Intelligent sound system and method

Abstract

The invention relates to the technical field of intelligent sound control, in particular to an intelligent sound system and method. Obtaining a first crossing mode, a second crossing mode and a third crossing mode according to the state of the crossing sliding door; identifying first crossing data and first movement data of the first crossing mode to obtain a first abnormal area and a first abnormal coefficient, and determining first orientation data by combining a first distance between the first abnormal area and a crossing sliding door; identifying second crossing data of a second crossing mode to obtain a second abnormal region and a second abnormal coefficient; determining second orientation data according to a second distance between the second abnormal area and the crossing sliding door and a second abnormal coefficient; third crossing data and the second moving data of the third crossing mode are recognized, a third abnormal area and a third abnormal coefficient are obtained, and third orientation data are determined by combining the third abnormal area with a third distance between the crossing sliding door. According to the invention, intelligent control of the sound equipment at the railway crossing is realized through orientation data of different modes.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent audio control, and in particular to an intelligent audio system and method. Background Art

[0002] A railroad crossing is a place where a railway meets a road, street or other traffic route. It is also called a railroad level crossing. In the rail transportation system, many places are not suitable for building bridges or tunnels due to geographical or economic reasons; therefore, railroad crossings become an inevitable choice. Railroad crossings are usually equipped with traffic signs, warning signals and control facilities to ensure the safety of rail transportation and other modes of transportation.

[0003] With the development of railway transportation, the safety of railway crossings has received more and more attention. In the existing technology, railway crossing alarm systems usually rely on sound alarms to alert pedestrians and vehicles; however, since the sound equipment used for early warning is generally in a fixed direction, it is not possible to make personalized adjustments to the sound direction according to the control status of the railway crossing, the running status of the train, the traffic conditions, and the behavior of personnel, and it is impossible to focus on areas with obvious abnormalities, resulting in unsatisfactory early warning effects.

[0004] With the development of data technology, intelligent algorithms can be used to provide dynamic and flexible regional early warning reminders based on train driving parameters, real-time traffic conditions at crossings, and personnel behavior, which can improve traffic safety and road efficiency and avoid the safety hazards of traditional early warning methods.

[0005] To this end, an intelligent audio system and method are proposed. Summary of the invention

[0006] The purpose of the present invention is to provide an intelligent sound system and method; obtain a first crossing mode, a second crossing mode and a third crossing mode according to the state of a crossing sliding door; identify the first crossing data and the first movement data of the first crossing mode, obtain a first abnormal area and a first abnormal coefficient, and then determine the first orientation data in combination with the first distance between the first abnormal area and the crossing sliding door; identify the second crossing data of the second crossing mode, obtain the second abnormal area and the second abnormal coefficient; determine the second orientation data according to the second distance between the second abnormal area and the crossing sliding door and the second abnormal coefficient; identify the third crossing data and the second movement data of the third crossing mode, obtain the third abnormal area and the third abnormal coefficient, and then determine the third orientation data in combination with the third distance between the third abnormal area and the crossing sliding door. The present invention realizes intelligent control of railway crossing sound equipment through orientation data of different modes.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] An intelligent audio system, comprising:

[0009] A crossing mode recognition module recognizes the crossing sliding door of the railway crossing, and divides the crossing sliding door into a first crossing mode, a second crossing mode and a third crossing mode according to the state of the crossing sliding door;

[0010] A first pattern analysis module is configured to obtain first crossing data and first movement data in the first crossing pattern; construct a first pattern recognition model to recognize the first crossing data and the first movement data, and obtain a first abnormal area and a first abnormal coefficient; obtain a first distance between the first abnormal area and the crossing sliding door, and determine first orientation data of the audio device according to the first distance and the first abnormal coefficient;

[0011] A second pattern analysis module is used to obtain the second crossing data in the second crossing pattern; a second pattern recognition model is constructed to recognize the second crossing data, and the second abnormal area and the second abnormal coefficient are obtained; a second distance between the second abnormal area and the crossing sliding door is obtained, and the second orientation data of the audio device is determined according to the second distance and the second abnormal coefficient;

[0012] A third mode analysis module is configured to obtain the third crossing data and the second movement data in the third crossing mode; construct a third mode recognition model to recognize the third crossing data and the second movement data, and obtain a third abnormal area and a third abnormal coefficient; obtain a third distance between the third abnormal area and the crossing sliding door, and determine the third orientation data of the audio device according to the third distance and the third abnormal coefficient;

[0013] The audio intelligent control module controls the audio device according to the first orientation data, the second orientation data and the third orientation data.

[0014] In the crossing mode recognition module, the division process of the first crossing mode, the second crossing mode and the third crossing mode is: the first crossing mode is the process of the crossing sliding door from opening to closing; the second crossing mode is that the crossing sliding door is in a closed state; the third crossing mode is the process of the crossing sliding door from closing to opening.

[0015] The first pattern recognition model includes a first data acquisition layer, a first crossing data recognition layer and a first anomaly recognition layer;

[0016] The first data acquisition layer acquires first crossing data and first movement data; the first crossing data is surveillance video data of the railway crossing during the first crossing mode; the first movement data includes the closing speed of the crossing sliding door and the first passable width in the first crossing mode;

[0017] The first crossing data recognition layer identifies and obtains first individual pass data and first traffic environment data in the first crossing data; the first individual pass data includes first individual pass type, first individual pass speed, first individual pass distance and first traffic congestion data; the first traffic environment data includes environmental visibility, road surface slipperiness, road surface flatness and environmental noise data in the first crossing mode;

[0018] The first abnormality recognition layer obtains the first abnormality coefficient of each passing individual according to the first passing individual data, the first passing environment data and the first movement data; obtains the first abnormal area according to the clustering of the passing individual position and the first passing abnormality coefficient; and obtains the first abnormality coefficient according to the first passing abnormality coefficient of the passing individual in the first abnormal area.

[0019] The process of determining the first orientation data according to the first distance and the first anomaly coefficient is:

[0020] A three-dimensional coordinate system is constructed with the location of the audio equipment as the origin; wherein the z-axis of the three-dimensional coordinate system is a vertical line of the ground plane passing through the origin;

[0021] According to the connection line between the first abnormal area and the audio device, first candidate orientation data of the audio device is obtained; the first candidate orientation data is vector data;

[0022] According to the first distance between the first abnormal area and the crossing sliding door and the first abnormal coefficient of the first abnormal area, the first orientation weight of the first alternative orientation data is calculated; and the first orientation data is obtained according to all the first alternative orientation data and the corresponding first orientation weights.

[0023] The second pattern recognition model includes a second data acquisition layer, a second crossing data recognition layer, and a second anomaly recognition layer;

[0024] The second data acquisition layer acquires second crossing data; the second crossing data is monitoring video data of the railway crossing during the second crossing mode;

[0025] The second crossing data recognition layer performs sound preprocessing on the second crossing data to obtain second preprocessed crossing data; identifies and obtains individuals with abnormal sound and individuals with abnormal behavior in the second preprocessed crossing data; obtains sound abnormality coefficients according to the sound type, sound decibel and sound frequency of the individuals with abnormal sound; obtains behavior abnormality coefficients according to the type, passing speed and behavior type of the individuals with abnormal behavior;

[0026] The second abnormality identification layer clusters the position data of individuals with abnormal voices, the abnormal voice coefficient, the position data and the abnormal behavior coefficient of individuals with abnormal behavior to obtain a second abnormal area; and obtains the second abnormality coefficient based on the abnormal voice coefficient of individuals with abnormal voices and the abnormal behavior coefficient of individuals with abnormal behavior in the second abnormal area.

[0027] In the second crossing data recognition layer, the process of sound preprocessing of the second crossing data is: according to the distance between the train and the railway crossing and the sound data of the train passing, the environmental noise caused by the passage of the train in the second crossing data is processed to obtain the second preprocessed crossing data.

[0028] The third pattern recognition model includes a third data acquisition layer, a third crossing data recognition layer and a third anomaly recognition layer;

[0029] The third data acquisition layer acquires third crossing data and second movement data; the third crossing data is monitoring video data of the railway crossing during the third crossing mode; the second movement data includes the opening speed of the crossing sliding door and the second passable width in the third crossing mode;

[0030] The third crossing data recognition layer recognizes the third crossing data and obtains third individual traffic data and third traffic environment data in the third crossing data; the third individual traffic data includes third individual traffic type, third traffic speed, third individual traffic distance and third traffic congestion data; the third traffic environment data includes environmental visibility, road surface slipperiness, road surface flatness and environmental noise data in the third crossing mode;

[0031] The third anomaly identification layer obtains the third anomaly coefficient of each passing individual based on the third passing individual data, the third passing environment data and the second movement data; obtains the third anomaly area based on the clustering of the passing individual positions and the third anomaly coefficients; and obtains the third anomaly coefficient based on the third anomaly coefficients of the passing individuals in the third anomaly area.

[0032] A smart audio method, comprising:

[0033] Identify the sliding gates at railway crossings, and divide them into the first crossing mode, the second crossing mode, and the third crossing mode according to the states of the sliding gates;

[0034] Acquire the first crossing data and the first movement data in the first crossing mode; construct a first pattern recognition model to recognize the first crossing data and the first movement data, and obtain a first abnormal area and a first abnormal coefficient; acquire a first distance between the first abnormal area and the crossing sliding door, and determine the first orientation data of the audio device according to the first distance and the first abnormal coefficient;

[0035] Acquire the second crossing data in the second crossing mode; construct a second pattern recognition model to recognize the second crossing data, and obtain a second abnormal area and a second abnormal coefficient; obtain a second distance between the second abnormal area and the crossing sliding door, and determine the second orientation data of the audio device according to the second distance and the second abnormal coefficient;

[0036] Obtain the third crossing data and the second movement data in the third crossing mode; construct a third pattern recognition model to identify the third crossing data and the second movement data, and obtain a third abnormal area and a third abnormal coefficient; obtain a third distance between the third abnormal area and the crossing sliding door, and determine the third orientation data of the audio device according to the third distance and the third abnormal coefficient;

[0037] The audio device is controlled according to the first orientation data, the second orientation data and the third orientation data.

[0038] The division process of the first crossing mode, the second crossing mode and the third crossing mode is: the first crossing mode is the process of the crossing sliding door from opening to closing; the second crossing mode is that the crossing sliding door is in a closed state; the third crossing mode is the process of the crossing sliding door from closing to opening.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] 1. The present invention identifies the state of the sliding door of a railway crossing, takes the process of the sliding door from opening to closing as the first crossing mode, takes the sliding door in the closed state as the second crossing mode, and takes the process of the sliding door from closing to opening as the third crossing mode; the first crossing mode, the second crossing mode, and the third crossing mode are used to accurately divide the stages of the process of a train passing through a railway crossing, providing support for subsequent abnormal area identification and early warning.

[0041] 2. The present invention constructs a first pattern recognition model to identify the first crossing data and the first movement data; firstly, the first passing individual data and the first passing environment data are obtained according to the first crossing data; the first passing abnormality coefficient of each passing individual is obtained by identifying the first passing individual data, the first passing environment data and the first movement data; the first abnormal area is obtained by clustering according to the passing individual position and the first passing abnormality coefficient; the first abnormality coefficient is obtained according to the first passing abnormality coefficient of the passing individual in the first abnormal area; the abnormal area and the degree of abnormality during the first crossing mode are accurately measured through the first abnormal area and the first abnormality coefficient.

[0042] 3. The present invention constructs a second pattern recognition model to identify the second crossing data; based on the second crossing data, individuals with abnormal sound passing, abnormal sound coefficients, individuals with abnormal behavior passing and abnormal behavior coefficients are obtained; based on the position data of individuals with abnormal sound passing, abnormal sound coefficients, position data of individuals with abnormal behavior passing and abnormal behavior coefficients, clustering is performed to obtain a second abnormal area; based on the sound abnormality coefficients of individuals with abnormal sound passing and the abnormal behavior coefficients of individuals with abnormal behavior passing in the second abnormal area, the second abnormal coefficient is obtained; through the second abnormal area and the second abnormal coefficient, the abnormal area and the degree of abnormality during the second crossing mode are accurately measured.

[0043] 4. The present invention constructs a third pattern recognition model to identify the third crossing data and the second movement data; firstly, the third passing individual data and the third passing environment data are obtained according to the third crossing data; the third passing abnormality coefficient of each passing individual is obtained by the third passing individual data, the third passing environment data and the second movement data; the third abnormal area is obtained by clustering the passing individual positions and the third passing abnormality coefficient; the third abnormality coefficient is obtained according to the third passing abnormality coefficient of the passing individuals in the third abnormal area; the abnormal area and the degree of abnormality during the third crossing mode are accurately measured by the third abnormal area and the third abnormality coefficient.

[0044] 5. The present invention uses the location of the audio equipment as the origin to construct a three-dimensional coordinate system; obtains alternative orientation data of the audio equipment based on the line between the abnormal area and the audio equipment; calculates the orientation weight of the alternative orientation data based on the distance between the abnormal area and the crossing sliding door and the abnormal coefficient of the abnormal area; obtains orientation data based on all the alternative orientation data and the corresponding orientation weights; and according to the orientation data, the warning focus of the audio equipment can accurately cover the abnormal area, effectively improving the warning effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 A schematic diagram of the structure of an intelligent audio system of the present invention;

[0046] Figure 2 is a structural schematic diagram of a first pattern recognition model of the present invention;

[0047] Figure 3 A schematic diagram of a flow chart of an intelligent audio method of the present invention;

[0048] Figure 4 A schematic diagram of a highway traffic scene at a railway crossing of the present invention;

[0049] Figure 5 It is a schematic diagram of the rail traffic scene at the railway crossing of the present invention. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] Embodiment 1

[0052] The present invention provides an intelligent sound system, the structure of which is as follows Figure 1 As shown; it includes: a crossing pattern recognition module, a first pattern analysis module, a second pattern analysis module, a third pattern analysis module and an audio intelligent control module.

[0053] The crossing mode recognition module recognizes the crossing sliding door of the railway crossing, and obtains the first crossing mode, the second crossing mode and the third crossing mode according to the state of the crossing sliding door.

[0054] In the crossing mode recognition module, the division process of the first crossing mode, the second crossing mode and the third crossing mode is: the first crossing mode is the process of the crossing sliding door from opening to closing; the second crossing mode is that the crossing sliding door is in a closed state; the third crossing mode is the process of the crossing sliding door from closing to opening.

[0055] The present invention identifies the state of a sliding door at a railway crossing, and uses the process of the sliding door being opened to closed as a first crossing mode, the sliding door being in a closed state as a second crossing mode, and the process of the sliding door being closed to open as a third crossing mode; the first crossing mode, the second crossing mode, and the third crossing mode are used to accurately divide the stages of a train passing through a railway crossing, thereby providing support for subsequent abnormal area identification.

[0056] The first pattern analysis module obtains the first crossing data and the first movement data in the first crossing pattern; constructs a first pattern recognition model to recognize the first crossing data and the first movement data to obtain a first abnormal area and a first abnormal coefficient; obtains a first distance between the first abnormal area and the crossing sliding door, and determines the first orientation data of the audio equipment according to the first distance and the first abnormal coefficient.

[0057] The first pattern recognition model is constructed based on a deep neural network, and its structure is as follows: Figure 2 As shown; including a first data acquisition layer, a first crossing data recognition layer and a first abnormality recognition layer;

[0058] The first data acquisition layer acquires first crossing data and first movement data; the first crossing data is surveillance video data of the railway crossing during the first crossing mode; the first movement data includes the closing speed of the crossing sliding door and the first passable width in the first crossing mode;

[0059] The first crossing data recognition layer identifies and obtains first individual pass data and first traffic environment data in the first crossing data; the first individual pass data includes first individual pass type, first individual pass speed, first individual pass distance and first traffic congestion data; the first traffic environment data includes environmental visibility, road surface slipperiness, road surface flatness and environmental noise data in the first crossing mode;

[0060] Among them, the first type of passing individuals includes pedestrians, electric vehicles, cars, trucks and tankers, etc.; the process of obtaining the first traffic congestion data is: taking the passing individual as the center of the circle, determining the circular area according to the set radius threshold, and determining the first traffic congestion data according to the number and type of passing individuals in the circular area; the environmental noise data is obtained through a noise measuring instrument.

[0061] The first abnormality recognition layer obtains the first abnormality coefficient of each passing individual according to the first passing individual data, the first passing environment data and the first movement data; obtains the first abnormal area according to the clustering of the passing individual position and the first passing abnormality coefficient; and obtains the first abnormality coefficient according to the first passing abnormality coefficient of the passing individual in the first abnormal area.

[0062] The present invention constructs a first pattern recognition model to identify the first crossing data and the first movement data; firstly, the first passing individual data and the first passing environment data are obtained according to the first crossing data; the first passing abnormality coefficient of each passing individual is obtained by identifying the first passing individual data, the first passing environment data and the first movement data; the first abnormal area is obtained by clustering according to the passing individual position and the first passing abnormality coefficient; the first abnormality coefficient is obtained according to the first passing abnormality coefficient of the passing individual in the first abnormal area; the abnormal area and the degree of abnormality during the first crossing mode are accurately measured through the first abnormal area and the first abnormality coefficient.

[0063] The process of determining the first orientation data according to the first distance and the first anomaly coefficient is:

[0064] A three-dimensional coordinate system is constructed with the location of the audio equipment as the origin; wherein the z-axis of the three-dimensional coordinate system is a vertical line of the ground plane passing through the origin;

[0065] According to the connection line between the first abnormal area and the audio device, first candidate orientation data of the audio device is obtained; the first candidate orientation data is vector data;

[0066] According to the first distance between the first abnormal area and the crossing sliding door and the first abnormal coefficient of the first abnormal area, the first orientation weight of the first alternative orientation data is calculated; and the first orientation data is obtained according to all the first alternative orientation data and the corresponding first orientation weights.

[0067] The present invention constructs a three-dimensional coordinate system with the position of the audio equipment as the origin; obtains the first alternative orientation data of the audio equipment according to the connection line between the first abnormal area and the audio equipment; calculates the first orientation weight of the first alternative orientation data according to the first distance between the first abnormal area and the crossing sliding door and the first abnormal coefficient of the first abnormal area; obtains the first orientation data according to all the first alternative orientation data and the corresponding first orientation weights; and according to the first orientation data, the warning focus of the audio equipment can accurately cover the abnormal area, effectively improving the warning effect.

[0068] The second pattern analysis module obtains the second crossing data in the second crossing pattern; constructs a second pattern recognition model to identify the second crossing data, and obtains the second abnormal area and the second abnormal coefficient; obtains the second distance between the second abnormal area and the crossing sliding door, and determines the second orientation data of the audio equipment according to the second distance and the second abnormal coefficient.

[0069] The second pattern recognition model includes a second data acquisition layer, a second crossing data recognition layer, and a second anomaly recognition layer;

[0070] The second data acquisition layer acquires second crossing data; the second crossing data is monitoring video data of the railway crossing during the second crossing mode;

[0071] The second crossing data recognition layer performs sound preprocessing on the second crossing data to obtain second preprocessed crossing data; identifies and obtains individuals with abnormal sound and individuals with abnormal behavior in the second preprocessed crossing data; obtains sound abnormality coefficients according to the sound type, sound decibel and sound frequency of the individuals with abnormal sound; obtains behavior abnormality coefficients according to the type, passing speed and behavior type of the individuals with abnormal behavior;

[0072] The sound types include the sound of vehicle horns, the sound of quarrels among people, etc.; the abnormal behaviors include pedestrians approaching, climbing over the sliding gate of the crossing, vehicles overtaking, vehicles approaching the sliding gate of the crossing, etc.

[0073] The second abnormality identification layer clusters the position data of individuals with abnormal voices, the abnormal voice coefficient, the position data and the abnormal behavior coefficient of individuals with abnormal behavior to obtain a second abnormal area; and obtains the second abnormality coefficient based on the abnormal voice coefficient of individuals with abnormal voices and the abnormal behavior coefficient of individuals with abnormal behavior in the second abnormal area.

[0074] The present invention constructs a second pattern recognition model to identify the second crossing data; based on the second crossing data, individuals with abnormal sound passing, abnormal sound coefficients, individuals with abnormal behavior passing and abnormal behavior coefficients are obtained; based on the position data of individuals with abnormal sound passing, the abnormal sound coefficients, the position data of individuals with abnormal behavior passing and the abnormal behavior coefficients, clustering is performed to obtain a second abnormal area; based on the abnormal sound coefficients of individuals with abnormal sound passing and the abnormal behavior coefficients of individuals with abnormal behavior passing in the second abnormal area, a second abnormal coefficient is obtained; and through the second abnormal area and the second abnormal coefficient, the abnormal area and the degree of abnormality during the second crossing mode are accurately measured.

[0075] In the second crossing data recognition layer, the process of sound preprocessing of the second crossing data is: according to the distance between the train and the railway crossing and the sound data of the train passing, the environmental noise caused by the passage of the train in the second crossing data is processed to obtain the second preprocessed crossing data.

[0076] The third mode analysis module obtains the third crossing data and the second movement data in the third crossing mode; constructs a third mode recognition model to recognize the third crossing data and the second movement data to obtain a third abnormal area and a third abnormal coefficient; obtains a third distance between the third abnormal area and the crossing sliding door, and determines the third orientation data of the audio equipment according to the third distance and the third abnormal coefficient.

[0077] The third pattern recognition model includes a third data acquisition layer, a third crossing data recognition layer and a third anomaly recognition layer;

[0078] The third data acquisition layer acquires third crossing data and second movement data; the third crossing data is monitoring video data of the railway crossing during the third crossing mode; the second movement data includes the opening speed of the crossing sliding door and the second passable width in the third crossing mode;

[0079] The third crossing data recognition layer recognizes the third crossing data and obtains third individual traffic data and third traffic environment data in the third crossing data; the third individual traffic data includes third individual traffic type, third traffic speed, third individual traffic distance and third traffic congestion data; the third traffic environment data includes environmental visibility, road surface slipperiness, road surface flatness and environmental noise data in the third crossing mode;

[0080] The third anomaly identification layer obtains the third anomaly coefficient of each passing individual based on the third passing individual data, the third passing environment data and the second movement data; obtains the third anomaly area based on the clustering of the passing individual positions and the third anomaly coefficients; and obtains the third anomaly coefficient based on the third anomaly coefficients of the passing individuals in the third anomaly area.

[0081] The present invention constructs a third mode recognition model to identify the third crossing data and the second movement data; firstly, the third passing individual data and the third passing environment data are obtained according to the third crossing data identification; the third passing abnormality coefficient of each passing individual is obtained by the third passing individual data, the third passing environment data and the second movement data identification; the third abnormal area is obtained by clustering according to the passing individual position and the third passing abnormality coefficient; the third abnormality coefficient is obtained according to the third passing abnormality coefficient of the passing individual in the third abnormal area; the abnormal area and the degree of abnormality during the third crossing mode are accurately measured through the third abnormal area and the third abnormality coefficient.

[0082] The audio intelligent control module controls the audio device according to the first orientation data, the second orientation data and the third orientation data.

[0083] The present invention also proposes an intelligent audio method, the process of which is as follows: Figure 3 As shown, including:

[0084] Identify the sliding gates at railway crossings, and divide them into the first crossing mode, the second crossing mode, and the third crossing mode according to the states of the sliding gates;

[0085] Acquire the first crossing data and the first movement data in the first crossing mode; construct a first pattern recognition model to recognize the first crossing data and the first movement data, and obtain a first abnormal area and a first abnormal coefficient; acquire a first distance between the first abnormal area and the crossing sliding door, and determine the first orientation data of the audio device according to the first distance and the first abnormal coefficient;

[0086] Acquire the second crossing data in the second crossing mode; construct a second pattern recognition model to recognize the second crossing data, and obtain a second abnormal area and a second abnormal coefficient; obtain a second distance between the second abnormal area and the crossing sliding door, and determine the second orientation data of the audio device according to the second distance and the second abnormal coefficient;

[0087] Obtain the third crossing data and the second movement data in the third crossing mode; construct a third pattern recognition model to identify the third crossing data and the second movement data, and obtain a third abnormal area and a third abnormal coefficient; obtain a third distance between the third abnormal area and the crossing sliding door, and determine the third orientation data of the audio device according to the third distance and the third abnormal coefficient;

[0088] The audio device is controlled according to the first orientation data, the second orientation data and the third orientation data.

[0089] The present application obtains the first crossing mode, the second crossing mode and the third crossing mode according to the state of the crossing sliding door; identifies the first crossing data and the first movement data of the first crossing mode, obtains the first abnormal area and the first abnormal coefficient, and then determines the first orientation data in combination with the first distance between the first abnormal area and the crossing sliding door; identifies the second crossing data of the second crossing mode, obtains the second abnormal area and the second abnormal coefficient; determines the second orientation data according to the second distance between the second abnormal area and the crossing sliding door and the second abnormal coefficient; identifies the third crossing data and the second movement data of the third crossing mode, obtains the third abnormal area and the third abnormal coefficient, and then determines the third orientation data in combination with the third distance between the third abnormal area and the crossing sliding door. The present invention realizes intelligent control of railway crossing sound equipment through orientation data of different modes.

[0090] Embodiment 2

[0091] The intelligent sound system of the present invention is used to control the sound of an actual railway crossing; the railway crossing is the intersection of a highway and a railway, and the highway traffic scene is as follows Figure 4 As shown in Figure 2, the rail transit scene is as follows: Figure 5 shown.

[0092] The present invention provides an intelligent sound system, the structure of which is as follows Figure 1 As shown; it includes: a crossing pattern recognition module, a first pattern analysis module, a second pattern analysis module, a third pattern analysis module and an audio intelligent control module.

[0093] The crossing mode recognition module recognizes the crossing sliding door of the railway crossing, and obtains the first crossing mode, the second crossing mode and the third crossing mode according to the state of the crossing sliding door.

[0094] In the crossing mode recognition module, the division process of the first crossing mode, the second crossing mode and the third crossing mode is: the first crossing mode is the process of the crossing sliding door from opening to closing; the second crossing mode is that the crossing sliding door is in a closed state; the third crossing mode is the process of the crossing sliding door from closing to opening.

[0095] The present invention identifies the state of a sliding door at a railway crossing, and uses the process of the sliding door being opened to closed as a first crossing mode, the sliding door being in a closed state as a second crossing mode, and the process of the sliding door being closed to open as a third crossing mode; the first crossing mode, the second crossing mode, and the third crossing mode are used to accurately divide the stages of a train passing through a railway crossing, thereby providing support for subsequent abnormal area identification and early warning.

[0096] The first pattern analysis module obtains the first crossing data and the first movement data in the first crossing pattern; constructs a first pattern recognition model to recognize the first crossing data and the first movement data to obtain a first abnormal area and a first abnormal coefficient; obtains a first distance between the first abnormal area and the crossing sliding door, and determines the first orientation data of the audio equipment according to the first distance and the first abnormal coefficient.

[0097] The first pattern recognition model is constructed based on a deep neural network, and its structure is as follows: Figure 2 As shown; including a first data acquisition layer, a first crossing data recognition layer and a first abnormality recognition layer;

[0098] The first data acquisition layer acquires first crossing data and first movement data; the first crossing data is surveillance video data of the railway crossing during the first crossing mode; the first movement data includes the closing speed of the crossing sliding door and the first passable width in the first crossing mode;

[0099] The first crossing data recognition layer identifies and obtains first individual pass data and first traffic environment data in the first crossing data; the first individual pass data includes first individual pass type, first individual pass speed, first individual pass distance and first traffic congestion data; the first traffic environment data includes environmental visibility, road surface slipperiness, road surface flatness and environmental noise data in the first crossing mode;

[0100] Among them, the first passing individual type includes pedestrians, electric vehicles, cars, trucks and tankers, etc.; the first passing individual distance is the distance between the passing individual and the sliding door of the crossing; the process of obtaining the first traffic congestion data is: taking the passing individual as the center of the circle, determining the circular area according to the set radius threshold, and determining the first traffic congestion data according to the number and type of passing individuals in the circular area; the environmental noise data is obtained through a noise measuring instrument.

[0101] The first individual data of the first crossing mode at the start time is identified, and the data obtained are shown in Table 1.

[0102] Table 1 First pass individual data table

[0103] First pass individual number First pass individual type First passing individual speed First pass individual distance S01 Battery car 2.24m / s 16.25m S02 car 4.58m / s 18.34m S03 car 5.24m / s 15.26m S04 bus 3.38m / s 10.85m

[0104] The first abnormality recognition layer obtains the first abnormality coefficient of each passing individual according to the first passing individual data, the first passing environment data and the first movement data; obtains the first abnormal area according to the clustering of the passing individual position and the first passing abnormality coefficient; and obtains the first abnormality coefficient according to the first passing abnormality coefficient of the passing individual in the first abnormal area.

[0105] The present invention constructs a first pattern recognition model to identify the first crossing data and the first movement data; firstly, the first passing individual data and the first passing environment data are obtained according to the first crossing data; the first passing abnormality coefficient of each passing individual is obtained by identifying the first passing individual data, the first passing environment data and the first movement data; the first abnormal area is obtained by clustering according to the passing individual position and the first passing abnormality coefficient; the first abnormality coefficient is obtained according to the first passing abnormality coefficient of the passing individual in the first abnormal area; the abnormal area and the degree of abnormality during the first crossing mode are accurately measured through the first abnormal area and the first abnormality coefficient.

[0106] The process of determining the first orientation data according to the first distance and the first anomaly coefficient is:

[0107] A three-dimensional coordinate system is constructed with the location of the audio equipment as the origin; wherein the z-axis of the three-dimensional coordinate system is a vertical line of the ground plane passing through the origin;

[0108] According to the connection line between the first abnormal area and the audio device, first candidate orientation data of the audio device is obtained; the first candidate orientation data is vector data;

[0109] According to the first distance between the first abnormal area and the crossing sliding door and the first abnormal coefficient of the first abnormal area, the first orientation weight of the first alternative orientation data is calculated; and the first orientation data is obtained according to all the first alternative orientation data and the corresponding first orientation weights.

[0110] The present invention constructs a three-dimensional coordinate system with the position of the audio equipment as the origin; obtains the first alternative orientation data of the audio equipment according to the connection line between the first abnormal area and the audio equipment; calculates the first orientation weight of the first alternative orientation data according to the first distance between the first abnormal area and the crossing sliding door and the first abnormal coefficient of the first abnormal area; obtains the first orientation data according to all the first alternative orientation data and the corresponding first orientation weights; and according to the first orientation data, the warning focus of the audio equipment can accurately cover the abnormal area, effectively improving the warning effect.

[0111] The second pattern analysis module obtains the second crossing data in the second crossing pattern; constructs a second pattern recognition model to identify the second crossing data, and obtains the second abnormal area and the second abnormal coefficient; obtains the second distance between the second abnormal area and the crossing sliding door, and determines the second orientation data of the audio equipment according to the second distance and the second abnormal coefficient.

[0112] The second pattern recognition model includes a second data acquisition layer, a second crossing data recognition layer, and a second anomaly recognition layer;

[0113] The second data acquisition layer acquires second crossing data; the second crossing data is monitoring video data of the railway crossing during the second crossing mode;

[0114] The second crossing data recognition layer performs sound preprocessing on the second crossing data to obtain second preprocessed crossing data; identifies and obtains individuals with abnormal sound and individuals with abnormal behavior in the second preprocessed crossing data; obtains sound abnormality coefficients according to the sound type, sound decibel and sound frequency of the individuals with abnormal sound; obtains behavior abnormality coefficients according to the type, passing speed and behavior type of the individuals with abnormal behavior;

[0115] The sound types include vehicle horn sounds, crowd quarrel sounds, etc.; the abnormal behaviors include pedestrians approaching, climbing over the sliding gate at the intersection, vehicles overtaking, changing lanes, vehicles approaching the sliding gate at the intersection, etc.

[0116] The behavioral abnormality data during the second crossing mode are shown in Table 2.

[0117] Table 2 Data table of individuals with abnormal behavior

[0118] Abnormal Behavior Pass Individual Number type Passing speed Behavior Type T01 pedestrian 1.86m / s Sliding door near the crossing T02 car 2.51m / s Lane Change T03 car 2.64m / s Sliding door near the crossing

[0119] The second abnormality identification layer clusters the position data of individuals with abnormal voices, the abnormal voice coefficient, the position data and the abnormal behavior coefficient of individuals with abnormal behavior to obtain a second abnormal area; and obtains the second abnormality coefficient based on the abnormal voice coefficient of individuals with abnormal voices and the abnormal behavior coefficient of individuals with abnormal behavior in the second abnormal area.

[0120] The present invention constructs a second pattern recognition model to identify the second crossing data; based on the second crossing data, individuals with abnormal sound passing, abnormal sound coefficients, individuals with abnormal behavior passing and abnormal behavior coefficients are obtained; based on the position data of individuals with abnormal sound passing, the abnormal sound coefficients, the position data of individuals with abnormal behavior passing and the abnormal behavior coefficients, clustering is performed to obtain a second abnormal area; based on the abnormal sound coefficients of individuals with abnormal sound passing and the abnormal behavior coefficients of individuals with abnormal behavior passing in the second abnormal area, a second abnormal coefficient is obtained; and through the second abnormal area and the second abnormal coefficient, the abnormal area and the degree of abnormality during the second crossing mode are accurately measured.

[0121] In the second crossing data recognition layer, the process of sound preprocessing of the second crossing data is: according to the distance between the train and the railway crossing and the sound data of the train passing, the environmental noise caused by the passage of the train in the second crossing data is processed to obtain the second preprocessed crossing data.

[0122] The third mode analysis module obtains the third crossing data and the second movement data in the third crossing mode; constructs a third mode recognition model to recognize the third crossing data and the second movement data to obtain a third abnormal area and a third abnormal coefficient; obtains a third distance between the third abnormal area and the crossing sliding door, and determines the third orientation data of the audio equipment according to the third distance and the third abnormal coefficient.

[0123] The third pattern recognition model includes a third data acquisition layer, a third crossing data recognition layer and a third anomaly recognition layer;

[0124] The third data acquisition layer acquires third crossing data and second movement data; the third crossing data is monitoring video data of the railway crossing during the third crossing mode; the second movement data includes the opening speed of the crossing sliding door and the second passable width in the third crossing mode;

[0125] The third crossing data recognition layer recognizes the third crossing data and obtains third individual traffic data and third traffic environment data in the third crossing data; the third individual traffic data includes third individual traffic type, third traffic speed, third individual traffic distance and third traffic congestion data; the third traffic environment data includes environmental visibility, road surface slipperiness, road surface flatness and environmental noise data in the third crossing mode;

[0126] The third anomaly identification layer obtains the third anomaly coefficient of each passing individual based on the third passing individual data, the third passing environment data and the second movement data; obtains the third anomaly area based on the clustering of the passing individual positions and the third anomaly coefficients; and obtains the third anomaly coefficient based on the third anomaly coefficients of the passing individuals in the third anomaly area.

[0127] The present invention constructs a third mode recognition model to identify the third crossing data and the second movement data; firstly, the third passing individual data and the third passing environment data are obtained according to the third crossing data identification; the third passing abnormality coefficient of each passing individual is obtained by the third passing individual data, the third passing environment data and the second movement data identification; the third abnormal area is obtained by clustering according to the passing individual position and the third passing abnormality coefficient; the third abnormality coefficient is obtained according to the third passing abnormality coefficient of the passing individual in the third abnormal area; the abnormal area and the degree of abnormality during the third crossing mode are accurately measured through the third abnormal area and the third abnormality coefficient.

[0128] The audio intelligent control module controls the audio device according to the first orientation data, the second orientation data and the third orientation data.

[0129] In order to verify the early warning effect of the intelligent sound system described in the present invention, it is verified;

[0130] The verification process uses system 1 and system 2, wherein system 1 is an intelligent sound system according to the present invention; system 2 is a sound warning system with a fixed direction;

[0131] During the verification process, the location selection of the audio device includes location 1, location 2 and location 3;

[0132] The verification is divided into three stages, corresponding to the first crossing mode, the second crossing mode and the third crossing mode respectively; the average value of the first passing abnormality coefficient of the passing individuals in the first crossing mode is used as the first verification index; the average value of the second abnormality coefficient of the passing individuals in the second crossing mode is used as the second verification index; the average value of the third passing abnormality coefficient of the passing individuals in the third crossing mode is used as the third verification index; the data obtained after multiple verifications are shown in Table 3.

[0133] Table 3 Intelligent sound system warning effect data table

[0134]

[0135] According to the data in Table 3, under the same location, the warning effect of system 1 is better than that of system 2.

[0136] The present application obtains the first crossing mode, the second crossing mode and the third crossing mode according to the state of the crossing sliding door; identifies the first crossing data and the first movement data of the first crossing mode, obtains the first abnormal area and the first abnormal coefficient, and then determines the first orientation data in combination with the first distance between the first abnormal area and the crossing sliding door; identifies the second crossing data of the second crossing mode, obtains the second abnormal area and the second abnormal coefficient; determines the second orientation data according to the second distance between the second abnormal area and the crossing sliding door and the second abnormal coefficient; identifies the third crossing data and the second movement data of the third crossing mode, obtains the third abnormal area and the third abnormal coefficient, and then determines the third orientation data in combination with the third distance between the third abnormal area and the crossing sliding door. The present invention realizes intelligent control of railway crossing sound equipment through orientation data of different modes.

[0137] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent sound system, characterized in that: include: A crossing mode recognition module recognizes the crossing sliding door of the railway crossing, and divides the crossing sliding door into a first crossing mode, a second crossing mode and a third crossing mode according to the state of the crossing sliding door; A first pattern analysis module is configured to obtain first crossing data and first movement data in the first crossing pattern; construct a first pattern recognition model to recognize the first crossing data and the first movement data, and obtain a first abnormal area and a first abnormal coefficient; obtain a first distance between the first abnormal area and the crossing sliding door, and determine first orientation data of the audio device according to the first distance and the first abnormal coefficient; A second mode analysis module is used to obtain second crossing data in the second crossing mode; a second mode recognition model is constructed to recognize the second crossing data, and a second abnormal area and a second abnormal coefficient are obtained; a second distance between the second abnormal area and the crossing sliding door is obtained, and second orientation data of the audio device is determined according to the second distance and the second abnormal coefficient; A third mode analysis module is configured to obtain the third crossing data and the second movement data in the third crossing mode; construct a third mode recognition model to recognize the third crossing data and the second movement data, and obtain a third abnormal area and a third abnormal coefficient; obtain a third distance between the third abnormal area and the crossing sliding door, and determine the third orientation data of the audio device according to the third distance and the third abnormal coefficient; The audio intelligent control module controls the audio device according to the first orientation data, the second orientation data and the third orientation data.

2. The intelligent sound system according to claim 1, characterized in that: In the road crossing pattern recognition module, the process of dividing the first road crossing pattern, the second road crossing pattern and the third road crossing pattern is as follows: the first road crossing pattern is the process of the road crossing sliding door from opening to closing; The second crossing mode is when the crossing sliding door is in a closed state; the third crossing mode is when the crossing sliding door is in a process from closed to open.

3. The intelligent sound system according to claim 1, characterized in that: The first pattern recognition model includes a first data acquisition layer, a first crossing data recognition layer and a first anomaly recognition layer; The first data acquisition layer acquires first crossing data and first movement data; the first crossing data is surveillance video data of the railway crossing during the first crossing mode; the first movement data includes the closing speed of the crossing sliding door and the first passable width in the first crossing mode; The first crossing data recognition layer identifies and obtains first individual pass data and first traffic environment data in the first crossing data; the first individual pass data includes first individual pass type, first individual pass speed, first individual pass distance and first traffic congestion data; the first traffic environment data includes environmental visibility, road surface slipperiness, road surface flatness and environmental noise data in the first crossing mode; the first individual pass distance is the distance between the individual pass and the crossing sliding door; The first abnormality recognition layer obtains the first abnormality coefficient of each passing individual according to the first passing individual data, the first passing environment data and the first movement data; obtains the first abnormal area according to the clustering of the passing individual position and the first passing abnormality coefficient; and obtains the first abnormality coefficient according to the first passing abnormality coefficient of the passing individual in the first abnormal area.

4. The intelligent sound system according to claim 1, characterized in that: The process of determining the first orientation data according to the first distance and the first anomaly coefficient is: A three-dimensional coordinate system is constructed with the location of the audio equipment as the origin; wherein the z-axis of the three-dimensional coordinate system is a vertical line of the ground plane passing through the origin; According to the connection line between the first abnormal area and the audio device, first candidate orientation data of the audio device is obtained; the first candidate orientation data is vector data; According to the first distance between the first abnormal area and the crossing sliding door and the first abnormal coefficient of the first abnormal area, the first orientation weight of the first alternative orientation data is calculated; and the first orientation data is obtained according to all the first alternative orientation data and the corresponding first orientation weights.

5. The intelligent sound system according to claim 4, characterized in that: The second pattern recognition model includes a second data acquisition layer, a second crossing data recognition layer, and a second anomaly recognition layer; The second data acquisition layer acquires second crossing data; the second crossing data is monitoring video data of the railway crossing during the second crossing mode; The second crossing data recognition layer performs sound preprocessing on the second crossing data to obtain second preprocessed crossing data; Identify and obtain individuals passing through with abnormal sounds and individuals passing through with abnormal behaviors in the second pre-processed crossing data; According to the voice type, voice decibel and voice frequency of individuals with abnormal voice, the voice abnormality coefficient is identified; according to the type, passing speed and behavior type of individuals with abnormal behavior, the behavior abnormality coefficient is identified; The second abnormality identification layer clusters the position data of individuals with abnormal voices, the abnormal voice coefficient, the position data and the abnormal behavior coefficient of individuals with abnormal behavior to obtain a second abnormal area; and obtains the second abnormality coefficient based on the abnormal voice coefficient of individuals with abnormal voices and the abnormal behavior coefficient of individuals with abnormal behavior in the second abnormal area.

6. The intelligent sound system according to claim 5, characterized in that: In the second crossing data recognition layer, the process of sound preprocessing of the second crossing data is: according to the distance between the train and the railway crossing and the sound data of the train passing, the environmental noise caused by the passage of the train in the second crossing data is processed to obtain the second preprocessed crossing data.

7. The intelligent sound system according to claim 1, characterized in that: The third pattern recognition model includes a third data acquisition layer, a third crossing data recognition layer and a third anomaly recognition layer; The third data acquisition layer acquires third crossing data and second movement data; the third crossing data is monitoring video data of the railway crossing during the third crossing mode; the second movement data includes the opening speed of the crossing sliding door and the second passable width in the third crossing mode; The third crossing data recognition layer recognizes the third crossing data and obtains third individual traffic data and third traffic environment data in the third crossing data; the third individual traffic data includes third individual traffic type, third traffic speed, third individual traffic distance and third traffic congestion data; the third traffic environment data includes environmental visibility, road surface slipperiness, road surface flatness and environmental noise data in the third crossing mode; The third anomaly identification layer obtains the third anomaly coefficient of each passing individual based on the third passing individual data, the third passing environment data and the second movement data; obtains the third anomaly area based on the clustering of the passing individual positions and the third anomaly coefficients; and obtains the third anomaly coefficient based on the third anomaly coefficients of the passing individuals in the third anomaly area.

8. An intelligent audio method, characterized in that: include: Identify the sliding gates at railway crossings, and divide them into the first crossing mode, the second crossing mode, and the third crossing mode according to the states of the sliding gates; Acquire the first crossing data and the first movement data in the first crossing mode; construct a first pattern recognition model to recognize the first crossing data and the first movement data, and obtain a first abnormal area and a first abnormal coefficient; acquire a first distance between the first abnormal area and the crossing sliding door, and determine the first orientation data of the audio device according to the first distance and the first abnormal coefficient; Acquire the second crossing data in the second crossing mode; construct a second pattern recognition model to recognize the second crossing data, and obtain a second abnormal area and a second abnormal coefficient; obtain a second distance between the second abnormal area and the crossing sliding door, and determine the second orientation data of the audio device according to the second distance and the second abnormal coefficient; Obtain the third crossing data and the second movement data in the third crossing mode; construct a third pattern recognition model to identify the third crossing data and the second movement data, and obtain a third abnormal area and a third abnormal coefficient; obtain a third distance between the third abnormal area and the crossing sliding door, and determine the third orientation data of the audio device according to the third distance and the third abnormal coefficient; The audio device is controlled according to the first orientation data, the second orientation data and the third orientation data.

9. The intelligent audio method according to claim 8, characterized in that: The division process of the first crossing mode, the second crossing mode and the third crossing mode is as follows: the first crossing mode is the process from opening to closing of the crossing sliding door; The second crossing mode is when the crossing sliding door is in a closed state; the third crossing mode is when the crossing sliding door is in a process from closed to open.

Images