Blind zone precision evaluation method and system based on AI camera shooting technology

Abstract

The invention discloses a blind area precision evaluation method and system based on an AI camera shooting technology, and belongs to the technical field of automobile blind area evaluation. Each influence factor is determined, and a monitoring main body corresponding to each influence factor is identified; setting a blind area comprehensive range of the vehicle; setting a comprehensive influence model corresponding to the blind area comprehensive range; monitoring the monitoring main body of each influence factor in real time to obtain main body monitoring data of each influence factor; determining a blind area, and generating a blind area initial image according to the blind area; analyzing the monitoring data of each main body through a comprehensive influence model to obtain a comprehensive influence value corresponding to each blind area position; dividing the blind area initial image into a plurality of unit areas according to each comprehensive influence value, and setting a unit precision value of each unit area; and estimating a reference risk value of each target object, identifying a unit precision value corresponding to the unit region where the target object is located, correcting the reference risk value according to the unit precision value, and performing blind area early warning through the corrected reference risk value.

Description

Technical Field

[0001] The present invention belongs to the technical field of automobile blind spot assessment, and specifically relates to a blind spot accuracy assessment method and system based on AI camera technology. Background Art

[0002] With the rapid development of the automobile industry and the continuous progress of intelligent transportation systems, the improvement of vehicle safety performance has become the focus of industry attention. However, during the driving process of the vehicle, due to the limited vision of the driver and the obstruction of the vehicle structure, the blind spot problem has always been an important factor affecting driving safety. The improvement of its evaluation accuracy is of great significance for preventing traffic accidents and ensuring driving safety. However, the blind spot evaluation system on the current market still faces many challenges in practical applications, especially in terms of the real-time and accuracy of the evaluation accuracy. As a result, many users currently lack trust in it; because most of the existing blind spot evaluation systems use fixed sensors and algorithms, lack of real-time consideration of dynamic factors such as equipment status, environmental impact, data delay and vehicle speed. This leads to the accuracy of blind spot evaluation is often restricted by multiple factors, so that the evaluation accuracy of different areas in the blind spot is different, but in practical applications, it is the same when presented to users; based on this, in order to realize the intelligent evaluation of blind spots, the present invention provides a blind spot accuracy evaluation method and system based on AI camera technology. Summary of the invention

[0003] In order to solve the problems existing in the above-mentioned solutions, the present invention provides a blind spot accuracy assessment method and system based on AI camera technology.

[0004] The purpose of the present invention can be achieved through the following technical solutions: A blind spot accuracy assessment system based on AI camera technology, including an information module, a subject monitoring module, a blind spot module and a blind spot analysis module; The information module is used to perform information management, determine various influencing factors, and identify the monitoring subject corresponding to each of the influencing factors; Set a comprehensive blind spot range for the vehicle; obtain the main material data of each of the monitoring subjects, perform a single simulation on the main material data of each of the monitoring subjects, and obtain the factor influence curves corresponding to each blind spot position within the comprehensive blind spot range of each influencing factor, the horizontal axis of the factor influence curve is the main monitoring data of the corresponding monitoring subject, and the vertical axis is the accuracy influence value corresponding to the corresponding subject monitoring data, and the accuracy influence value range is [0, 100]; perform a comprehensive analysis based on the factor influence curves corresponding to each blind spot position to form a comprehensive influence model corresponding to the comprehensive blind spot range.

[0005] The subject monitoring module performs real-time monitoring on the monitoring subject corresponding to each influencing factor to obtain subject monitoring data corresponding to each influencing factor.

[0006] The blind area module is used to generate an initial image of the blind area in real time.

[0007] Furthermore, the method for generating the initial image of the blind area includes: Determine the blind spot area, collect image data around the vehicle in real time through the AI ​​camera, perform feature recognition on the image data through a preset image recognition model, and obtain target information of each target object; Determine relative data between the target object and the vehicle according to each target information, and integrate the relative data with the target information to form comprehensive target information of each target object; A blind spot image is generated according to the blind spot area and the target comprehensive information of each target object, and is marked as a blind spot initial image.

[0008] Furthermore, the method for determining the blind area includes: Obtaining blind spot material data of the vehicle, the blind spot material data including rear mirror parameters and corresponding material areas; performing statistics on each blind spot material data to obtain each material area corresponding to each rear mirror parameter and a share ratio corresponding to each material area; Establishing a three-dimensional appearance model corresponding to the vehicle, marked as a vehicle appearance model; restricting and adjusting the vehicle appearance model according to each of the material areas to form a dynamic restriction model; Identify rear mirror parameters of the current vehicle, match corresponding benchmark material areas according to the rear mirror parameters, display the dynamic restriction model to the user according to the benchmark material area, the user determines the setting screen through the dynamic restriction model, adjusts the rear mirror parameters according to the setting screen, and determines the blind spot area.

[0009] Furthermore, the method of matching the corresponding reference material area according to the rear mirror parameters includes: Identify each material area corresponding to the rear mirror parameter, and obtain the share proportion corresponding to each material area; obtain the number of times the user uses each material area under the rear mirror parameter; Calculate the priority value of each material area in real time according to the formula PY=FE+YC; Where: PY is the priority value; FE is the share value; YC is the number of applications; The material area with the highest priority value is selected as the reference material area.

[0010] Furthermore, after the user applies the vehicle appearance model to determine the blind spot area, the parameters and setting screens of each rear mirror are recorded in real time, integrated into adjustment record data, and each adjustment record data is deduplicated; Classifying each of the adjustment record data to obtain each application classification; setting an application classification label corresponding to each application classification in the dynamic restriction model; When the user needs to determine the blind spot area again, the determination is performed through the corresponding application classification label in the dynamic restriction model.

[0011] The blind spot analysis module is used to evaluate and analyze the initial blind spot image, obtain monitoring data of each subject, analyze the monitoring data of each subject through the comprehensive impact model, and obtain the comprehensive impact value corresponding to each blind spot position in the initial blind spot image; divide the initial blind spot image into a plurality of unit areas according to the comprehensive impact value corresponding to each blind spot position, and set a unit precision value for each unit area, wherein the unit precision value is the average value of the comprehensive impact value of each blind spot position in the unit area; Mark the accuracy value of each unit in each unit area; estimate the baseline risk value of each target object, and issue a blind spot warning based on the baseline risk value of each target object.

[0012] Furthermore, the method of dividing the initial blind spot map into a plurality of unit areas according to the comprehensive influence value of each blind spot position includes: Step SA1: identifying the comprehensive impact value of each blind spot position; determining each initial position; marking each blind spot position adjacent to the initial position as a selected position; The comprehensive impact value of the initial position is marked as QY1 (x，y，z) , (x, y, z) represents the corresponding position coordinates; the comprehensive influence value of the selected position is marked as QY2 (x，y，z) ; According to the formula YM=|QY1 (x，y，z) -QY2 (x，y，z) | Calculate the first impact difference between the initial position and the selected position; Where: YM is the first impact difference; Merge the initial position where the first impact difference is less than the threshold X1 with the selected position to obtain an initial area; Step SA2: Mark each blind spot position adjacent to the initial area as a selected position; According to the formula calculating a second influence difference between the initial region and the selected location; Where: YM´ is the second impact difference; i represents the corresponding blind spot position in the initial area, i=1, 2, ..., n, n is a positive integer; QY (x，y，z)i Indicates the comprehensive impact value of the corresponding blind spot position in the initial area; QY´ (x，y，z) Indicates the comprehensive impact value corresponding to the selected position; Merge the initial region whose second impact difference is less than the threshold value X1 with the selected position to obtain a new initial region; Step SA3: loop step SA2 until the initial area cannot be merged, and mark the initial area as a unit area; When there is no unmerged blind spot position in the blind spot area, the analysis ends; When there are unmerged blind spot positions in the blind spot area, return to step SA1.

[0013] Furthermore, before performing a blind spot warning based on a baseline risk value, a target object corresponding to the baseline risk value is identified, a unit accuracy value corresponding to the target object is identified, the baseline risk value is corrected based on the unit accuracy value, and a blind spot warning is performed using the corrected baseline risk value.

[0014] Furthermore, the revised benchmark risk value is recorded in real time, the revised benchmark risk value is marked as the assessed risk value, and the risk difference between the assessed risk value and the corresponding benchmark risk value is calculated; the abnormal optimization data is determined based on the risk difference and the benchmark risk value, and each abnormal optimization data is sent to the platform.

[0015] A blind spot accuracy assessment method based on AI camera technology, the method comprising: Determine each influencing factor and identify the monitoring subject corresponding to each influencing factor; Set the comprehensive range of the blind spot of the vehicle; obtain the main material data of each monitoring subject, perform a single simulation on the main material data of each monitoring subject, and obtain the factor influence curve corresponding to each blind spot position within the comprehensive range of the blind spot for each influencing factor; perform a comprehensive analysis based on the factor influence curves of each blind spot position to form a comprehensive influence model corresponding to the comprehensive range of the blind spot; Conduct real-time monitoring of the monitoring subjects of each influencing factor to obtain the main monitoring data of each influencing factor; Determine the blind spot area, collect image data around the vehicle in real time through the AI ​​camera, perform feature recognition on the image data through the preset image recognition model, and obtain the target information of each target object; determine the relative data between the target object and the vehicle based on each target information, integrate the relative data with the target information to form the target comprehensive information of each target object; generate a blind spot image based on the blind spot area and the target comprehensive information of each target object, and mark it as the blind spot initial image; The monitoring data of each subject is analyzed through the comprehensive impact model to obtain the comprehensive impact value corresponding to each blind spot position in the initial blind spot image; the initial blind spot image is divided into several unit areas according to the comprehensive impact value corresponding to each blind spot position, and the unit accuracy value of each unit area is set; Mark the accuracy value of each unit in each unit area; estimate the baseline risk value of each target object, identify the unit accuracy value corresponding to the unit area where the target object is located, correct the baseline risk value according to the unit accuracy value, and issue a blind spot warning through the corrected baseline risk value.

[0016] Compared with the prior art, the present invention has the following beneficial effects: Through the cooperation between the information module, blind spot module and blind spot analysis module, the blind spot can be accurately evaluated and the blind spot risk can be reduced. The existing problems of the current blind spot evaluation system can be solved, providing strong technical support for the development of intelligent transportation and autonomous driving technology, and continuously improving the accuracy and real-time performance of blind spot evaluation. The driver can more intuitively understand the driving environment around the vehicle and reduce driving troubles caused by blind spots. At the same time, the system can also continuously optimize the blind spot monitoring and analysis algorithm based on historical accuracy evaluation records to further improve driving comfort and convenience. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 This is a principle block diagram of the present invention. DETAILED DESCRIPTION

[0019] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. 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.

[0020] like Figure 1 As shown, a blind spot accuracy assessment system based on AI camera technology includes an information module, a blind spot module and a blind spot analysis module; The information module is used for information management. It determines various influencing factors that are applicable to blind spot assessment based on historical blind spot assessment data, such as environmental factors, equipment factors, transmission factors, vehicle speed factors and other influencing factors; it matches the existing application blind spot assessment system based on various influencing factors, determines the subject corresponding to each influencing factor, and marks it as a monitoring subject. For example, the monitoring subject of the equipment factor is each device related to the equipment factor, and the impact analysis needs to be performed based on the equipment performance and equipment monitoring status of the monitoring subject later; Analyze the possible blind spot positions of the vehicle based on historical blind spot assessment data to form a comprehensive blind spot range, which refers to the range relative to the vehicle; The possible data of each monitoring subject is obtained and integrated into the main material data. A single simulation is performed on the main material data of each monitoring subject to understand the influence of the single main material data on the blind spot image under different circumstances, and form the factor influence curve corresponding to each blind spot position within the comprehensive range of the blind spot for each influencing factor; the horizontal axis of the factor influence curve is the main monitoring data of the corresponding monitoring subject, and the vertical axis is the accuracy influence value corresponding to the corresponding main monitoring data. The accuracy influence value is set according to the degree of influence on the real accuracy, and the value range is [0, 100]. If there is no influence, the accuracy influence value is 0. In the case of the same influence as complete distortion, the accuracy influence value is 0, such as too large delay, resulting in loss of effect, abnormal equipment, resulting in inability to identify the target object, etc.; the influence ratio on the accuracy can be multiplied by 100; it can also be directly set manually during the simulation process; A comprehensive analysis is performed on the factor influence curves of each influencing factor corresponding to each blind spot position to form a comprehensive influence model corresponding to the comprehensive range of the blind spot; that is, a correlation analysis is performed on the factor influence curves of the same blind spot position to understand the comprehensive influence of different subject material data corresponding to each influencing factor on the accuracy, and a comprehensive influence value is formed. The value range is also [0, 100], which is determined in combination with the corresponding historical data and simulation method; because this module step is the preliminary preparation of the platform, the platform's resources can be used for analysis to form a sufficient training set, and then a comprehensive influence model can be developed based on the training set. Through the comprehensive influence model, each blind spot position can be analyzed according to the monitoring data of each subject to obtain the corresponding comprehensive influence value.

[0021] The subject monitoring module performs real-time monitoring on the monitoring subject corresponding to each influencing factor to obtain subject monitoring data corresponding to each influencing factor.

[0022] The blind spot module is used to generate an initial image of the blind spot in real time; determine the blind spot area, collect image data around the vehicle in real time through the AI ​​camera, perform feature recognition on the image data through a preset image recognition model, and obtain target information of each target object, such as position, size, shape, etc.; the image recognition model is established based on existing image technology to identify image data, determine each target object included in the image data, and then determine target information, and the target object is a vehicle, pedestrian, obstacle, etc. that needs to be monitored; calculate relative data such as the distance and relative speed between the target object and the vehicle based on the target information of each target object, integrate the obtained relative data with the target information, and form comprehensive target information of each target object; A blind spot image is generated according to the target comprehensive information of each target object and the blind spot area, and is marked as a blind spot initial image.

[0023] In one embodiment, the blind spot area is determined according to an existing blind spot determination method; for example, it is determined based on factors such as a rear-view mirror, a driving seat, and a driver.

[0024] In one embodiment, the method for determining the blind area includes: Establishing a three-dimensional appearance model corresponding to the vehicle, marked as a vehicle appearance model; Obtain a large amount of historical blind spot area and rear mirror angle parameter relationship data of the vehicle, marked as blind spot material data, the blind spot material data is composed of rear mirror parameters and their corresponding material areas; perform statistics on each blind spot material data, obtain each material area corresponding to each rear mirror parameter and the corresponding share ratio of each material area; The vehicle appearance model is restricted and adjusted according to each material area to form a dynamic restriction model, that is, the dynamic restriction model can only see the vehicle appearance corresponding to each material area through manual rotation and other methods, which is used for users to adjust the dynamic restriction model according to the vehicle rotation they can actually see, and then determine their blind spot conditions; Identify the rear mirror parameters of the current vehicle, match the corresponding reference material area according to the rear mirror parameters, display the dynamic restriction model to the user according to the reference material area, and the user adjusts the dynamic restriction model according to the vehicle data actually observed by the user. The screen corresponding to the dynamic restriction model after successful adjustment is marked as the setting screen, and the rear mirror parameters are adjusted according to the setting screen to determine the blind area. That is, when the user determines the setting screen, the user's actual viewing angle can be understood according to the setting screen, and the rear mirror simulation adjustment can be performed to determine whether the setting screen is accurate according to the changes in the screen seen by the user; when the user's actual viewing angle is clear, it can be determined how to adjust the rear mirror parameters under this condition to achieve the best driving viewing angle, and then determine the blind area under this condition; it can also realize intelligent adjustment of the rear mirror parameters without collecting user private information; subsequently, it can intelligently determine whether each driver is based on the user's setting screen, and then quickly adjust and determine, that is, it can realize subsequent rapid adjustment of drivers with records; because the user's adjustment habit for the driver's seat often adjusts it to a predetermined position, so that the subsequent rapid adjustment of the record can be realized only by relying on the rear mirror parameters; if the rapid adjustment is unqualified, the user can also adjust according to the above operation and re-determine the setting screen. From a safety perspective, the blind spot area can be adaptively expanded later.

[0025] In one embodiment, each rear-view mirror parameter and setting screen is recorded in real time, integrated into adjustment record data, and each adjustment record data is deduplicated; each adjustment record data is classified according to its corresponding user perspective to obtain each application classification, and the application classification is composed of its corresponding adjustment record data, that is, one application classification corresponds to a driver who has applied the vehicle; an application classification label corresponding to each application classification is set in the dynamic restriction model, and the user can clearly know which application classification it corresponds to, and can adjust each application classification, such as deleting and other adjustment operations; the user quickly adjusts the rear-view mirror parameters according to each application classification label and determines the blind spot area.

[0026] In one embodiment, a method for matching a corresponding reference material area according to rear mirror parameters includes: Identify each material area corresponding to the rear mirror parameter, and identify the share proportion corresponding to each material area; obtain the setting screen selected by the user under the rear mirror parameter, identify the material area corresponding to the setting screen, and count the number of times the user applies each material area under the rear mirror parameter; Calculate the priority value of each material area in real time according to the formula PY=FE+YC; Where: PY is the priority value; FE is the share value; YC is the number of applications; The material area with the highest priority value is selected as the base material area.

[0027] The blind spot analysis module is used to evaluate and analyze the initial image of the blind spot, obtain the monitoring data of each subject, analyze the monitoring data of each subject through the comprehensive impact model, and obtain the comprehensive impact value corresponding to each blind spot position in the initial image of the blind spot; divide the initial image of the blind spot into a number of unit areas according to the comprehensive impact value corresponding to each blind spot position; and use the average value of the comprehensive impact value corresponding to each blind spot position in the unit area as the unit accuracy value of the unit area; Mark the accuracy value of each unit in each unit area; estimate the risk value of each target object and mark it as the benchmark risk value. The benchmark risk value is estimated based on the initial image of the blind spot, that is, the risk of the current target object position is estimated using the existing risk estimation technology to obtain the corresponding risk value; perform blind spot warning based on the benchmark risk value of each target object.

[0028] In one embodiment, the method of dividing the initial blind spot map into a plurality of unit areas according to the comprehensive impact value of each blind spot position includes: Step SA1: identifying the comprehensive impact value of each blind spot position; taking each blind spot position close to the blind spot area boundary or unit area boundary of the vehicle as the initial position; identifying each blind spot position adjacent to the initial position, marking them as selected positions, and adjacent initial positions are regarded as selected positions for each other; The comprehensive impact value of the initial position is marked as QY1 (x，y，z), (x, y, z) represents the corresponding position coordinates; the comprehensive influence value of the selected position is marked as QY2 (x，y，z) ; According to the formula YM=|QY1 (x，y，z) -QY2 (x，y，z) | Calculate the first impact difference between the initial position and the selected position; Where: YM is the first impact difference; The initial position whose first influence difference is less than the threshold value X1 is merged with the selected position to obtain the initial area, where the threshold value X1 is set according to a standard that can be regarded as correspondingly the same, such as 0, 1, 2, 3, etc.; Step SA2: Mark each blind spot position adjacent to the initial area as a selected position; According to the formula calculating a second influence difference between the initial region and the selected location; Where: YM´ is the second impact difference; i represents the corresponding blind spot position in the initial area, i=1, 2, ..., n, n is a positive integer; QY (x，y，z)i Indicates the comprehensive impact value of the corresponding blind spot position in the initial area; QY´ (x，y，z) Indicates the comprehensive influence value corresponding to the selected position; merges the initial area whose second influence difference is less than the threshold X1 with the selected position to obtain a new initial area; Step SA3: looping step SA2 until the initial areas cannot be merged, marking the initial areas as unit areas; identifying whether there are still unmerged blind areas, which means that the initial areas cannot be merged any more; When there are no unmerged blind zone positions, the analysis ends; When there are unmerged blind spot positions, return to step SA1.

[0029] In one embodiment, the problem of blind spot assessment accuracy causes a certain error in the baseline risk. Therefore, the baseline risk value can be corrected in combination with the unit accuracy value corresponding to the target object, and a blind spot warning is performed using the corrected baseline risk value.

[0030] In one embodiment, the method for correcting the benchmark risk value with the unit precision value includes: A correction model is established based on a neural network such as a CNN network or a DNN network, and a corresponding training set is established manually for training. The training set includes input data and output data. The input data is the target object position, the baseline risk value, and the unit accuracy value; the output data is the corrected baseline risk value; and the correction is performed through the correction model after successful training. In one embodiment, corrections may also be made based on other existing technologies.

[0031] In one embodiment, when a large number of corrected baseline risk values ​​are available, one will be able to intuitively understand under what circumstances the risk difference exceeds the range, determine the corresponding difference cause data, that is, the corresponding monitoring data of each subject, and integrate them into abnormal optimization data; the baseline risk value can also be combined for abnormal judgment. The larger the baseline risk value, the higher the requirement, that is, the smaller the risk difference allowed between the two; subsequently, it is convenient for the platform to optimize each monitoring subject according to the abnormal optimization data to reduce possible deviation risks.

[0032] A blind spot accuracy assessment method based on AI camera technology, the method comprising: Determine each influencing factor and identify the monitoring subject corresponding to each influencing factor; Set the comprehensive range of the vehicle's blind spot; obtain the main material data of each monitoring subject, perform a single simulation on the main material data of each monitoring subject, and obtain the factor influence curve corresponding to each blind spot position within the comprehensive range of the blind spot for each influencing factor, the horizontal axis of the factor influence curve is the main monitoring data of the corresponding monitoring subject, and the vertical axis is the accuracy influence value corresponding to the corresponding subject monitoring data, and the value range of the accuracy influence value is [0, 100]; perform a comprehensive analysis based on the influence curves of each factor corresponding to each blind spot position, and form a comprehensive influence model corresponding to the comprehensive range of the blind spot; Conduct real-time monitoring of the monitoring subjects corresponding to each influencing factor to obtain the subject monitoring data corresponding to each influencing factor; Determine the blind spot area, collect image data around the vehicle in real time through the AI ​​camera, perform feature recognition on the image data through the preset image recognition model, and obtain target information of each target object; Determine relative data between the target object and the vehicle according to each target information, integrate the relative data with the target information, and form comprehensive target information of each target object; Generate a blind spot image according to the comprehensive target information of the blind spot area and each target object, and mark it as the blind spot initial image; The monitoring data of each subject is analyzed through the comprehensive impact model to obtain the comprehensive impact value corresponding to each blind spot position in the initial blind spot image; the initial blind spot image is divided into several unit areas according to the comprehensive impact value corresponding to each blind spot position, and the unit accuracy value of each unit area is set, and the unit accuracy value is the average value of the comprehensive impact value of each blind spot position in the unit area; Mark the accuracy value of each unit in each unit area; estimate the baseline risk value of each target object, identify the target object corresponding to the baseline risk value, identify the unit accuracy value corresponding to the target object, correct the baseline risk value according to the unit accuracy value, and perform blind spot warning through the corrected baseline risk value.

[0033] The above formulas are all calculated by removing dimensions and taking numerical values. The formula is a formula that is closest to the actual situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and preset thresholds in the formula are set by technical personnel in this field according to actual conditions or obtained by simulating a large amount of data.

[0034] 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, device, article or method 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, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0035] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, is also included in the patent protection scope of the present invention. It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

[0036] For the convenience of description, the above devices are described separately according to their functions. Of course, when implementing the present application, the functions of each unit can be implemented in one or more software and / or hardware. It should be understood by those skilled in the art that the embodiments of the present invention can be provided as methods, systems or computer program products. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0037] 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 processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 The functionality specified in one or more boxes.

[0038] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A function specified in a process or multiple boxes.

[0039] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

Claims

1. A blind spot accuracy assessment system based on AI camera technology, characterized in that: It includes information module, subject monitoring module, blind spot module and blind spot analysis module; The information module is used to perform information management, determine various influencing factors, and identify the monitoring subject corresponding to each of the influencing factors; The comprehensive range of the blind spot of the vehicle is set; the main material data of each monitoring subject is obtained, and the main material data of each monitoring subject is simulated in a single manner to obtain the factor influence curve corresponding to each blind spot position of each influencing factor within the comprehensive range of the blind spot, wherein the horizontal axis of the factor influence curve is the main monitoring data of the corresponding monitoring subject, and the vertical axis is the accuracy influence value corresponding to the corresponding subject monitoring data, and the value range of the accuracy influence value is [0, 100]; a comprehensive analysis is performed according to the influence curves of each factor corresponding to each blind spot position to form a comprehensive influence model corresponding to the comprehensive range of the blind spot; The subject monitoring module performs real-time monitoring on the monitoring subject corresponding to each influencing factor to obtain subject monitoring data corresponding to each influencing factor; The blind area module is used to generate an initial image of the blind area in real time; The blind spot analysis module is used to evaluate and analyze the initial blind spot image, obtain monitoring data of each subject, analyze the monitoring data of each subject through the comprehensive impact model, and obtain the comprehensive impact value corresponding to each blind spot position in the initial blind spot image; divide the initial blind spot image into a plurality of unit areas according to the comprehensive impact value corresponding to each blind spot position, and set a unit precision value for each unit area, wherein the unit precision value is the average value of the comprehensive impact value of each blind spot position in the unit area; Mark each unit precision value in each unit area; Estimate the baseline risk value of each target object and issue blind spot warning based on the baseline risk value of each target object.

2. According to claim 1, a blind spot accuracy assessment system based on AI camera technology is characterized in that: The method for generating the initial image of the blind area includes: Determine the blind spot area, collect image data around the vehicle in real time through the AI ​​camera, perform feature recognition on the image data through a preset image recognition model, and obtain target information of each target object; Determine relative data between the target object and the vehicle according to each target information, and integrate the relative data with the target information to form comprehensive target information of each target object; A blind spot image is generated according to the blind spot area and the target comprehensive information of each target object, and is marked as a blind spot initial image.

3. According to claim 2, a blind spot accuracy assessment system based on AI camera technology is characterized in that: Methods for determining the blind spot area include: Obtaining blind spot material data of the vehicle, the blind spot material data including rear mirror parameters and corresponding material areas; performing statistics on each blind spot material data to obtain each material area corresponding to each rear mirror parameter and a share ratio corresponding to each material area; Establishing a three-dimensional appearance model corresponding to the vehicle, marked as a vehicle appearance model; restricting and adjusting the vehicle appearance model according to each of the material areas to form a dynamic restriction model; Identify rear mirror parameters of the current vehicle, match corresponding benchmark material areas according to the rear mirror parameters, display the dynamic restriction model to the user according to the benchmark material area, the user determines the setting screen through the dynamic restriction model, adjusts the rear mirror parameters according to the setting screen, and determines the blind spot area.

4. According to claim 3, a blind spot accuracy assessment system based on AI camera technology is characterized in that: The method for matching the corresponding reference material area according to the rear mirror parameters includes: Identify each material area corresponding to the rear mirror parameter, and obtain the share proportion corresponding to each material area; obtain the number of times the user uses each material area under the rear mirror parameter; Calculate the priority value of each material area in real time according to the formula PY=FE+YC; Where: PY is the priority value; FE is the share value; YC is the number of applications; The material area with the highest priority value is selected as the reference material area.

5. According to claim 3, a blind spot accuracy assessment system based on AI camera technology is characterized in that: After the user uses the vehicle appearance model to determine the blind spot area, the parameters and setting screens of each rear mirror are recorded in real time, integrated into adjustment record data, and each adjustment record data is deduplicated; Classifying each of the adjustment record data to obtain each application classification; setting an application classification label corresponding to each application classification in the dynamic restriction model; When the user needs to determine the blind spot area again, the determination is performed through the corresponding application classification label in the dynamic restriction model.

6. According to claim 1, a blind spot accuracy assessment system based on AI camera technology is characterized in that: The method of dividing the initial blind area map into several unit areas according to the comprehensive influence value of each blind area position includes: Step SA1: identifying the comprehensive impact value of each blind spot position; determining each initial position; marking each blind spot position adjacent to the initial position as a selected position; The comprehensive impact value of the initial position is marked as QY1 (x，y，z) , (x, y, z) represents the corresponding position coordinates; the comprehensive influence value of the selected position is marked as QY2 (x，y，z) ; According to the formula YM=|QY1 (x，y，z) -QY2 (x，y，z) | Calculate the first impact difference between the initial position and the selected position; Where: YM is the first impact difference; Merge the initial position where the first impact difference is less than the threshold X1 with the selected position to obtain an initial area; Step SA2: Mark each blind spot position adjacent to the initial area as a selected position; According to the formula calculating a second influence difference between the initial region and the selected location; Where: YM´ is the second impact difference; i represents the corresponding blind spot position in the initial area, i=1, 2, ..., n, n is a positive integer; QY (x，y，z)i Indicates the comprehensive impact value of the corresponding blind spot position in the initial area; QY´ (x，y，z) Indicates the comprehensive impact value corresponding to the selected position; Merge the initial region whose second impact difference is less than the threshold value X1 with the selected position to obtain a new initial region; Step SA3: loop step SA2 until the initial area cannot be merged, and mark the initial area as a unit area; When there is no unmerged blind spot position in the blind spot area, the analysis ends; When there are unmerged blind spot positions in the blind spot area, return to step SA1.

7. According to claim 1, a blind spot accuracy assessment system based on AI camera technology is characterized in that: Before performing a blind spot warning based on a baseline risk value, a target object corresponding to the baseline risk value is identified, a unit precision value corresponding to the target object is identified, the baseline risk value is corrected based on the unit precision value, and a blind spot warning is performed using the corrected baseline risk value.

8. The blind spot accuracy assessment system based on AI camera technology according to claim 7, characterized in that: Record the revised benchmark risk value in real time, mark the revised benchmark risk value as the assessed risk value, and calculate the risk difference between the assessed risk value and the corresponding benchmark risk value; Determine abnormal optimization data based on the risk difference and the benchmark risk value, and send each abnormal optimization data to the platform.

9. A blind spot accuracy assessment method based on AI camera technology, characterized in that: A blind spot accuracy assessment system based on AI camera technology applied to any one of claims 1 to 8, the method comprising: Determine each influencing factor and identify the monitoring subject corresponding to each influencing factor; Set the comprehensive range of the blind spot of the vehicle; obtain the main material data of each monitoring subject, perform a single simulation on the main material data of each monitoring subject, and obtain the factor influence curve corresponding to each blind spot position within the comprehensive range of the blind spot for each influencing factor; perform a comprehensive analysis based on the factor influence curves of each blind spot position to form a comprehensive influence model corresponding to the comprehensive range of the blind spot; Conduct real-time monitoring of the monitoring subjects of each influencing factor to obtain the main monitoring data of each influencing factor; Determine the blind spot area, collect image data around the vehicle in real time through the AI ​​camera, perform feature recognition on the image data through the preset image recognition model, and obtain the target information of each target object; determine the relative data between the target object and the vehicle based on each target information, integrate the relative data with the target information to form the target comprehensive information of each target object; generate a blind spot image based on the blind spot area and the target comprehensive information of each target object, and mark it as the blind spot initial image; The monitoring data of each subject is analyzed through the comprehensive impact model to obtain the comprehensive impact value corresponding to each blind spot position in the initial blind spot image; the initial blind spot image is divided into several unit areas according to the comprehensive impact value corresponding to each blind spot position, and the unit accuracy value of each unit area is set; Mark the accuracy value of each unit in each unit area; estimate the baseline risk value of each target object, identify the unit accuracy value corresponding to the unit area where the target object is located, correct the baseline risk value according to the unit accuracy value, and issue a blind spot warning through the corrected baseline risk value.

Images