Violation inspection system and method based on self-driving car visual sensing

Abstract

A violation inspection system and method based on self-driving car visual sensing, which receives road condition images taken by self-driving cars to identify the image objects contained therein through artificial intelligence, and continuously detects the movement status of image objects to generate image object dynamics. When the image object satisfies both the trigger object factor and at least one of the associated object factors, the image object and its image object dynamics are input into a context recognition model based on artificial intelligence to output the corresponding current context, and when the current context is matched with a violation context template, the image object and the corresponding violation item corresponding to the trigger object factor and the associated object factor corresponding to the violation context template are output, so as to achieve the technical effect of improving the accuracy and diversity of judging violation behaviors.

Description

Technical Field

[0001] The present invention relates to a violation inspection system and method thereof, in particular to a violation inspection system and method thereof based on self-driving car visual sensing. Background Art

[0002] In recent years, with the popularization and vigorous development of self-driving car technology, it is no longer uncommon to see self-driving cars on the road. Since self-driving cars themselves have camera components and distance measurement components, and can identify various objects (or "objects") on the road, and cooperate with positioning systems to obtain positioning coordinates, there is the possibility of combining with other applications. Therefore, how to make full use of self-driving cars to provide more added value has become one of the problems that manufacturers are eager to solve.

[0003] Generally speaking, when a traditional self-driving car is moving, it will continue to obtain environmental conditions through its own camera components and distance measurement components, so that it can drive on the road based on the environmental conditions, and make real-time adjustments and corrections according to changes in environmental conditions, so it is very suitable to replace the current fixed or fixed-route violation prevention system. However, the current self-driving car cannot automatically determine whether there is a violation in front of the vehicle through the camera components and distance measurement components. It can only record video for evidence first, and then manually identify it afterwards, so there is a problem of insufficient convenience.

[0004] In view of this, some manufacturers have proposed a technical means to automatically identify whether a violation has occurred by combining image recognition. The method inputs image data into a trained artificial intelligence module (Model) to identify the image object, and determines whether a violation has occurred based on the position change of the image object. However, this method can usually only identify a single violation in the same picture, such as: vehicles not giving way to pedestrians, illegal left or right turns, crossing double white lines, illegal parking, pedestrians crossing the road, etc., and as the complexity of the image increases, the recognition accuracy decreases. In other words, it is impossible to accurately identify multiple violations at the same time, so there is a problem of insufficient accuracy and lack of diversity in the judgment of violations.

[0005] In summary, it can be seen that the existing technology has long had the problem of insufficient accuracy and lack of diversity in judging violations, so it is necessary to propose improved technical means to solve this problem. Summary of the invention

[0006] The invention discloses a violation inspection system and method based on self-driving car visual sensing.

[0007] Firstly, the present invention discloses a violation inspection system based on self-driving car visual sensing, which is used to perform violation inspection on the inspection path. The system comprises: a microprocessor, a violation item database, an object type database, a context recognition module and a violation determination module. Among them, the microprocessor is used to connect with the visual sensing module of the self-driving car, and continuously receive the road condition image transmitted by the visual sensing module, and then identify the image objects contained in the road condition image through the object recognition algorithm based on artificial intelligence, and continuously detect the movement state of each image object to generate the image object dynamics corresponding to each image object; the violation item database is used to store violation scenario templates, and each violation scenario template corresponds to at least one violation item; the object type database is used to store associated object factors and trigger object factors; the scenario recognition module is connected to the microprocessor and the object type database, and is used to determine that when the image object satisfies at least one of the trigger object factor and the associated object factor at the same time, the image object and its corresponding image object dynamics are input into the scenario recognition model based on artificial intelligence to output the corresponding current scenario; and the violation determination module is connected to the scenario recognition module and the violation item database, and is used to compare the current scenario with the violation scenario template, and when the comparison is consistent, the image object and the corresponding violation item corresponding to the trigger object factor and the associated object factor corresponding to the violation scenario template are output.

[0008] In addition, the present invention also discloses a violation inspection method based on self-driving car visual sensing, which is used to perform violation inspection on the inspection path, and its steps include: providing a violation item database and an object type database, the violation item database stores violation scenario templates, each violation scenario template corresponds to at least one violation item, and the object type database stores associated object factors and trigger object factors; continuously receiving road condition images, and identifying image objects contained in the road condition images through an object recognition calculation model based on artificial intelligence, and continuously detecting the movement state of each image object to generate image object dynamics corresponding to each image object; when it is determined that the image object simultaneously satisfies at least one of the trigger object factor and the associated object factor, the image object and its corresponding image object dynamics are input into the scenario recognition model based on artificial intelligence to output the corresponding current scenario; and comparing the current scenario with the violation scenario template, and when the comparison is consistent, outputting the image object and the corresponding violation item corresponding to the trigger object factor and the associated object factor corresponding to the violation scenario template.

[0009] The system and method disclosed in the present invention are as described above. The difference from the prior art is that the present invention receives road condition images taken by a self-driving car to identify the image objects contained therein through artificial intelligence, and continuously detects the movement status of the image objects to generate image object dynamics. When the image object simultaneously satisfies the trigger object factor and at least one of the associated object factors, the image object and its image object dynamics are input into a situation recognition model based on artificial intelligence to output the corresponding current situation, and when the current situation is matched with the violation situation template, the image object and the corresponding violation item corresponding to the trigger object factor and the associated object factor corresponding to the violation situation template are output.

[0010] Through the above-mentioned technical means, the present invention can achieve the technical effect of improving the accuracy and diversity of judging illegal behaviors. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 The system block diagram of the violation inspection system based on self-driving car visual sensing of the present invention.

[0012] Figure 2A The present invention is a method flow chart of the violation inspection method based on self-driving car visual sensing.

[0013] Figure 2B This is a method flow chart of another embodiment of the violation inspection method based on self-driving car visual sensing of the present invention.

[0014] Figure 3 The figure is a schematic diagram of applying the present invention to conduct illegal inspection.

[0015] Figure 4 A schematic diagram of outputting image objects and violation items related to violation events using the present invention.

[0016] The following are the descriptions of the reference numerals:

[0017] 100 Vision Sensing Module

[0018] 110 Microprocessor

[0019] 120 Illegal Project Database

[0020] 130 Object Type Database

[0021] 140 Context Identification Module

[0022] 150 Violation determination module

[0023] 300 Road Condition Images

[0024] 311 Previous location

[0025] 312,322: pedestrian items

[0026] 313,323 Image object dynamics

[0027] 314 Line marking objects

[0028] Register Object No. 320

[0029] 321 Vehicle Objects

[0030] 400 Output screen

[0031] 410,420 Violation items display block

[0032] 411,423 pedestrian items

[0033] Register Object No. 421

[0034] 422 Vehicle Objects

[0035] Step 210 provides a violation item database and an object type database. The violation item database stores at least one violation scenario template, each violation scenario template corresponds to at least one violation item. The object type database stores at least one associated object factor and one trigger object factor.

[0036] Step 220 continuously receives a road image, and identifies a plurality of image objects contained in the road image through an object recognition algorithm model based on artificial intelligence, and continuously detects the movement state of each of the image objects to generate an image object dynamic state corresponding to each of the image objects.

[0037] Step 230 determines that when the image object satisfies at least one of the trigger object factor and the associated object factor, the image object and its corresponding image object are dynamically input into at least one context recognition model based on artificial intelligence to output a corresponding current context.

[0038] Step 240 compares the current situation with the violation situation template, and when the comparison is consistent, outputs the trigger object factor corresponding to the violation situation template and the image object corresponding to the associated object factor and the corresponding violation item

[0039] Step 250 receives a hotspot of illegal activities reported dynamically, and dynamically adjusts the inspection route according to the hotspot of illegal activities, so that the self-driving car goes to the hotspot of illegal activities and performs illegal activities in real time.

[0040] Step 260 performs face recognition and license plate recognition or both on the output image object to generate a recognition result, and transmits the recognition result to an external database to query whether it matches a notification object, and if it matches, transmits the output image object and its corresponding violation coordinates to a notification unit. DETAILED DESCRIPTION

[0041] The following will describe the implementation methods of the present invention in detail with reference to drawings and examples, so that the implementation process of how the present invention applies technical means to solve technical problems and achieve technical effects can be fully understood and implemented accordingly.

[0042] Please read first Figure 1 , Figure 1 The system block diagram of the violation inspection system based on the visual sensing of the self-driving car of the present invention is used to perform violation inspection on the inspection path. The system includes: a microprocessor 110, a violation item database 120, an object type database 130, a situation recognition module 140 and a violation determination module 150. The microprocessor 110 is used to connect with the visual sensing module 100 of the self-driving car, and continuously receive the road condition image transmitted by the visual sensing module 100, and then identify the image objects contained in the road condition image through the object recognition algorithm based on artificial intelligence, and continuously detect the movement state of each image object to generate the image object dynamics corresponding to each image object. For example, assuming that the road image includes pedestrians, vehicles, road markings and traffic signs, through artificial intelligence object recognition algorithms, such as: Deep Learning's Region based Convolutional Neural Network (R-CNN), Fast Region based Convolutional Neural Network (Fast R-CNN), Faster Region based Convolutional Neural Network (Faster R-CNN), "Mask R-CNN" and "YOLO" models, pedestrians, vehicles, road markings and traffic signs can be identified as image objects, and the movement state (e.g., movement trajectory) of each image object can be continuously detected to generate image object dynamics of the corresponding image object. In actual implementation, the microprocessor 110 can be replaced or implemented using a microcontroller unit (MCU), a central processing unit (CPU), a system on a chip (SoC), etc.

[0043] The violation item database 120 is used to store violation scenario templates, each violation scenario template corresponding to at least one violation item. For example, scenarios including at least one violation item, such as: vehicles not giving way to pedestrians, illegal left or right turns, crossing double white lines, illegal parking, pedestrians crossing the road, etc. or any combination thereof, are provided in advance as violation scenario templates. Taking the violation scenario template of pedestrians crossing the road as an example, it includes the image object dynamics of a pedestrian object crossing the road on a road without a line object.

[0044] The object type database 130 is used to store associated object factors and trigger object factors. Taking the associated object factor as an example, it may include: driving objects, object objects, signal objects, marking objects and scene objects, etc.; taking the trigger object factor as an example, it may include pedestrian objects. In actual implementation, the trigger object factor means that as long as an object that meets the trigger object factor appears in the road image, the violation judgment process is triggered, and at this time, the object of the associated object factor will be coordinated to judge the violation items together.

[0045] The situation recognition module 140 is connected to the microprocessor 110 and the object type database 130, and is used to determine that when the image object satisfies at least one of the trigger object factor and the associated object factor, the image object and its corresponding image object are dynamically input into the situation recognition model based on artificial intelligence to output the corresponding current situation. For example, assuming that the image object determined is a pedestrian object, a driving object, and a line object, since the pedestrian object is the trigger object factor, and the driving object and the line object are the associated object factors, it is determined that the image object satisfies at the same time, so the image object and its corresponding image object are dynamically input into the situation recognition model to output the corresponding current situation, such as: a situation of pedestrians walking, a situation of vehicles turning left, or a situation that includes pedestrians walking and vehicles turning at the same time. In fact, the artificial intelligence-based situation recognition model can be completed by long-term situation modeling training with big data (objects, scenes) and deep learning. For example, two convolutional neural network (CNN) models are used, wherein a fast regional convolutional neural network (Faster R-CNN) is used to detect the presence of pedestrians or vehicles, and a deep convolutional neural network (DCNN) is used to identify the movement state of pedestrians or vehicles, and finally the integration output is the current situation. Similarly, this method can also be applied to establish a violation situation template, and the violation situation template can include image objects and their movement states of associated object factors and triggering object factors. In addition, the current situation can include object images and their corresponding image object dynamics. When it is detected that the image object dynamics of the triggering object factor are blocked by the associated object factor and meet the violation situation template, the corresponding image object is cancelled as a violation object, that is, the violation event caused by the blocking of the associated object factor and the image object corresponding to the triggering object factor will be regarded as the image object corresponding to the triggering object factor. For example, if the dynamic image object of a pedestrian object is blocked by an object and causes a violation, then the pedestrian object will be considered not to have violated the law.

[0046] The violation determination module 150 is connected to the situation identification module 140 and the violation item database 120 to compare the current situation with the violation situation template, and when the comparison is consistent, the trigger object factor corresponding to the violation situation template and the image object corresponding to the associated object factor and the corresponding violation item are output. For example, when the current situation is a situation that includes both pedestrians walking and vehicles turning, it will be compared with the violation situation template that includes both pedestrians crossing the road and vehicles not giving way to pedestrians. In actual implementation, when the current situation is matched with the violation situation template, the camera component can be triggered to record, and the positioning device can be triggered synchronously to obtain the violation coordinates, and the violation coordinates are stored in the violation item database 120 to correspond to the corresponding violation situation template. In this way, when the self-driving car moves to the violation coordinates in the future, the violation situation template corresponding to the violation coordinates can be loaded first to compare with the current situation, and the judgment efficiency and accuracy can be effectively improved by giving priority to the violation situation that has occurred at the fixed coordinate position. In addition, when the violation coordinates are obtained, the optimal inspection path can be dynamically calculated according to the distribution status of the violation coordinates, and the original inspection path can be replaced by this optimal inspection path. For example, assuming there are two violation coordinates, the shortest path passing through the two violation coordinates can be calculated according to the image information as the optimal inspection path. In addition, the violation determination module 150 can also perform face recognition and license plate recognition or either of the two according to the output image object to generate recognition results (such as: facial features, license plate information, etc.), and transmit this recognition result to an external database (such as: household registration database, vehicle registration database, etc.) to query whether it matches the notification object, and if it matches, the output image object and its corresponding violation coordinates are transmitted to the notification unit (such as: procuratorate, police agency and other judicial units).

[0047] It should be particularly noted that in actual implementation, the modules described in the present invention can be implemented in various ways, including software, hardware or any combination thereof. For example, in some embodiments, each module can be implemented using software and hardware or one of them. In addition, the present invention can also be partially or completely implemented based on hardware. For example, one or more modules in the system can be implemented by integrated circuit chips, system single chips, complex programmable logic devices (CPLD), field programmable gate arrays (FPGA), etc. The present invention can be a system, method and / or computer program. The computer program can include a computer-readable storage medium, which carries computer-readable program instructions for enabling a processor to implement various aspects of the present invention. The computer-readable storage medium can be a tangible device that can hold and store instructions used by an instruction execution device. The computer-readable storage medium can be, but is not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device or any suitable combination of the above. More specific examples of computer-readable storage media (a non-exhaustive list) include: a hard disk, a random access memory, a read-only memory, a flash memory, an optical disk, a floppy disk, and any suitable combination of the above. The computer-readable storage medium used herein is not interpreted as a transient signal itself, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagated through a waveguide or other transmission medium (e.g., an optical signal through a fiber optic cable), or an electrical signal transmitted through a wire. In addition, the computer-readable program instructions described herein can be downloaded from the computer-readable storage medium to each computing / processing device, or downloaded to an external computer device or external storage device through a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmission, wireless transmission, routers, firewalls, switches, hubs, and / or gateways. The network card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.The computer program instructions for performing the operations of the present invention may be assembly language instructions, instruction set architecture instructions, machine instructions, machine-dependent instructions, microinstructions, firmware instructions, or source code or object code (Object Code) written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Common Lisp, Python, C++, Objective-C, Smalltalk, Delphi, Java, Swift, C#, Perl, Ruby and PHP, and conventional procedural programming languages ​​such as C language or similar programming languages. The computer program instructions may be executed completely on the computer, partially on the computer, as a stand-alone software, partially on the client computer and partially on the remote computer, or completely on the remote computer or server.

[0048] See also Figure 2A , Figure 2A The method flow chart of the violation inspection method based on self-driving car visual sensing of the present invention is used to perform violation inspection on the inspection path, and the steps include: providing a violation item database 120 and an object type database 130, wherein the violation item database 120 stores violation scenario templates, each violation scenario template corresponds to a violation item, and the object type database 130 stores associated object factors and trigger object factors (step 210); continuously receiving road condition images, and identifying image objects contained in the road condition images through an object recognition calculation model based on artificial intelligence, and continuously detecting the movement state of each image object to generate image object dynamics corresponding to each image object (step 220); when it is determined that the image object satisfies at least one of the trigger object factor and the associated object factor at the same time, the image object and its corresponding image object dynamics are input into the context recognition model based on artificial intelligence to output the corresponding current context (step 230); and comparing the current context with the violation scenario template, and when the comparison is consistent, outputting the image object and the corresponding violation item corresponding to the trigger object factor and the associated object factor corresponding to the violation scenario template (step 240). In this way, it is possible to receive road images taken by self-driving cars to identify the image objects contained therein through artificial intelligence, and continuously detect the movement status of image objects to generate image object dynamics. When the image object satisfies the trigger object factor and at least one of the associated object factors at the same time, the image object and its image object dynamics are input into the artificial intelligence-based situation recognition model to output the corresponding current situation, and when the current situation is matched with the violation situation sample, the trigger object factor and the associated object factor corresponding to the violation situation sample and the corresponding image object and the corresponding violation item are output, so as to improve the accuracy and diversity of judging violations.

[0049] In addition, after step 240, Figure 2B Indicated, Figure 2B This is a method flow chart of another embodiment of the violation inspection method based on self-driving car visual sensing of the present invention, which further includes: receiving a dynamically reported violation hotspot, and dynamically adjusting the inspection path according to the violation hotspot, so that the self-driving car goes to the violation hotspot to perform violation removal in real time (step 250). For example, when a location A is dynamically reported as a violation hotspot, the self-driving car can dynamically adjust the inspection path according to the violation hotspot in real time to go to perform violation removal. In addition, the self-driving car can also perform face recognition and license plate recognition or either of the two according to the output image object to generate a recognition result, and transmit the recognition result to an external database to query whether it matches the notification object, and if it matches, transmit the output image object and its corresponding violation coordinates to the notification unit (step 260). For example, facial recognition can be performed on the image object to obtain facial features, or license plate recognition can be performed to obtain license plate information, and then these recognition results can be sent to an external database for comparison to check whether they match the reporting object, such as a wanted criminal, a stolen vehicle, etc. If so, the image object and the violation coordinates are sent to the reporting unit, such as the procuratorate, the police agency or other similar judicial units.

[0050] The following match Figure 3 and Figure 4 The following description is given by way of example. Please refer to Figure 3 , Figure 3Schematic diagram of applying the present invention to conduct violation inspection. When the self-driving car performs violation inspection on the inspection path, the self-driving car will continue to generate road condition images 300 through its own visual sensing module. At this time, the microprocessor 110 will continue to receive the road condition images, and then identify multiple image objects contained in the road condition image 300 through an object recognition algorithm based on artificial intelligence, such as pedestrian objects (312, 322), signal objects 320, vehicle objects 321, and road marking objects 314, etc. Specifically, the object recognition algorithm can be implemented by "Bounding Box Predictions", "Non-Max Suppression" or similar algorithms. Then, the moving state of each image object is continuously detected. Taking the pedestrian object 312 as an example, its moving state (moving from the previous position 311 to the current position of the pedestrian object 312) will be detected to generate the corresponding image object dynamics 313; taking the vehicle object 321 as an example, its moving state will be detected to generate the corresponding image object dynamics 323. For example, the optical flow method is used to assign a velocity vector to each pixel in the image, and when moving, the image points are matched one-to-one with the points of the three-dimensional object through projection. The image object is then dynamically analyzed based on the velocity vector characteristics of each pixel, and the corresponding image object dynamics are generated (313, 323).

[0051] Next, the context recognition module 140 determines whether the image object satisfies both the trigger object factor and at least one associated object factor. For example, the pedestrian object (312, 322) is the trigger object factor, and the line object 314, the signal object 320, and the vehicle object 321 are all associated object factors. Since the trigger object factor and the associated object factor exist at the same time, they are considered satisfied. At this time, the image object and its corresponding image object dynamics are input into the context recognition model based on artificial intelligence to output the corresponding current context. That is to say, according to the pedestrian object 312, the vehicle object 321, the signal object 320, the line object 314, and the image object dynamics corresponding to each object, the current contexts are identified, for example: a context including the pedestrian object 312, the signal object 320, and the vehicle object 321, a context including the pedestrian object 312 and the line object 314, and so on. Then, the violation determination module 150 compares the current situation with the violation situation template to confirm which violation items exist in the current situation. For example, the violation situation templates of "pedestrians crossing the road" and "vehicles not giving way to pedestrians" are matched, so the trigger object factors corresponding to the two violation situation templates are output. Taking the first violation situation template (pedestrians crossing the road) that matches as an example, the image objects corresponding to the trigger object factor (i.e., pedestrian object 312) and the associated object factor (line object 314), such as pedestrian and line image, and the corresponding violation item, i.e., pedestrian crossing the road, are output; taking the second violation situation template (vehicles not giving way to pedestrians) that matches as an example, the image objects corresponding to the trigger object factor (i.e., pedestrian object 322) and the associated object factor (i.e., signal object 320 and vehicle object 321), such as pedestrian, signal and vehicle image, and the corresponding violation item, i.e., vehicle not giving way to pedestrians, are output. It should be noted that, in addition to the above examples, a violation scenario template can also include more than one violation item, for example, it can include both the violation items of pedestrians crossing the road and the violation items of vehicles not yielding to pedestrians. Thus, when there are multiple violation items in the same scene, the image objects related to all violation items are stored at the same time.

[0052] like Figure 4 Indicated, Figure 4A schematic diagram of outputting image objects and violation items related to violation events for the application of the present invention. In actual implementation, if the current situation matches the violation situation template, the image related to the violation event can be displayed through the output screen 400 to facilitate subsequent evidence. Taking the above-mentioned pedestrian crossing the road as an example, "Pedestrian crossing the road" can be displayed in the violation item display block 410 in the form of text, and the pedestrian object 411 is displayed below it; taking the vehicle not giving way to pedestrians as an example, "Vehicle not giving way to pedestrians" can be displayed in the violation item display block 420 in the form of text, and the signal object 421, vehicle object 422 and pedestrian object 423 are displayed below it. These objects can be stored in the violation item database through images, images or a combination of the two.

[0053] In summary, it can be seen that the difference between the present invention and the prior art lies in that the road condition images taken by the self-driving car are received to identify the image objects contained therein through artificial intelligence, and the moving state of the image objects is continuously detected to generate image object dynamics. When the image object satisfies the trigger object factor and at least one of the associated object factors at the same time, the image object and its image object dynamics are input into the context recognition model based on artificial intelligence to output the corresponding current context, and when the current context is matched with the violation context template, the image objects and corresponding violation items corresponding to the trigger object factor and the associated object factor corresponding to the violation context template are output, so as to improve the accuracy and diversity of judging violations. This technical means can solve the problems existing in the prior art, and thus achieve the technical effect of improving the accuracy and diversity of judging violations.

[0054] Although the present invention is disclosed as above with the aforementioned embodiments, it is not intended to limit the present invention. Any ordinary technician in the field can make some changes and modifications without departing from the spirit and scope of the present invention. Therefore, the patent protection scope of the present invention shall be based on the definition of the claims attached to this specification.

Claims

1. A violation inspection system based on self-driving car visual sensing, characterized in that: Used to perform illegal inspection on an inspection path, the illegal inspection system comprises: a microprocessor, connected to a visual sensing module of the self-driving car, and continuously receiving a road condition image transmitted by the visual sensing module, and then identifying a plurality of image objects contained in the road condition image through an object recognition algorithm based on artificial intelligence, and continuously detecting a moving state of each of the image objects to generate an image object dynamic corresponding to each of the image objects; A violation item database for storing at least one violation scenario template, each violation scenario template corresponding to at least one violation item; an object type database for storing at least one associated object factor and one triggering object factor; a context recognition module, connected to the microprocessor and the object type database, for determining that the image object satisfies at least one of the trigger object factor and the associated object factor at the same time, and dynamically inputting the image object and its corresponding image object into at least one context recognition model based on artificial intelligence to output a corresponding current context; as well as a violation determination module, connected to the context recognition module and the violation item database, for comparing the current context with the violation context template, and outputting the trigger object factor corresponding to the violation context template and the image object corresponding to the associated object factor and the corresponding violation item when the comparison is consistent; The violation scenario template includes the objects of the associated object factor and the trigger object factor and their movement states, and the current scenario includes the image object and its corresponding image object dynamics. When it is detected that the image object dynamics of the trigger object factor are blocked by the associated object factor and meet the violation scenario template, the corresponding image object is cancelled as a violation object.

2. The violation inspection system based on self-driving car visual sensing according to claim 1 is characterized in that: When the current situation matches the violation situation template, at least one photographic component is triggered to record, and a positioning device is synchronously triggered to obtain a violation coordinate, and the violation coordinate is stored in the violation item database to correspond to the corresponding violation situation template.

3. The violation inspection system based on self-driving car visual sensing according to claim 2 is characterized in that: When the self-driving car moves to the violation coordinates, the violation situation template corresponding to the violation coordinates is preferentially loaded to be compared with the current situation.

4. The violation inspection system based on self-driving car visual sensing according to claim 2 is characterized in that: The microprocessor dynamically calculates an optimal inspection path according to the distribution state of the illegal coordinates, and replaces the original inspection path with the optimal inspection path.

5. The violation inspection system based on self-driving car visual sensing according to claim 1 is characterized in that: The microprocessor receives a hot spot of illegal activities reported dynamically, and dynamically adjusts the inspection path according to the hot spot of illegal activities, so that the self-driving car goes to the hot spot of illegal activities and performs illegal removal in real time.

6. The violation inspection system based on self-driving car visual sensing according to claim 2 is characterized in that: The violation determination module further includes performing a face recognition and a license plate recognition or either of the two based on the output image object to generate a recognition result, and transmitting the recognition result to an external database to query whether it matches a notification object, and when it matches, transmitting the output image object and its corresponding violation coordinates to a notification unit.

7. A violation inspection method based on self-driving car visual sensing, used to perform violation inspection on an inspection path, characterized in that: The steps include: Providing a violation item database and an object type database, wherein the violation item database stores at least one violation scenario template, each violation scenario template corresponds to at least one violation item, and the object type database stores at least one associated object factor and one triggering object factor; Continuously receiving a road condition image, and identifying a plurality of image objects contained in the road condition image through an object recognition algorithm model based on artificial intelligence, and continuously detecting a movement state of each of the image objects to generate an image object dynamic corresponding to each of the image objects; When it is determined that the image object satisfies at least one of the trigger object factor and the associated object factor, the image object and its corresponding image object are dynamically input into at least one context recognition model based on artificial intelligence to output a corresponding current context; and Comparing the current situation with the violation situation template, and when the comparison is consistent, outputting the trigger object factor corresponding to the violation situation template and the image object corresponding to the associated object factor and the corresponding violation item; The violation scenario template includes the objects of the associated object factor and the trigger object factor and their movement states, and the current scenario includes the image object and its corresponding image object dynamics. When it is detected that the image object dynamics of the trigger object factor are blocked by the associated object factor and meet the violation scenario template, the corresponding image object is cancelled as a violation object.

8. The violation inspection method based on self-driving car visual sensing according to claim 7 is characterized in that: The violation inspection method further includes triggering at least one photographic component to record a video when the current situation matches the violation situation template, and synchronously triggering a positioning device to obtain a violation coordinate, and storing the violation coordinate in the violation item database to correspond to the corresponding violation situation template.

9. The violation inspection method based on self-driving car visual sensing according to claim 8 is characterized in that: When the self-driving car moves to the violation coordinates, the violation situation template corresponding to the violation coordinates is preferentially loaded to be compared with the current situation.

10. The violation inspection method based on self-driving car visual sensing according to claim 8, characterized in that: The method further includes the steps of dynamically calculating an optimal inspection path according to the distribution state of the violation coordinates, and replacing the original inspection path with the optimal inspection path.

11. The violation inspection method based on self-driving car visual sensing according to claim 7 is characterized in that: The violation inspection method further includes receiving a violation hotspot reported dynamically, and dynamically adjusting the inspection path according to the violation hotspot, so that the self-driving car goes to the violation hotspot to execute the violation elimination step in real time.

12. The violation inspection method based on self-driving car visual sensing according to claim 8, characterized in that: The violation inspection method further includes performing a face recognition and a license plate recognition or either of the two based on the output image object to generate a recognition result, and transmitting the recognition result to an external database to query whether it matches a notification object, and when it matches, transmitting the output image object and its corresponding violation coordinates to a notification unit.

Images