Intelligent driving scene recognition model creation method, device, equipment and storage medium

Abstract

The present application relates to a method, device, equipment and storage medium for creating an intelligent driving scene recognition model, and relates to the field of intelligent driving technology, including dividing intelligent driving scenes based on GPS signal strength and lane line clarity to obtain a state scene set; creating a scene state transition probability matrix based on the mutual transition probability between different state scenes in the state scene set; creating an observation state transition probability matrix based on the state scene set and observation objects, and the observation objects include GPS reliability observations, light intensity observations based on camera collection, road sign credibility observations and lane line credibility observations; creating an intelligent driving scene recognition model based on hidden Markov based on the scene state transition probability matrix and the observation state transition probability matrix. The scene recognition model based on hidden Markov created in the present application can realize intelligent recognition of driving scenes, and effectively improve the efficiency and accuracy of scene recognition.

Description

Technical Field

[0001] The present application relates to the field of intelligent driving technology, and in particular to an intelligent driving scene recognition model creation method, device, equipment and storage medium. Background Art

[0002] At present, with the rapid development of new generation information technology, artificial intelligence and other technologies, the global automobile industry is also in a period of profound transformation, which will completely change the way people travel, making cars no longer simple means of transportation, but will become smart terminals like mobile phones, with more powerful entertainment and service functions. Among them, intelligent driving is the product of the "intelligent manufacturing" and "Internet +" era, which refers to the use of computer systems to achieve automatic driving with almost no human intervention.

[0003] The technical environment requirements for intelligent driving mainly include high-precision maps, vehicle networking and 5G. Among them, due to various positioning errors, the moving vehicles on the electronic map coordinates cannot maintain the correct positional relationship with the surrounding objects. Therefore, the use of high-precision map matching can accurately locate the position of the intelligent driving vehicle on the lane, thereby improving the accuracy of vehicle positioning; and the high-precision map, as the memory system of intelligent driving, will also supplement the parts that the sensor cannot detect, monitor the real-time situation and feedback external information. For example, the sensor, as the eyes of unmanned driving, has its limitations, such as being easily affected by bad weather. At this time, high-precision maps can be used to obtain accurate traffic conditions at the current location.

[0004] It can be seen that when smart driving vehicles are driving on actual roads, they need lane line information identified by cameras and GPS (Global Positioning System) to provide accurate positioning information. If the lane line information and accurate positioning information can be clearly marked on the high-precision map, many dangerous scenes in the smart driving process will be identified in advance, which can effectively reduce the possibility of safety accidents.

[0005] In the related technologies, the scenes in intelligent driving are mainly identified manually, and the scenes are manually marked in the national map, which is time-consuming, labor-intensive and inefficient. In addition, the current scene recognition only involves indoors and outdoors, which cannot meet the needs of intelligent driving. When scene recognition is performed based only on GPS as the model basis, it is easily blocked by objects such as outdoor leaves, which leads to misjudgment. Summary of the invention

[0006] The present application provides a method, device, equipment and storage medium for creating an intelligent driving scene recognition model to solve the problems of low efficiency and poor accuracy of intelligent driving scene recognition methods in related technologies.

[0007] In a first aspect, a method for creating an intelligent driving scene recognition model is provided, comprising the following steps:

[0008] Intelligent driving scenarios are divided based on GPS signal strength and lane line clarity to obtain a set of state scenarios;

[0009] Create a scenario state transition probability matrix according to the mutual transition probabilities between different state scenarios in the state scenario set;

[0010] An observation state transition probability matrix is ​​created according to the state scene set and the observation objects, wherein the observation objects include GPS reliability observation O1, light intensity observation O2 based on camera acquisition, road sign credibility observation O3 based on camera acquisition, and lane line credibility observation O4 based on camera acquisition;

[0011] A hidden Markov-based intelligent driving scene recognition model is created according to the scene state transition probability matrix and the observation state transition probability matrix.

[0012] In some embodiments, the state scene set includes a first state scene S1, a second state scene S2, a third state scene S3 and a fourth state scene S4. The first state scene S1 is that the GPS signal strength is greater than or equal to a strength threshold and the clarity of the lane line recognized by the camera is greater than or equal to a clarity threshold. The second state scene S2 is that the GPS signal strength is greater than or equal to the strength threshold and the clarity of the lane line recognized by the camera is less than the clarity threshold. The third state scene S3 is that the GPS signal strength is less than the strength threshold and the clarity of the lane line recognized by the camera is greater than or equal to the clarity threshold. The fourth state scene S4 is that the GPS signal strength is less than the strength threshold and the clarity of the lane line recognized by the camera is less than the clarity threshold.

[0013] In some embodiments, the scene state transition probability matrix X(t) is:

[0014] X(t)=P(S1(0))×……×P(S i (t)|S j (t-1)) i∈{1,2,3,4},j∈{1,2,3,4}

[0015] Among them, S represents the state scene, P(S1(0)) represents the probability that the initial moment is the first state scene S1, and P(S i (t)|S j (t-1)) represents the probability of the i-th state scenario occurring at time t under the condition that the j-th state scenario occurs at time t-1.

[0016] In some embodiments, the observed state transition probability matrix Y(t) is:

[0017] Y(t)=P(O k (t)|S j (t-1)) k∈{1,2,3,4},j∈{1,2,3,4}

[0018] Among them, P(O k (t)|S j (t-1)) represents the probability of the state scenario observed based on the kth observation quantity occurring at the tth time under the condition that the jth state scenario occurs at the t-1th time.

[0019] In some embodiments, the intelligent driving scene recognition model is:

[0020] S max =max(X(t)×Y(t))=max(P(S i (t)|S j (t-1),O k (t)))

[0021] Among them, S max Indicates the state scenario with the highest probability of occurring at the current moment.

[0022] In some embodiments, after the step of creating a hidden Markov-based intelligent driving scene recognition model according to the scene state transition probability matrix and the observation state transition probability matrix, the method further includes:

[0023] The actual value of the acquired observation object is input into the intelligent driving scenario recognition model, so that the intelligent driving scenario recognition model solves the occurrence probability of each state scenario based on the actual value of the observation object, and takes the state scenario with the highest occurrence probability as the actual state scenario.

[0024] In some embodiments, after the step of taking the state scenario with the highest probability of occurrence as the actual state scenario, the following step is further included:

[0025] The actual status scene is marked on the high-precision map.

[0026] In a second aspect, a device for creating an intelligent driving scene recognition model is provided, comprising:

[0027] A scene division unit, which is used to divide the intelligent driving scene based on GPS signal strength and lane line clarity to obtain a state scene set;

[0028] A first creation unit, which is used to create a scene state transition probability matrix according to the mutual transition probabilities between different state scenes in the state scene set;

[0029] A second creation unit, which is used to create an observation state transition probability matrix according to the state scene set and the observation object, wherein the observation object includes a GPS reliability observation O1, a light intensity observation O2 based on camera acquisition, a road sign credibility observation O3 based on camera acquisition, and a lane line credibility observation O4 based on camera acquisition;

[0030] The third creation unit is used to create an intelligent driving scene recognition model based on hidden Markov according to the scene state transition probability matrix and the observation state transition probability matrix.

[0031] In a third aspect, a smart driving scene recognition model creation device is provided, comprising: a memory and a processor, wherein at least one instruction is stored in the memory, and the at least one instruction is loaded and executed by the processor to implement the aforementioned smart driving scene recognition model creation method.

[0032] In a fourth aspect, a computer-readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the aforementioned intelligent driving scene recognition model creation method is implemented.

[0033] The beneficial effects brought about by the technical solution provided in this application include: realizing intelligent recognition of driving scenes and effectively improving the efficiency and accuracy of scene recognition.

[0034] The present application provides a method, device, equipment and storage medium for creating an intelligent driving scene recognition model, including dividing the intelligent driving scene based on GPS signal strength and lane line clarity to obtain a state scene set; creating a scene state transition probability matrix according to the mutual transition probability between different state scenes in the state scene set; creating an observation state transition probability matrix according to the state scene set and the observation object, the observation object includes GPS reliability observation O1, light intensity observation O2 based on camera acquisition, road sign credibility observation O3 based on camera acquisition and lane line credibility observation O4 based on camera acquisition; creating an intelligent driving scene recognition model based on hidden Markov according to the scene state transition probability matrix and the observation state transition probability matrix. The present application creates a scene recognition model based on hidden Markov based on GPS data, camera data and the correlation between GPS data, camera data and scenes. Through this model, the occurrence probability of each state scene can be accurately identified, and the intelligent recognition of driving scenes is realized, which effectively improves the efficiency and accuracy of scene recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0036] Figure 1 A schematic diagram of a process for creating an intelligent driving scene recognition model provided in an embodiment of the present application;

[0037] Figure 2 A schematic diagram of the structure of an intelligent driving scene recognition model creation device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0038] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0039] The embodiments of the present application provide a method, device, equipment and storage medium for creating an intelligent driving scene recognition model, which can solve the problems of low efficiency and poor accuracy of intelligent driving scene recognition methods in related technologies.

[0040] Figure 1 A method for creating an intelligent driving scene recognition model provided in an embodiment of the present application includes the following steps:

[0041] Step S10: dividing the intelligent driving scene based on GPS signal strength and lane line clarity to obtain a state scene set;

[0042] Furthermore, the state scene set includes a first state scene S1, a second state scene S2, a third state scene S3 and a fourth state scene S4. The first state scene S1 is that the GPS signal strength is greater than or equal to the strength threshold and the clarity of the lane line recognized by the camera is greater than or equal to the clarity threshold, the second state scene S2 is that the GPS signal strength is greater than or equal to the strength threshold and the clarity of the lane line recognized by the camera is less than the clarity threshold, the third state scene S3 is that the GPS signal strength is less than the strength threshold and the clarity of the lane line recognized by the camera is greater than or equal to the clarity threshold, and the fourth state scene S4 is that the GPS signal strength is less than the strength threshold and the clarity of the lane line recognized by the camera is less than the clarity threshold.

[0043] For example, in the embodiment of the present application, the intelligent driving scene is first divided according to the GPS signal strength and the clarity of the lane line, that is, the scene is divided according to the relationship between the GPS signal strength and the lane line clarity and the corresponding thresholds. It should be noted that the specific setting of the strength threshold and the clarity threshold can be determined according to the actual situation and is not limited here.

[0044] Specifically, intelligent driving scenarios are divided into the following four types:

[0045] State scenario 1: GPS signal strength is greater than or equal to the strength threshold (i.e., GPS signal is strong), and the clarity of the lane line identified by the camera is greater than or equal to the clarity threshold (i.e., the lane line is clear), and the state is set to S1;

[0046] State scenario 2: GPS signal strength is greater than or equal to the strength threshold (i.e., GPS signal is strong), and the clarity of the lane line identified by the camera is less than the clarity threshold (i.e., the lane line is blurred), and the state is set to S2;

[0047] State scenario 3: GPS signal strength is less than the strength threshold (i.e., GPS signal is weak), and the clarity of the lane line identified by the camera is greater than or equal to the clarity threshold (i.e., the lane line is clear), and the state is set to S3;

[0048] State scenario 4: The GPS signal strength is less than the strength threshold (i.e., the GPS signal is weak), and the clarity of the lane line recognized by the camera is less than the clarity threshold (i.e., the lane line is blurred), and the state is set to S4.

[0049] In addition, the probability of transferring to other state scenarios under the condition of state scenario S1 satisfies formula (1):

[0050]

[0051] It can be seen from formula (1) that the transition probability of state scene S1 to state scene S2 is a1%, the transition probability of state scene S1 to state scene S2 is b1%, the transition probability of state scene S1 to state scene S3 is c1%, and the transition probability of state scene S1 to state scene S4 is d1%. According to the probability principle of the hidden Markov model, the sum of the above transition probability values ​​satisfies the following formula:

[0052] a1%+b1%+c1%+d1%=100% (2).

[0053] Similarly, the probability of transferring to other state scenarios under the state scenario S2 satisfies formula (3):

[0054]

[0055] It can be seen from formula (3) that the transition probability of state scene S2 to state scene S1 is a2%, the transition probability of state scene S2 to state scene S2 is b2%, the transition probability of state scene S2 to state scene S3 is c2%, and the transition probability of state scene S2 to state scene S4 is d2%; and according to the probability principle of the hidden Markov model, the sum of the above transition probability values ​​satisfies the following formula:

[0056] a2%+b2%+c2%+d2%=100% (4).

[0057] Similarly, the probability of transferring to other state scenarios under the condition of state scenario S3 satisfies formula (5):

[0058]

[0059] It can be seen from formula (5) that the transition probability of state scene S3 to state scene S1 is a3%, the transition probability of state scene S3 to state scene S2 is b3%, the transition probability of state scene S3 to state scene S3 is c3%, and the transition probability of state scene S3 to state scene S4 is d3%; and according to the probability principle of the hidden Markov model, the sum of the above transition probability values ​​satisfies the following formula:

[0060] a3%+b3%+c3%+d3%=100% (6).

[0061] Similarly, the probability of transferring to other state scenarios under the condition of state scenario S4 satisfies formula (7):

[0062]

[0063] From formula (7), we can know that the transition probability of state scene S4 to state scene S1 is a4%, the transition probability of state scene S4 to state scene S2 is b4%, the transition probability of state scene S4 to state scene S3 is c4%, and the transition probability of state scene S4 to state scene S4 is d4%. According to the probability principle of the hidden Markov model, the sum of the above transition probability values ​​satisfies the following formula:

[0064] a4%+b4%+c4%+d4%=100% (8).

[0065] Step S20: creating a scene state transition probability matrix according to the mutual transition probabilities between different state scenes in the state scene set;

[0066] Furthermore, the scene state transition probability matrix X(t) is:

[0067] X(t)=P(S1(0))×……×P(S i(t)|S j (t-1)) i∈{1,2,3,4},j∈{1,2,3,4}

[0068] Among them, S represents the state scene, P(S1(0)) represents the probability that the initial moment is the first state scene S1, and P(S i (t)|S j (t-1)) represents the probability of the i-th state scenario occurring at time t under the condition that the j-th state scenario occurs at time t-1.

[0069] For example, in the embodiment of the present application, each state scene is initialized. Specifically, referring to formula (9), the state value of each state scene at the initial moment is set:

[0070]

[0071] Where P(S1(0)), P(S2(0)), P(S3(0)) and P(S4(0)) represent the probability of the initial moment being the corresponding state scene S1, and p1, p2, p3 and p4 represent specific probability values.

[0072] Since the state transition probability matrix X of the hidden Markov model can be expressed as:

[0073] X=P(S(t)|S(t-1)) (10)

[0074] Therefore, by substituting formulas (1), (3), (5), (7) and (9) into formula (10), we can obtain the scene state transition probability matrix X(t):

[0075] X(t)=P(S1(0))×……×P(S i (t)|S j (t-1)) i∈{1,2,3,4},j∈{1,2,3,4} (11)

[0076] Among them, P(S i (t)|S j (t-1)) represents the probability of the i-th state scenario occurring at time t under the condition that the j-th state scenario occurs at time t-1.

[0077] Step S30: creating an observation state transition probability matrix according to the state scene set and the observation objects, wherein the observation objects include GPS reliability observation O1, light intensity observation O2 based on camera acquisition, road sign credibility observation O3 based on camera acquisition, and lane line credibility observation O4 based on camera acquisition;

[0078] Furthermore, the observed state transition probability matrix Y(t) is:

[0079] Y(t)=P(O k (t)|S j (t-1)) k∈{1,2,3,4},j∈{1,2,3,4}

[0080] Among them, P(O k (t)|S j (t-1)) represents the probability of the state scenario observed based on the kth observation quantity occurring at the tth time under the condition that the jth state scenario occurs at the t-1th time.

[0081] Exemplarily, in the embodiment of the present application, the GPS reliability observation O1, the light intensity observation O2 based on camera acquisition, the road sign credibility observation O3 based on camera acquisition, and the lane line credibility observation O4 based on camera acquisition are designed respectively.

[0082] Specifically, set GPS reliability as the observation value, represented by O1, then:

[0083]

[0084] In the formula, n represents the number of satellites searched, N th Indicates the satellite number threshold, represents the average signal-to-noise ratio, S th Represents the signal-to-noise ratio threshold, 1 indicates that GPS reliability is available, and 0 indicates that GPS reliability is unavailable; the solution for the average signal-to-noise ratio is as follows:

[0085]

[0086] Wherein, sn represents the signal-to-noise ratio fed back to the receiver by the nth satellite.

[0087] Therefore, in the state scenario S j (t-1) condition, if the current frame GPS information is reliable, then the corresponding observation state transition probability is As shown in formula (14):

[0088]

[0089] If the current frame GPS information is unreliable, the corresponding observation state transition probability is As shown in formula (15):

[0090]

[0091] The light intensity collected by the camera is set as the observed value, represented by O2. In order to avoid objects such as leaves blocking the GPS, which may cause the model to misjudge indoors and outdoors, this embodiment adds the light intensity collected by the camera under daytime conditions to satisfy the following formula:

[0092]

[0093] In the formula, L represents the light intensity information detected by the camera, L th Indicates the light intensity threshold, 1 means indoors and 0 means outdoors.

[0094] Therefore, in the state scenario S j (t-1) condition, if the current state is indoors, then the corresponding observed state transition probability is As shown in formula (17):

[0095]

[0096] If it is currently outdoors, then the corresponding observation state transition probability is As shown in formula (18):

[0097]

[0098] The credibility of the road sign collected by the camera is set as the observation value, represented by O3, then:

[0099]

[0100] In the formula, M represents the credibility of the road sign detected by the camera, which can be directly collected and output by the camera, M th Represents the road sign credibility threshold, 1 means the real road sign is detected, 0 means the real road sign is not detected.

[0101] Therefore, in the state scenario S j (t-1) condition, if the real road sign is detected, then the corresponding observation state transition probability is As shown in formula (20):

[0102]

[0103] If the real road sign is not detected at present, the corresponding observation state transition probability is As shown in formula (21):

[0104]

[0105] The lane line credibility based on camera acquisition is set as the observation value, represented by O4, then:

[0106]

[0107] In the formula, LL represents the credibility of the lane line detected by the camera, which can be directly collected and output by the camera. th Indicates the lane line credibility threshold, 1 means the real lane line is detected, and 0 means the real lane line is not detected.

[0108] Therefore, in the state scenario S j (t-1) condition, if the real lane line is currently detected, then the corresponding observation state transition probability As shown in formula (23):

[0109]

[0110] If the lane line is not detected at present, the corresponding observation state transition probability is As shown in formula (24):

[0111]

[0112] By abstracting formulas (14), (15), (17), (18), (20), (21), (23) and (24), we can obtain the observed state transition probability matrix Y(t) of the hidden Markov model:

[0113] Y(t)=P(O k (t)|S j (t-1)) k∈{1,2,3,4},j∈{1,2,3,4} (25)

[0114] In the formula, P(O k (t)|S j (t-1)) represents the probability of the state scenario observed based on the kth observation quantity occurring at the tth time under the condition that the jth state scenario occurs at the t-1th time.

[0115] Step S40: Creating a hidden Markov-based intelligent driving scene recognition model according to the scene state transition probability matrix and the observation state transition probability matrix.

[0116] Furthermore, the intelligent driving scene recognition model is:

[0117] S max =max(X(t)×Y(t))=max(P(S i (t)|S j (t-1),O k (t)))

[0118] Among them, S max Indicates the state scenario with the highest probability of occurring at the current moment.

[0119] Furthermore, after the step of creating a hidden Markov-based intelligent driving scene recognition model according to the scene state transition probability matrix and the observation state transition probability matrix, the method further includes:

[0120] The actual value of the acquired observation object is input into the intelligent driving scenario recognition model, so that the intelligent driving scenario recognition model solves the occurrence probability of each state scenario based on the actual value of the observation object, and takes the state scenario with the highest occurrence probability as the actual state scenario.

[0121] Furthermore, after the step of taking the state scenario with the highest probability of occurrence as the actual state scenario, the method further includes:

[0122] The actual status scene is marked on the high-precision map.

[0123] Exemplarily, in an embodiment of the present application, the scene state transition probability matrix and the observation state transition probability matrix are multiplied to obtain an intelligent driving scene recognition model based on hidden Markov; then the actual values ​​of each observation object received are input into the intelligent driving scene recognition model, and the intelligent driving scene recognition model can calculate the probability of occurrence of each state scene corresponding to the current moment through the state scene probability of the previous moment and the probability of a certain state scene at the current moment observed under the state scene at the previous moment, and then select the state scene with the highest probability of occurrence from the multiple state scenes corresponding to the current moment as the actual state scene at the current moment; finally, the actual state scene is marked in the high-precision map, so that when the intelligent driving vehicle uses the high-precision map, it can obtain the scene that the vehicle is about to arrive at through the position of the vehicle in the map, thereby realizing the positioning of the intelligent driving scene.

[0124] It can be seen that the present application creates a scene recognition model based on hidden Markov based on GPS data, camera data, and the correlation between GPS data, camera data and scenes. Through this model, the occurrence probability of each state scene can be accurately identified, and the state scene with the highest probability of occurrence can be used as the actual state scene, thereby realizing intelligent recognition of driving scenes and effectively improving the efficiency and accuracy of scene recognition; and the identified scenes are automatically annotated directly to the high-precision map without manual annotation, which effectively improves the scene annotation efficiency.

[0125] The embodiment of the present application also provides a device for creating an intelligent driving scene recognition model, including:

[0126] A scene division unit, which is used to divide the intelligent driving scene based on GPS signal strength and lane line clarity to obtain a state scene set;

[0127] A first creation unit, which is used to create a scene state transition probability matrix according to the mutual transition probabilities between different state scenes in the state scene set;

[0128] A second creation unit, which is used to create an observation state transition probability matrix according to the state scene set and the observation object, wherein the observation object includes a GPS reliability observation O1, a light intensity observation O2 based on camera acquisition, a road sign credibility observation O3 based on camera acquisition, and a lane line credibility observation O4 based on camera acquisition;

[0129] The third creation unit is used to create an intelligent driving scene recognition model based on hidden Markov according to the scene state transition probability matrix and the observation state transition probability matrix.

[0130] Furthermore, the state scene set includes a first state scene S1, a second state scene S2, a third state scene S3 and a fourth state scene S4. The first state scene S1 is that the GPS signal strength is greater than or equal to the strength threshold and the clarity of the lane line recognized by the camera is greater than or equal to the clarity threshold, the second state scene S2 is that the GPS signal strength is greater than or equal to the strength threshold and the clarity of the lane line recognized by the camera is less than the clarity threshold, the third state scene S3 is that the GPS signal strength is less than the strength threshold and the clarity of the lane line recognized by the camera is greater than or equal to the clarity threshold, and the fourth state scene S4 is that the GPS signal strength is less than the strength threshold and the clarity of the lane line recognized by the camera is less than the clarity threshold.

[0131] Furthermore, the scene state transition probability matrix X(t) is:

[0132] X(t)=P(S1(0))×……×P(S i (t)|S j (t-1)) i∈{1,2,3,4},j∈{1,2,3,4}

[0133] Among them, S represents the state scene, P(S1(0)) represents the probability that the initial moment is the first state scene S1, and P(S i (t)|S j (t-1)) represents the probability of the i-th state scenario occurring at time t under the condition that the j-th state scenario occurs at time t-1.

[0134] Furthermore, the observed state transition probability matrix Y(t) is:

[0135] Y(t)=P(O k (t)|S j (t-1)) k∈{1,2,3,4},j∈{1,2,3,4}

[0136] Among them, P(Ok (t)|S j (t-1)) represents the probability of the state scenario observed based on the kth observation quantity occurring at the tth time under the condition that the jth state scenario occurs at the t-1th time.

[0137] Furthermore, the intelligent driving scene recognition model is:

[0138] S max =max(X(t)×Y(t))=max(P(S i (t)|S j (t-1),O k (t)))

[0139] Among them, S max Indicates the state scenario with the highest probability of occurring at the current moment.

[0140] Furthermore, the intelligent driving scenario recognition model is used to: input the actual value of the acquired observation object into the intelligent driving scenario recognition model, so that the intelligent driving scenario recognition model can solve the occurrence probability of each state scenario based on the actual value of the observation object, and take the state scenario with the highest probability of occurrence as the actual state scenario.

[0141] Furthermore, the intelligent driving scene recognition model is also used to: mark the actual state scene into a high-precision map.

[0142] This application creates a scene recognition model based on hidden Markov based on GPS data, camera data, and the correlation between GPS data, camera data and scenes. Through this model, the probability of occurrence of each state scene can be accurately identified, and the state scene with the highest probability of occurrence can be used as the actual state scene, thereby realizing intelligent recognition of driving scenes and effectively improving the efficiency and accuracy of scene recognition; and the identified scenes are automatically annotated directly to the high-precision map without manual annotation, which effectively improves the scene annotation efficiency.

[0143] It should be noted that technicians in the relevant field can clearly understand that for the convenience and conciseness of description, the specific working process of the above-described device and each unit can refer to the corresponding process in the aforementioned intelligent driving scene recognition model creation method embodiment, and will not be repeated here.

[0144] The intelligent driving scene recognition model creation device provided in the above embodiment can be implemented in the form of a computer program. The computer program can be used in Figure 2 The intelligent driving scene recognition model shown is created to run on the device.

[0145] An embodiment of the present application also provides an intelligent driving scene recognition model creation device, including: a memory, a processor and a network interface connected via a system bus, the memory storing at least one instruction, and the at least one instruction being loaded and executed by the processor to implement all or part of the steps of the aforementioned intelligent driving scene recognition model creation method.

[0146] The network interface is used for network communication, such as sending assigned tasks, etc. Those skilled in the art will understand that Figure 2 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0147] The processor may be a CPU, or other general-purpose processors, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor, or the processor may be any conventional processor, etc. The processor is the control center of the computer device, and uses various interfaces and lines to connect various parts of the entire computer device.

[0148] The memory can be used to store computer programs and / or modules. The processor realizes various functions of the computer device by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a video playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as video data, image data, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (SmartMediacard, SMC), a secure digital (Secure digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device or other volatile solid-state storage device.

[0149] An embodiment of the present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, all or part of the steps of the aforementioned intelligent driving scene recognition model creation method are implemented.

[0150] The embodiment of the present application implements all or part of the aforementioned process, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of each of the above methods when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electric carrier signals and telecommunication signals.

[0151] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, servers or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program codes.

[0152] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or system including the element.

[0153] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the process in the flowchart. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0154] The above description is only a specific implementation of the present application, so that those skilled in the art can understand or implement the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest range consistent with the principles and novel features applied for herein.

Claims

1. A method for creating an intelligent driving scene recognition model, characterized in that: The following steps are involved: Intelligent driving scenarios are divided based on GPS signal strength and lane line clarity to obtain a set of state scenarios; Create a scenario state transition probability matrix according to the mutual transition probabilities between different state scenarios in the state scenario set; An observation state transition probability matrix is ​​created according to the state scene set and the observation objects, wherein the observation objects include GPS reliability observation O1, light intensity observation O2 based on camera acquisition, road sign credibility observation O3 based on camera acquisition, and lane line credibility observation O4 based on camera acquisition; Creating a hidden Markov-based intelligent driving scene recognition model according to the scene state transition probability matrix and the observation state transition probability matrix; Among them, the state scene set includes a first state scene S1, a second state scene S2, a third state scene S3 and a fourth state scene S4. The first state scene S1 is that the GPS signal strength is greater than or equal to the strength threshold and the clarity of the lane line recognized by the camera is greater than or equal to the clarity threshold, the second state scene S2 is that the GPS signal strength is greater than or equal to the strength threshold and the clarity of the lane line recognized by the camera is less than the clarity threshold, the third state scene S3 is that the GPS signal strength is less than the strength threshold and the clarity of the lane line recognized by the camera is greater than or equal to the clarity threshold, and the fourth state scene S4 is that the GPS signal strength is less than the strength threshold and the clarity of the lane line recognized by the camera is less than the clarity threshold.

2. The method for creating an intelligent driving scene recognition model according to claim 1, characterized in that: The scene state transition probability matrix X(t) is: X(t)=P(S1(0))×……×P(S i (t)|S j (t-1))i∈{1,2,3,4},j∈{1,2,3,4} Among them, S represents the state scene, P(S1(0)) represents the probability that the initial moment is the first state scene S1, and P(S i (t)|S j (t-1)) represents the probability of the i-th state scenario occurring at time t under the condition that the j-th state scenario occurs at time t-1.

3. The method for creating an intelligent driving scene recognition model according to claim 2, characterized in that: The observed state transition probability matrix Y(t) is: Y(t)=P(O k (t)|S j (t-1))k∈{1,2,3,4},j∈{1,2,3,4} Among them, P(O k (t)|S j (t-1)) represents the probability of the state scenario observed based on the kth observation quantity occurring at the tth time under the condition that the jth state scenario occurs at the t-1th time.

4. The method for creating an intelligent driving scene recognition model according to claim 3, characterized in that: The intelligent driving scene recognition model is: S max =max(X(t)×Y(t))=max(P(S i (t)|S j (t-1),O k (t))) Among them, S max Indicates the state scenario with the highest probability of occurring at the current moment.

5. The method for creating an intelligent driving scene recognition model according to claim 1, characterized in that: After the step of creating a hidden Markov-based intelligent driving scene recognition model according to the scene state transition probability matrix and the observation state transition probability matrix, the method further includes: The actual value of the acquired observation object is input into the intelligent driving scenario recognition model, so that the intelligent driving scenario recognition model solves the occurrence probability of each state scenario based on the actual value of the observation object, and takes the state scenario with the highest occurrence probability as the actual state scenario.

6. The method for creating an intelligent driving scene recognition model according to claim 5, characterized in that: After the step of taking the state scenario with the highest probability of occurrence as the actual state scenario, the method further includes: The actual status scene is marked on the high-precision map.

7. An intelligent driving scene recognition model creation device, characterized in that: include: A scene division unit, which is used to divide the intelligent driving scene based on GPS signal strength and lane line clarity to obtain a state scene set; A first creation unit, which is used to create a scene state transition probability matrix according to the mutual transition probabilities between different state scenes in the state scene set; A second creation unit, which is used to create an observation state transition probability matrix according to the state scene set and the observation object, wherein the observation object includes a GPS reliability observation O1, a light intensity observation O2 based on camera acquisition, a road sign credibility observation O3 based on camera acquisition, and a lane line credibility observation O4 based on camera acquisition; A third creation unit, which is used to create an intelligent driving scene recognition model based on hidden Markov according to the scene state transition probability matrix and the observation state transition probability matrix; Among them, the state scene set includes a first state scene S1, a second state scene S2, a third state scene S3 and a fourth state scene S4. The first state scene S1 is that the GPS signal strength is greater than or equal to the strength threshold and the clarity of the lane line recognized by the camera is greater than or equal to the clarity threshold, the second state scene S2 is that the GPS signal strength is greater than or equal to the strength threshold and the clarity of the lane line recognized by the camera is less than the clarity threshold, the third state scene S3 is that the GPS signal strength is less than the strength threshold and the clarity of the lane line recognized by the camera is greater than or equal to the clarity threshold, and the fourth state scene S4 is that the GPS signal strength is less than the strength threshold and the clarity of the lane line recognized by the camera is less than the clarity threshold.

8. An intelligent driving scene recognition model creation device, characterized in that: include: A memory and a processor, wherein the memory stores at least one instruction, and the at least one instruction is loaded and executed by the processor to implement the intelligent driving scene recognition model creation method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer storage medium stores a computer program, and when the computer program is executed by a processor, the method for creating an intelligent driving scene recognition model according to any one of claims 1 to 6 is implemented.

Images