Search

CN-122017807-A - State detection method and system for mass production laser radar

CN122017807ACN 122017807 ACN122017807 ACN 122017807ACN-122017807-A

Abstract

The invention relates to the technical field of intelligent sensors and discloses a state detection method and a state detection system for mass production of a laser radar, wherein the method comprises the steps of processing laser radar sensing data and environment sensor data through multi-mode data fusion to obtain an operation feature set; the method comprises the steps of extracting space-time correlation modes by a convolutional neural network, determining boundary distinguishing indexes of environmental interference and state change of equipment, adjusting feature weights by a self-adaptive filtering algorithm to obtain optimized feature fusion vectors, analyzing dynamic sequence changes by a long-term memory network and a short-term memory network, judging fault modes and obtaining state classification probability distribution, updating sensing parameters according to a fault positioning coordinate system, and generating a real-time adjustment instruction sequence by environmental compensation and abnormal propagation analysis. The method can solve the problem of insufficient laser radar state detection precision in the prior art.

Inventors

  • YU YANWU
  • LIU SHENGRONG

Assignees

  • 深圳光秒传感科技有限公司

Dates

Publication Date
20260512
Application Date
20260212

Claims (10)

  1. 1. A state detection method for mass-production lidar, comprising: acquiring space information through a laser radar, acquiring environmental data through an environmental sensor, integrating the acquired data, and performing anomaly correction and consistency verification processing to obtain an operation feature set; Extracting space-time change features from the operation feature set, matching the space-time change features with a pre-established environment interference feature library to obtain an interference mark, determining the boundary between environment interference and the state change of equipment according to the interference mark, and calculating a boundary distinguishing index; if the boundary distinguishing index exceeds a preset distinguishing index threshold, extracting an environment feature subset from the operation feature set and recombining to obtain a feature fusion vector; Acquiring a high-dimensional feature subset according to the feature fusion vector, extracting key node data from the high-dimensional feature subset, comparing the key node data with a pre-established fault mode library, judging a fault mode type according to a comparison result, and calculating state classification probability distribution; dividing a low confidence coefficient region according to the state classification probability distribution, acquiring a feature subset of the low confidence coefficient region, performing deviation correction and space conversion, and determining a fault positioning coordinate system; Updating sensing parameters of the laser radar and collecting data again according to the fault positioning coordinate system to obtain a calibrated running state vector; if the running state vector displays abnormal deviation, compensating the abnormal deviation according to environmental influence, updating the running state vector after compensation and carrying out state detection to obtain a state detection result; Acquiring abnormal points from the state detection result, and determining a comprehensive fault positioning path by carrying out signal comparison and path simulation on the abnormal points; And extracting the environmental data of the fault position according to the comprehensive fault positioning path to obtain an environmental feature set, generating an adjustment instruction according to the environmental feature set, and performing multi-scene adaptation verification and layering processing on the instruction to obtain a real-time adjustment instruction sequence.
  2. 2. The method for detecting the state of a mass production laser radar according to claim 1, wherein the steps of acquiring spatial information by the laser radar, acquiring environmental data by an environmental sensor, integrating the acquired data and performing anomaly correction and consistency check processing to obtain an operation feature set, include: acquiring space information through a laser radar, acquiring environmental data through an environmental sensor, and integrating to obtain initial operation characteristics; If the initial operation characteristic exceeds a preset characteristic threshold range, marking abnormal characteristic data, and determining an environment interference type corresponding to the abnormal characteristic data; according to the environment interference type, corresponding correction parameters are obtained from a pre-established environment interference database, and the abnormal characteristic data are adjusted to obtain an adjustment characteristic data set; And carrying out consistency check on the adjustment characteristic data set, and if the consistency check is passed, determining the adjustment characteristic data set as an operation characteristic set.
  3. 3. The method for detecting the state of the mass production lidar according to claim 1, wherein the steps of extracting the space-time variation feature from the operation feature set and matching with a pre-established environmental interference feature library to obtain an interference mark, determining the boundary between the environmental interference and the state variation of the device itself according to the interference mark, and calculating the boundary distinguishing index include: acquiring signal data according to the operation feature set, and extracting features of the signal data to obtain space-time variation features; Matching the space-time variation characteristics with a pre-established environment interference characteristic library to obtain characteristic matching degree, and judging the characteristic as environment interference dominant characteristics if the characteristic matching degree exceeds a preset characteristic matching threshold value to obtain an interference mark; determining the boundary between the environmental interference and the state change of the equipment according to the interference mark to obtain preliminary boundary division data; and carrying out space-time correlation analysis on the preliminary boundary dividing data, and calculating to obtain a boundary distinguishing index.
  4. 4. The method for detecting the state of the mass production lidar according to claim 1, wherein if the boundary distinguishing indicator exceeds a preset distinguishing indicator threshold, extracting and recombining an environmental feature subset from the operation feature set to obtain a feature fusion vector, comprising: if the boundary distinguishing index exceeds a preset distinguishing index threshold, carrying out layering decomposition on the data content of the operation feature set to obtain an environment feature subset related to environment adaptation; calculating initial feature weights of the environmental feature subsets and performing dynamic adjustment to obtain adjusted weight distribution data; and recombining the environmental feature subsets through the weight distribution data to obtain feature fusion vectors.
  5. 5. The method for detecting the state of the mass production lidar according to claim 1, wherein the obtaining a high-dimensional feature subset according to the feature fusion vector, extracting key node data from the high-dimensional feature subset and comparing the key node data with a pre-established failure mode library, judging a failure mode class according to a comparison result, and calculating a state classification probability distribution comprises: Acquiring a high-dimensional feature subset according to the feature fusion vector, analyzing the dynamic sequence change of the high-dimensional feature subset, and determining the dynamic sequence change as key node data if the change amplitude exceeds a preset change amplitude range; And comparing the key node data with a pre-established fault mode library, determining fault mode types according to comparison results, and calculating state classification probability distribution.
  6. 6. The method for detecting the state of the mass-produced lidar according to claim 1, wherein the classifying the low-confidence regions according to the state classification probability distribution, acquiring feature subsets of the low-confidence regions and performing bias correction and space conversion, and determining a fault location coordinate system comprises: Acquiring feature subset distribution from the state classification probability distribution, and carrying out layering treatment on the feature subset distribution to acquire feature core data; Separating the environmental interference factors from the characteristic core data to obtain a clean data set; If the similarity between the clean data set and the fault mode type is lower than a preset similarity threshold, determining the clean data set as a low confidence region and clustering and grouping the feature subsets of the low confidence region to obtain candidate data clusters; and carrying out deviation correction and space conversion on the candidate data clusters to obtain a fault positioning coordinate system.
  7. 7. The method for detecting the state of the mass production lidar according to claim 1, wherein if the operation state vector shows an abnormal deviation, compensating the abnormal deviation for an environmental impact, updating the operation state vector after the compensation, and performing the state detection to obtain a state detection result, comprising: acquiring abnormal deviation data from the running state vector, judging that abnormal deviation occurs and labeling the abnormal deviation data if the abnormal deviation data exceeds a preset deviation threshold value, and obtaining a labeled deviation data set; acquiring environmental impact data from the deviation data set, and compensating the deviation data set according to the environmental impact data to obtain a compensation data set; And updating the running state vector according to the compensation data set, and carrying out state detection on the updated running state vector to obtain a state detection result.
  8. 8. The method for detecting the state of a mass-produced laser radar according to claim 1, wherein the obtaining an abnormal point from the state detection result, performing signal comparison and path simulation on the abnormal point, and determining a comprehensive fault location path includes: acquiring abnormal points from the state detection result, and classifying the abnormal points to obtain an abnormal point classification set; According to the abnormal point classification set, carrying out signal comparison on the abnormal points by combining with environmental interference factors to obtain a signal attenuation source region; performing path simulation according to the signal attenuation source region to obtain a simulated propagation path, and performing priority ranking on the simulated propagation path to obtain a positioning path set; and carrying out consistency check on the positioning path set, and determining the comprehensive fault positioning path if a consistency check result meets a preset verification standard.
  9. 9. The method for detecting the state of the mass production lidar according to claim 1, wherein the extracting the environmental data of the fault location according to the comprehensive fault location path to obtain an environmental feature set, generating an adjustment instruction according to the environmental feature set, and performing multi-scene adaptation verification and layering processing on the instruction to obtain a real-time adjustment instruction sequence includes: According to the comprehensive fault positioning path, extracting the characteristics of the environmental data of the fault position to obtain an environmental characteristic set; If the characteristic value in the environment characteristic set exceeds a preset environment threshold value, generating an adjustment instruction and determining an instruction sequence; if the instruction sequence passes the multi-scene adaptation verification, obtaining an instruction optimization combination; Layering the instruction optimization combination to obtain a layering result, and if the layering result meets the complex scene adaptation requirement, obtaining a real-time adjustment instruction sequence.
  10. 10. A state detection system for mass-production lidar, comprising: The operation feature acquisition module is used for acquiring space information through a laser radar, acquiring environment data through an environment sensor, integrating the acquired data, and performing anomaly correction and consistency check processing to obtain an operation feature set; the boundary distinguishing module is used for extracting space-time change characteristics from the operation characteristic set and matching the space-time change characteristics with a pre-established environment interference characteristic library to obtain an interference mark, determining the boundary between environment interference and the state change of equipment according to the interference mark, and calculating a boundary distinguishing index; The weight distribution module is used for extracting an environment feature subset from the operation feature set and recombining the environment feature subset to obtain a feature fusion vector if the boundary distinguishing index exceeds a preset distinguishing index threshold; the fault mode identification module is used for acquiring a high-dimensional feature subset according to the feature fusion vector, extracting key node data from the high-dimensional feature subset, comparing the key node data with a pre-established fault mode library, judging a fault mode type according to a comparison result, and calculating state classification probability distribution; The fault coordinate positioning module is used for dividing a low confidence coefficient region according to the state classification probability distribution, acquiring a feature subset of the low confidence coefficient region, carrying out deviation correction and space conversion, and determining a fault positioning coordinate system; The sensing parameter updating module is used for updating the sensing parameters of the laser radar and collecting data again according to the fault positioning coordinate system to obtain a calibrated running state vector; The environment compensation module is used for compensating the abnormal deviation according to the environmental influence if the running state vector displays the abnormal deviation, updating the running state vector after compensation and carrying out state detection to obtain a state detection result; the abnormal point analysis module is used for acquiring abnormal points from the state detection result and determining a comprehensive fault positioning path by carrying out signal comparison and path simulation on the abnormal points; the instruction generation module is used for extracting the environmental data of the fault position according to the comprehensive fault positioning path to obtain an environmental feature set, generating an adjustment instruction according to the environmental feature set, and carrying out multi-scene adaptation verification and layering processing on the instruction to obtain a real-time adjustment instruction sequence.

Description

State detection method and system for mass production laser radar Technical Field The invention relates to the technical field of intelligent sensors, in particular to a state detection method and system for mass production of a laser radar. Background At present, in the field of intelligent sensors, a laser radar is used as a core environment sensing sensor, and the reliability of the running state of the laser radar directly determines the performance and safety of the whole system. In particular, under a complex industrial environment, the laser radar needs to keep stable working under the condition that a plurality of interference factors such as temperature fluctuation, humidity change, dust accumulation and the like coexist, and therefore, higher requirements are provided for real-time monitoring and accurate diagnosis of equipment states. In one prior art, lidar condition detection relies primarily on fixed threshold judgment or statistical analysis based on a single data source. The system judges the state of the equipment by collecting the original perception signal of the laser radar and comparing the original perception signal with a preset static threshold value. For example, when the signal strength continues to fall below a certain fixed threshold, an equipment anomaly alarm is triggered. Another approach attempts to introduce environmental sensors, but simply compares or independently analyzes the perceived data with the environmental data, failing to achieve deep fusion of the multi-source information. When the method is used for dealing with complex and changeable working scenes, similar abnormal performances caused by environmental interference (such as signal attenuation caused by mirror surface condensation) and equipment self faults (such as performance reduction caused by laser aging) are difficult to effectively distinguish. In the prior art, due to the lack of effective fusion and deep feature association analysis of multi-mode data, the boundary between the environmental interference and the state change of the equipment is fuzzy, and accurate fault tracing and state classification are difficult to realize. Meanwhile, the static threshold mechanism and the single-dimension analysis are difficult to adapt to the working environment with dynamic change, so that the accuracy of state evaluation is insufficient, and the timeliness of fault early warning is limited. Therefore, the prior art has the problem of insufficient laser radar state detection precision. Disclosure of Invention The invention provides a state detection method and a state detection system for mass production of a laser radar, which are used for solving the problem of insufficient state detection precision of the laser radar in the prior art. In order to solve the above technical problems, the present invention provides a state detection method for mass-producing a laser radar, including: acquiring space information through a laser radar, acquiring environmental data through an environmental sensor, integrating the acquired data, and performing anomaly correction and consistency verification processing to obtain an operation feature set; Extracting space-time change features from the operation feature set, matching the space-time change features with a pre-established environment interference feature library to obtain an interference mark, determining the boundary between environment interference and the state change of equipment according to the interference mark, and calculating a boundary distinguishing index; if the boundary distinguishing index exceeds a preset distinguishing index threshold, extracting an environment feature subset from the operation feature set and recombining to obtain a feature fusion vector; Acquiring a high-dimensional feature subset according to the feature fusion vector, extracting key node data from the high-dimensional feature subset, comparing the key node data with a pre-established fault mode library, judging a fault mode type according to a comparison result, and calculating state classification probability distribution; dividing a low confidence coefficient region according to the state classification probability distribution, acquiring a feature subset of the low confidence coefficient region, performing deviation correction and space conversion, and determining a fault positioning coordinate system; Updating sensing parameters of the laser radar and collecting data again according to the fault positioning coordinate system to obtain a calibrated running state vector; if the running state vector displays abnormal deviation, compensating the abnormal deviation according to environmental influence, updating the running state vector after compensation and carrying out state detection to obtain a state detection result; Acquiring abnormal points from the state detection result, and determining a comprehensive fault positioning path by carrying out signal comparison and path simulation