CN-121982866-A - Firework and firecracker warehouse monitoring system and method based on scanning radar and vision fusion

CN121982866ACN 121982866 ACN121982866 ACN 121982866ACN-121982866-A

Abstract

The invention discloses a firework and firecracker warehouse monitoring system and method based on scanning radar and vision fusion, and relates to the technical field of warehouse supervision, wherein the system comprises the steps of extracting historical data of the firework and firecracker warehouse monitoring system, determining a scene only containing goods as an initial scene, and dividing two types of characteristic scenes; the method comprises the steps of determining early warning events of high risk blockage in an initial scene, taking a characteristic scene determination moment as the initial moment, taking a target shielding duration time as a monitoring period, collecting and calculating minimum allowed passing width and stacking density under each time stamp in real time, matching the time stamps to form a dynamic monitoring data set, judging whether early warning characteristic values of the time stamps meet early warning conditions, marking the time stamps meeting the conditions as observation nodes, judging that the high risk blockage is caused when all the time periods are observation nodes, and extracting differentiated data of the characteristic scene and the initial scene through a difference analysis algorithm if non-observation nodes exist, so as to generate temporary occupation data characteristics.

Inventors

JIN YONGSHUANG
LI PING
ZHOU BO
ZOU YANG

Assignees

江苏誉德行言智能科技有限公司

Dates

Publication Date: 20260505
Application Date: 20260206

Claims (8)

1. A firework and firecracker warehouse monitoring method based on the integration of scanning radar and vision is characterized by comprising the following steps: step S1, extracting historical data of a firework and firecracker warehouse monitoring system, determining a scene only containing goods as an initial scene, and dividing two types of characteristic scenes of personnel and goods and personnel and cart and goods; Step S2, determining an early warning event of high risk blockage in an initial scene, and defining early warning characteristics and early warning conditions, wherein the early warning characteristics comprise the minimum allowable passage width of a main channel, stacking density and target shielding duration, and the early warning conditions are a threshold value of the minimum allowable passage width less than or equal to the width or a threshold value of the stacking density more than or equal to the density in the target shielding duration; S3, taking the characteristic scene determination moment as an initial moment, collecting and calculating the minimum allowable passing width and stacking density under each time stamp in real time according to the target shielding duration as a monitoring period, and matching the time stamps to form a dynamic monitoring data set; step S4, judging whether the early warning characteristic value of each time stamp meets the early warning condition, marking the time stamp meeting the condition as an observation node, and judging that the time stamp is high-risk blockage if the time stamp is the observation node in the whole period; S5, if a non-observation node exists, extracting differential data of the characteristic scene and the initial scene through a differential analysis algorithm, and generating temporary occupied bit characteristics; and S6, extracting the duration of the observation node, and switching from the temporary occupation early warning state to the high risk blockage early warning state when the duration is more than or equal to the target shielding duration.
2. The method for monitoring the firework and firecracker warehouse based on the integration of scanning radar and vision as claimed in claim 1, wherein the step S1 comprises the following specific processes: Extracting radar point cloud data, visual image data and personnel positioning data of a monitoring system for about 12 months, wherein the radar point cloud data comprises three-dimensional coordinates and volumes of targets, the visual image data comprises contours and texture features of the targets, and the personnel positioning data comprises personnel position coordinates and moving tracks; Performing scene clustering on the historical data by adopting a K-means clustering algorithm, and calibrating an initial scene data set only containing cargoes by taking a 'no personnel positioning signal, no trolley characteristic and only containing cargo point cloud density characteristic' as a clustering condition; Based on a target attribute recognition algorithm, two types of scenes including personnel and goods and personnel and trolley and goods are extracted from non-initial scene data, wherein recognition features are that personnel features with locating signals and visual flexible contours are recorded, trolley features with specific volume ranges are recorded, and goods features with density of point cloud being greater than or equal to a density threshold are recorded.
3. The method for monitoring the firework and firecracker warehouse based on the integration of scanning radar and vision as claimed in claim 1, wherein the step S2 comprises the following specific steps: s21, screening high-risk blocking events from initial scene data, wherein the screening condition is that early warning events continuously exist and no characteristic change exists; step S22, the calculation process of the early warning features is as follows: Based on a channel boundary fitting algorithm of Lei Dadian cloud data, calculating that the distance difference of nearest targets on two sides of a channel is the minimum allowable traffic width W of a main channel, wherein W=W 0 -∑S i , W 0 is the total channel design width, and S i is the occupation width of a single-side target; Calculating stacking density through the volume integral of the point cloud and the physical space ratio, wherein the formula is ρ=V 1 /V 0 , wherein V 1 is the actual occupied volume of goods, and V 0 is the channel verification stacking volume; continuously counting and acquiring a target shielding duration time in minutes through a time stamp; Step S23, based on the initial scene history data, the system sets a width threshold W and a density threshold rho.
4. The method for monitoring a firework and firecracker warehouse based on the combination of scanning radar and vision as claimed in claim 1, wherein the matching time stamp forming the dynamic monitoring data set comprises the following steps: Step S31, collecting data according to the frequency of 1 second/time by taking the time of the identified characteristic scene as T 0 and the monitoring period T=the target shielding duration; Step S32, taking in the occupied space of personnel and a trolley, calculating the minimum allowed passing width W 'and W' =W 0 -(∑S Goods (e.g. a cargo) +∑S Human body +∑S Vehicle with a frame ),S Goods (e.g. a cargo) in a characteristic scene to represent the occupied width of goods, S Human body to represent the occupied width of personnel, S Vehicle with a frame to represent the occupied width of the trolley, and taking a value according to the size of an equipment account, wherein the stacking density in the characteristic scene is the same as the stacking density in an initial scene in a calculation mode; Step S33, the corresponding W' and rho values in the time stamps t0 and t 0+1 s、…、t 0+T are associated to form a dynamic monitoring dataset (t i ,W' i ,ρ i ) }, i is [0, T ].
5. The method for monitoring the firework and firecracker warehouse based on the integration of scanning radar and vision as claimed in claim 1, wherein the step S4 comprises the following steps: judging whether W' i is less than or equal to W or rho i is more than or equal to rho is true for each time stamp t i , and marking the time stamp as an observation node if the time stamp is true; Calculating fluctuation degree r of W' and rho in a monitoring period by adopting a standard deviation algorithm, wherein the formula is shown as r= [ Sigma (x i -μ)²/n] 1/2 ; wherein x i represents characteristic values of each time stamp, mu represents a characteristic value mean value, and n represents the number of time stamps; if all t i in the monitoring period are observation nodes and the output early warning characteristic fluctuation is stable, the high risk blockage is judged to exist in the characteristic scene.
6. The method for monitoring the firework and firecracker warehouse based on the fusion of scanning radar and vision according to claim 1, wherein the step of generating temporary occupation data features comprises the following specific steps: step S51, calculating a W 'difference value delta W=W' Features (e.g. a character) -W' Initial initiation 、ρ Difference value ：Δρ=ρ Features (e.g. a character) -ρ Initial initiation of the characteristic scene and the initial scene under the same time dimension, setting a minimum allowable passing width difference value delta W 0 and a stacking density difference value delta rho 0 , and judging that scene differentiation exists when delta W is more than or equal to delta W 0 and delta rho is less than or equal to delta rho 0 ; And S52, extracting statistical features of the differentiated data, wherein the statistical features comprise a mean value, a variance and a peak value, and constructing a temporary occupied bit feature model.
7. The method for monitoring the firework and firecracker warehouse based on the combination of the scanning radar and the vision as claimed in claim 1, wherein the step S6 comprises the following steps: adopting a continuous timestamp counting algorithm to count the continuous duration T' of the observation node; When T' is not less than T and the temporary occupation bit feature model judges that no differentiation exists, switching from the temporary occupation early warning state to the high risk blockage early warning state; Triggering a first-level alarm after switching, and synchronously outputting early warning characteristic values (W', ρ) and time stamp records.
8. The firework and firecracker warehouse monitoring system based on the scanning radar and vision fusion is characterized by comprising a data acquisition module, a scene division module, an early warning reference establishment module, a dynamic monitoring module, a volatility analysis module, a temporary occupation feature modeling module and a state switching judgment module, wherein the firework and firecracker warehouse monitoring system based on the scanning radar and vision fusion is applied to the firework and firecracker warehouse monitoring method based on the scanning radar and vision fusion of any one of claims 1-7; the data acquisition module is used for acquiring radar point cloud data, visual image data and personnel positioning data, wherein the radar point cloud data comprises target three-dimensional coordinates and volumes, the visual image data comprises target contours and texture features, and the personnel positioning data comprises personnel position coordinates and moving tracks; The scene dividing module is used for extracting historical data of a monitoring system for 12 months, calibrating an initial scene only containing cargoes by adopting a K-means clustering algorithm, and dividing two types of characteristic scenes of personnel and cargoes and personnel and trolley and cargoes by a target attribute recognition algorithm; The early warning standard establishment module is used for screening high risk blocking events from initial scene data, calculating the minimum allowed passing width, stacking density and target shielding duration of a main channel, determining a width threshold and a density threshold according to a 3 sigma statistical principle, and setting early warning conditions as the minimum allowed passing width less than or equal to the width threshold or the stacking density more than or equal to the density threshold in the target shielding duration; The dynamic monitoring module is used for setting a monitoring period T by taking the characteristic scene determination moment as an initial moment, collecting data according to the frequency of 1 second/time, calculating the actual passing width of the occupied space of an incorporating person and a trolley, and constructing a dynamic monitoring data set; the fluctuation analysis module is used for judging whether the early warning characteristic value of each time stamp meets the early warning condition, marking the time stamp meeting the condition as an observation node, quantifying the characteristic fluctuation degree by adopting a standard deviation algorithm, and judging that the system is blocked at high risk when all the time stamps are the observation nodes and sigma is less than or equal to 0.05; the temporary occupation feature modeling module is used for calculating the feature difference value of the feature scene and the initial scene by adopting a contrast analysis algorithm when a non-observation node exists, judging whether scene differentiation exists or not based on the feature difference value, and constructing a temporary occupation feature model; The state switching judging module is used for counting the continuous duration of the observation node by adopting a continuous timestamp counting algorithm, and when the continuous duration is more than or equal to 15 minutes and the temporary occupation data characteristic model judges that no differentiation exists, the state switching judging module is switched from the temporary occupation early-warning state to the high risk blockage early-warning state, and triggers the primary alarm and outputs the early-warning characteristic value and the timestamp record.

Description

Firework and firecracker warehouse monitoring system and method based on scanning radar and vision fusion Technical Field The invention relates to the field of monitoring and early warning of safety production risk of fireworks and crackers, in particular to a system and a method for monitoring a warehouse of fireworks and crackers based on integration of scanning radar and vision. Background The smoothness of the channel of the firework and firecracker warehouse is the key for ensuring emergency evacuation and rescue, the existing channel blockage monitoring system mainly carries out early warning based on the goods stacking state, and the blockage situation is judged by monitoring static characteristics such as stacking side length, channel width and the like. However, in warehouse dynamic operations, personnel, carts and goods frequently coexist, which results in the following core drawbacks in the prior art: The scene distinguishing is missing, namely, scenes such as 'only cargoes', 'personnel+cargoes', 'personnel+carts+cargoes' are not divided, temporary occupation of the personnel and the carts is misjudged as high-risk blockage, or high-risk blockage and leakage judgment are caused by the fact that space is temporarily occupied is not considered; judging the blocking state only by the characteristic value of a single time point, and not analyzing the fluctuation of the early warning characteristic in the monitoring period, so that the 'temporarily occupied dynamic change' and the 'high risk blocking stable state' cannot be distinguished; the differential features are not mined, namely differential data features of temporary occupation and high risk blockage are not extracted, so that the judgment of two states lacks quantitative basis, and the risks of false alarm and missing alarm are high. Therefore, a monitoring method based on scene division and dynamic feature analysis is needed to accurately and quantitatively distinguish temporary occupation from high risk blockage, and improve monitoring reliability. Disclosure of Invention The invention aims to provide a firework and firecracker warehouse monitoring system and method based on scanning radar and vision fusion, so as to solve the problems in the prior art. In order to achieve the purpose, the invention provides the following technical scheme that the method for monitoring the firework and firecracker warehouse based on the integration of scanning radar and vision comprises the following steps: step S1, extracting historical data of a firework and firecracker warehouse monitoring system, determining a scene only containing goods as an initial scene, and dividing two types of characteristic scenes of personnel and goods and personnel and cart and goods; Step S2, determining an early warning event of high risk blockage in an initial scene, and defining early warning characteristics and early warning conditions, wherein the early warning characteristics comprise the minimum allowable passage width of a main channel, stacking density and target shielding duration, and the early warning conditions are that the minimum allowable passage width in the target shielding duration is less than or equal to a width threshold value or the stacking density is more than or equal to a density threshold value; S3, taking the characteristic scene determination moment as an initial moment, collecting and calculating the minimum allowable passing width and stacking density under each time stamp in real time according to the target shielding duration as a monitoring period, and matching the time stamps to form a dynamic monitoring data set; step S4, judging whether the early warning characteristic value of each time stamp meets the early warning condition, marking the time stamp meeting the condition as an observation node, and judging that the time stamp is high-risk blockage if the time stamp is the observation node in the whole period; S5, if a non-observation node exists, extracting differential data of the characteristic scene and the initial scene through a differential analysis algorithm, and generating temporary occupied bit characteristics; and S6, extracting the duration of the observation node, and switching from the temporary occupation early warning state to the high risk blockage early warning state when the duration is more than or equal to the target shielding duration. Further, the step S1 includes the following specific procedures: Extracting radar point cloud data, visual image data and personnel positioning data of a monitoring system for about 12 months, wherein the radar point cloud data comprises three-dimensional coordinates and volumes of targets, the visual image data comprises contours and texture features of the targets, and the personnel positioning data comprises personnel position coordinates and moving tracks; Performing scene clustering on the historical data by adopting a K-means clustering algorithm, and calibrating an initial scene