficient; s5, acquiring a target neighborhood radius through a neighborhood radius optimization model based on the cluster adaptation degree, the task difficulty coefficient and the neighborhood radius.
2. The method for detecting the weak signal of the ship network based on the density clustering of claim 1, wherein in S1, the method for acquiring the task difficulty coefficient comprises the following steps: s11, acquiring signal-to-noise ratio, signal duty ratio and time-frequency diagram background uniformity; s12, carrying out maximum-minimum normalization treatment on the signal to noise ratio, and then taking the complement thereof to obtain a signal to noise ratio index; S13, carrying out maximum-minimum normalization processing on the signal duty ratio and the background uniformity of the time-frequency chart to obtain a signal duty ratio index and a background uniformity index; S14, acquiring a task difficulty coefficient through a task difficulty model based on the signal-to-noise ratio index, the signal duty ratio index and the background uniformity index.
3. The method for detecting the weak signal of the ship network based on the density clustering of claim 2, wherein the task difficulty model is expressed as: ; Wherein, the The coefficient of difficulty of the task is represented, Represents the signal-to-noise ratio index, Representing the duty cycle index of the signal, The background uniformity index is indicated as such, And the larger the value, the more difficult the detection task.
4. The method for detecting the weak signal of the ship network based on the density clustering of claim 1, wherein in the step S2, the method for acquiring the characteristic quality coefficient comprises the following steps: s21, obtaining the total number of the feature points, the signal-to-noise ratio distribution of the feature points and the feature space dimension; S22, carrying out ratio processing on the total number of the characteristic points and the signal-to-noise ratio distribution of the characteristic points and the corresponding maximum value to obtain a characteristic point quantity index and a signal-to-noise ratio distribution index; S23, carrying out ratio processing on the difference value between the feature space dimension and the optimal feature space dimension and the allowable deviation from the optimal feature space dimension value to obtain a feature space dimension index; S24, acquiring a characteristic quality coefficient through a characteristic quality model based on the characteristic point quantity index, the signal-to-noise ratio distribution index and the space dimension index.
5. The method for detecting weak signals of the ship network based on density clustering of claim 4, wherein the characteristic quality model is expressed as: ; Wherein, the The characteristic quality coefficient is represented by a characteristic, An index indicating the number of feature points, Represents the signal-to-noise ratio distribution index, The index of the spatial dimension is represented, And the larger the value, the higher the feature quality.
6. The method for detecting the weak signal of the ship network based on the density clustering of claim 1, wherein in the step S3, the method for acquiring the process state coefficient comprises the following steps: s31, obtaining the proportion of core points, the iteration times and the final clustering quantity; S32, carrying out maximum-minimum normalization processing on the proportion of the core points to obtain a proportion index of the core points; S33, carrying out ratio processing on the difference value between the iteration times and the expected iteration times and the allowable deviation expected value to obtain an iteration times index; s34, carrying out ratio processing on the difference value between the final cluster number and the expected cluster number and the allowable deviation expected cluster number value to obtain a cluster number index; s35, acquiring a process state coefficient through a process state model based on the core point proportion index, the iteration number index and the clustering number index.
7. The method for detecting weak signals of a ship network based on density clustering as set forth in claim 6, wherein said process state model is expressed as: ; Wherein, the The process state coefficients are represented by a set of values, The core point scale index is represented by a graph, Represents an index of the number of iterations, Represents an index of the number of clusters, And the larger the value, the more desirable the clustering process state.
8. The method for detecting the weak signals of the ship network based on the density clustering according to claim 1, wherein the method for acquiring the clustering fitness in S4 comprises the following steps: s41, acquiring a noise point proportion and a clustering quality index (such as a contour coefficient); S42, carrying out absolute difference processing on the noise point proportion and the optimal noise point proportion, and carrying out ratio processing on the noise point proportion value which is allowed to deviate from the optimal noise point proportion value to obtain the noise point proportion value; S43, importing the clustering quality index into a formula And obtaining a cluster quality index, wherein, Representing a cluster quality index; S44, acquiring a comprehensive quality coefficient through a comprehensive quality model based on the characteristic quality coefficient and the process state coefficient; s45, acquiring the clustering adaptation degree through a clustering adaptation degree model based on the noise point proportion index and the clustering quality index under the comprehensive quality coefficient.
9. The method for detecting weak signals of the ship network based on density clustering of claim 8, wherein the comprehensive quality model is expressed as: ; The cluster fitness model is expressed as: ; Wherein, the The degree of adaptation of the clusters is indicated, The coefficient of mass of the composite is represented, Representing the noise point scale index (plf), The index of the cluster quality index is represented, , And the larger the value is, the higher the matching degree between the current clustering parameter setting and the data characteristic is, and the better the detection performance is.
10. The method for detecting weak signals of a ship network based on density clustering according to claim 1, wherein in the step S5, a neighborhood radius optimization model is expressed as follows: ; Wherein, the Representing the radius of the target neighborhood, Representing the radius of the current field, The coefficient of difficulty of the task is represented, The degree of adaptation of the clusters is indicated, Representing the adjustment amplitude.

Description

Method for detecting weak signals of ship network based on density clustering Technical Field The invention belongs to the technical field of weak signal detection, and particularly relates to a ship network weak signal detection method based on density clustering. Background The ship network often faces the problem of weak signal detection in a complex marine environment, and the performance of the traditional method based on energy detection or fixed threshold value is obviously reduced under the background of low signal-to-noise ratio and non-stationary noise. In recent years, an unsupervised learning method based on density clustering is introduced into the field of signal detection, and the separation of signals and noise is realized by mining dense areas in a feature space, so that a new idea is provided for weak signal detection. In the prior art, research has been directed to improving the parameter adaptation problem of density clustering algorithms. For example, the patent of the invention with the grant bulletin number of CN113325383B and the name of self-adaptive vehicle millimeter wave radar clustering algorithm and device based on grids and DBSCAN discloses a method for dynamically calculating local minimum neighborhood radius and adjacent minimum point number according to the density of each grid by gridding a radar data area so as to solve the problems of the traditional DBSCAN algorithm that parameters are static and the clustering effect on multi-density data is poor. However, this prior art still has the following drawbacks related to the technical problem to be solved by the present application: The environmental adaptability is insufficient, the parameter adjustment is only dependent on the static space distribution density of data points, and the dynamic time-varying environmental challenges (such as signal-to-noise ratio changing in real time and non-stationary background noise) faced by the ship network cannot be quantified and responded. The evaluation dimension is single, namely an effective evaluation mechanism for the quality of the input features is lacked, and the running state (such as core point proportion and convergence efficiency) of the clustering process is not monitored. When the feature extraction is poor or the clustering process falls into a suboptimal state, the system cannot perceive. The optimization does not form a closed loop, namely, the adjustment of parameters and the final clustering effect (such as the proportion of separated noise points, the clustering contour coefficient and other indexes which directly reflect the detection performance) do not establish feedback association, and the optimization target and the final detection performance are improved and disjointed. Therefore, although the prior art improves on parameter self-adaption, the self-adaption dimension is single, the mechanism is simple, and the prior art is essentially an open-loop and local parameter adjustment method facing static data space characteristics. Under the complex scene that the signal and noise characteristics of the ship network are both severely and dynamically changed, stable and robust weak signal detection is difficult to realize. Disclosure of Invention The invention aims to provide a ship network weak signal detection method based on density clustering, which aims to solve the problems in the background technology. The method for detecting the weak signals of the ship network based on the density clustering has the characteristics of realizing comprehensive perception, intelligent evaluation and dynamic optimization of the weak signal detection process of the ship network, thereby providing a more reliable and efficient weak signal detection solution in a complex and changeable marine environment. In order to achieve the purpose, the invention provides the technical scheme that the method for detecting the weak signals of the ship network based on density clustering comprises the following steps: s1, acquiring a task difficulty coefficient through a task difficulty model based on signal-to-noise ratio, signal duty ratio and time-frequency diagram background uniformity; S2, acquiring a characteristic quality coefficient through a characteristic quality model based on the total number of the characteristic points, the signal-to-noise ratio distribution of the characteristic points and the characteristic space dimension; s3, acquiring a process state coefficient through a process state model based on the proportion of core points, the iteration times and the final clustering quantity; s4, acquiring a clustering adaptation degree through a clustering adaptation degree model based on the characteristic quality coefficient and the noise point proportion and the clustering quality index under the process state coefficient; s5, acquiring a target neighborhood radius through a neighborhood radius optimization model based on the cluster adaptation degree, the task difficulty coef