Traffic supervision system based on AI technology and method thereof

Abstract

The application discloses a traffic supervision system and method based on AI technology, which comprises a data acquisition module, a pretreatment module, a comparative analysis model, a data control module, and the comparative analysis model further comprises a data analysis unit and a data processing unit, and relates to the technical field of traffic supervision. The traffic supervision system and method based on AI technology acquire monitoring images on a specified road through a monitoring camera, then analyze and process the monitoring images to obtain a congestion distance of vehicles on a corresponding analysis lane, combine the congestion distance interval preset in a control list, then reset the passing time of a road intersection, effectively optimize the road condition, and improve the vehicle passing speed; meanwhile, the panoramic images on the specified road are acquired in a multi-point acquisition and collection and summary mode, the accuracy of road data acquisition is improved, the acquisition speed of analysis results is improved, the passing time of the road intersection is conveniently set in advance, and good real-time performance is achieved.

Description

Technical Field

[0001] This invention relates to the field of traffic monitoring technology, specifically to a traffic monitoring system and method based on AI technology. Background Technology

[0002] With the research and advancement of Artificial Intelligence (AI) technology, AI is being applied to various fields, including transportation. Particularly in intelligent traffic management, intelligent regulation stems from AI research, and smart transportation is a crucial component of smart city construction.

[0003] In recent years, the number of cars has grown rapidly. The widespread adoption of automobiles has promoted industrial progress and economic development, but it has also brought about a series of traffic and urban management problems, the most significant of which is traffic congestion. Currently, to address urban congestion and traffic problems, restrictions on predefined vehicle categories (such as dump trucks and large trucks) are often implemented through road signs or fixed displays in lanes. Restrictions are also imposed based on license plate numbers. While these methods can alleviate some traffic congestion, a certain amount of congestion still occurs during weekday rush hours. To address this, relevant departments assign traffic management personnel to direct traffic in designated areas. While this helps resolve congestion, it increases the workload of traffic management personnel, and directing traffic in the center of the road also poses certain safety hazards.

[0004] Therefore, how to achieve intelligent transportation regulation and management has become a major issue in this field. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a traffic monitoring system and method based on AI technology, which solves the problems mentioned in the background section.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a traffic monitoring system based on AI technology, comprising:

[0007] The data acquisition module is used to collect monitoring information 1 and monitoring information 2 of the designated road and send them to the preprocessing module. The designated road is a road area between two traffic lights. Monitoring information 1 is the monitoring image collected at different positions in the corresponding lane when the traffic light at the intersection of the designated road is red. Monitoring information 2 is the monitoring image collected at different positions in the corresponding lane when there are no vehicles on the designated road.

[0008] The preprocessing module is used to divide the specified road into multiple analysis lanes, and to perform image stitching and image grayscale processing on monitoring information one and monitoring information two to obtain background image and real-time image respectively, which are then imported into the pre-trained comparative analysis model.

[0009] The comparative analysis model is communicatively connected to the preprocessing module and the data control module, and includes a data analysis unit and a data processing unit.

[0010] The data analysis unit is used to obtain the route of the corresponding analysis lane from the identification samples and real-time images, and take the center position of the route as the analysis route. Then, it divides multiple analysis nodes on the analysis route and obtains the gray values ​​of the analysis nodes in the identification samples and real-time images. Then, it calculates the gray value difference by using the gray values ​​of the identification samples and real-time images at the analysis nodes. Then, it calculates and analyzes the gray value difference to obtain the analysis sequence list. Then, it calculates the absolute value of the difference between each adjacent analysis node in the analysis sequence list and then sends the result to the data processing unit.

[0011] The distance between two adjacent analysis nodes is a preset value, and the distance between multiple adjacent analysis lanes is the same.

[0012] The data processing unit is used to calculate the distance difference between two sets of adjacent analysis nodes in the analysis sequence table based on the distance values ​​of two sets of adjacent analysis nodes, then compare the distance difference with a preset distance comparison value, and calculate the congestion distance of vehicles in the corresponding analysis lane based on the comparison result, and then send the congestion distance to the data control module.

[0013] The data control module is used to import the congestion distance of vehicles in the corresponding analysis lane into a preset control list, and obtain the corresponding traffic light control time based on the congestion distance and the preset congestion distance range in the control list. Then, the traffic light control time is set to the green light duration in the corresponding analysis lane.

[0014] Preferably, the surveillance images are acquired by multiple surveillance cameras installed on the designated road, and the monitoring areas of two adjacent surveillance cameras are connected.

[0015] Preferably, the specific processing method of the preprocessing module is as follows:

[0016] A1. First, obtain the left-turn lane, straight lane, and right-turn lane on the specified road, and then divide the specified road into multiple analysis lanes according to the left-turn lane, straight lane, and right-turn lane;

[0017] A2. Then, the monitoring images collected by multiple monitoring cameras in each analysis lane are obtained from the monitoring information 1, and they are stitched together to obtain a stitched image, which is then used as the background image of each analysis lane.

[0018] A3. The background images of each analysis lane are converted to grayscale. Then, the grayscale background images are used as recognition samples and imported into the pre-trained comparative analysis model.

[0019] A4. Then, obtain the monitoring images collected by multiple monitoring cameras in each analysis lane from monitoring information 2, stitch them together to obtain a stitched image, and use it as the real-time image of each analysis lane.

[0020] A5. Convert the real-time images of each analysis lane to grayscale, and then import the grayscale-processed real-time images into the pre-trained comparative analysis model.

[0021] Preferably, in steps A2 and A4, the stitched images are obtained in the following way:

[0022] B1. Based on the direction of vehicle travel on the designated road, assign codes to multiple surveillance cameras starting from the end point of the direction of travel on the designated road, and assign the codes in ascending order.

[0023] B2. Subsequently, the monitoring images collected by multiple monitoring cameras in each analysis lane are obtained from monitoring information one and monitoring information two, sorted and stitched according to the corresponding coding order to obtain the corresponding stitched image.

[0024] Preferably, the specific analysis method of the data analysis unit is as follows:

[0025] D11. The center position of the path of the analyzed lane in the identified sample and real-time image is taken as the analysis route;

[0026] D12. The analysis route is then divided into several equally spaced analysis nodes, and the gray values ​​of the corresponding analysis nodes are obtained from the identification samples and real-time images based on the analysis nodes.

[0027] The gray values ​​of the corresponding analysis nodes in the identified samples and real-time images are labeled as Yi and Si, respectively, i = 1, 2, ..., n, where i represents the nth analysis node and n represents the number of analysis nodes;

[0028] D13. Then, by Ci = |Si-Yi|, the grayscale difference Ci of each analysis node is obtained;

[0029] D14, then through Obtain the discrete value CL of the grayscale difference, where Cp is the average value of all grayscale differences when calculating the corresponding discrete value;

[0030] Subsequently, compare the calculated discrete value CL with the preset discrete threshold CL0. If CL > CL0, it is considered that the discrete value CL of this set of data is too large. Delete the corresponding Ci values in descending order of |Ci - Cp| and calculate the remaining discrete values correspondingly until Ci < CL0, and then obtain all the deleted Ci;

[0031] D15. Then obtain the corresponding i values from all the deleted Ci, and sort the corresponding i values in ascending order to obtain the analysis sequence list;

[0032] D16. After that, in the analysis sequence list, obtain the absolute value of the difference between two adjacent i values, and denote it as Gj, where j = 1, 2,..., m. j represents the number of adjacent i values, and m represents the number of pairs of adjacent i values in the sorted analysis sequence list.

[0033] Preferably, the specific processing method of the data processing unit is as follows:

[0034] D21. Subsequently, obtain the distance values between two adjacent analysis nodes, and then mark the distance values as W;

[0035] D22. Then, through the formula GCj = Gj * W, calculate the distance difference GCj between two adjacent i values in the analysis sequence list, and then compare GCj with the preset distance comparison value GC0, and then calculate the congestion distance of the vehicle on the corresponding analysis lane according to the comparison result;

[0036] D23. Subsequently, send the congestion distance of the corresponding analysis lane to the data control module.

[0037] Preferably, the comparison method between GCj and GC0 in step D22 is as follows:

[0038] First, set the value of j to 1. If GCj > GC0, when the value of j is 1, obtain the corresponding two pairs of adjacent i values, and obtain the pair of i values with the smallest i value among the two pairs of adjacent i values;

[0039] Subsequently, through the formula Z = i * W, obtain the congestion distance Z of the vehicle on this analysis lane;

[0040] If GCj < GC0, then increase the value of j by 1, and compare GCj with the preset comparison value GC0 again until the corresponding value of j satisfies GCj > GC0. Subsequently, when GCj > GC0, obtain the corresponding value of j, and then obtain the corresponding two pairs of adjacent i values according to the corresponding value of j, and obtain the pair of i values with the smallest i value among the two pairs of adjacent i values. After that, the same as the above steps, obtain the congestion distance of the vehicle on this analysis lane.

[0041] An image analysis method based on AI technology is provided, which is implemented in the aforementioned AI-based image analysis system.

[0042] Beneficial effects

[0043] This invention provides a traffic monitoring system and method based on AI technology. Compared with existing technologies, it has the following advantages:

[0044] This invention acquires surveillance images of designated roads through monitoring cameras, then analyzes and processes them to obtain the congestion distance of vehicles in the corresponding analysis lanes. By combining this with the preset congestion distance range in the control list, the corresponding traffic light control time is obtained, and the passage time of the traffic lights at intersections is reset, which can effectively optimize road conditions and improve vehicle passage speed.

[0045] This invention employs a multi-point acquisition and aggregation method to obtain panoramic images of a specified road, improving the accuracy of road data acquisition. Simultaneously, the images are preprocessed to facilitate rapid analysis of traffic congestion at the same time point, increasing the speed of obtaining analysis results and facilitating the advance setting of traffic light passage times, thus exhibiting good real-time performance. Attached Figure Description

[0046] Figure 1 This is a system block diagram of the present invention;

[0047] Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0048] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] As an embodiment of the present invention

[0050] Please see Figure 1 This invention provides a technical solution: a traffic monitoring system based on AI technology, comprising:

[0051] The data acquisition module is used to collect monitoring information 1 of the designated road when the traffic light at the intersection is red, and to collect monitoring information 2 of the designated road when there are no vehicles driving. The monitoring information 1 and monitoring information 2 are both monitoring images collected from different locations in the corresponding lanes of the designated road. The monitoring images are obtained by multiple monitoring cameras installed on the designated road.

[0052] The designated road is a road area located between two traffic lights within a road;

[0053] The preprocessing module is used to divide the specified road into multiple analysis lanes and perform image stitching and grayscale processing on monitoring information one and monitoring information two. This technology is existing and will not be elaborated here. The background image and real-time image are obtained separately and then imported into the pre-trained comparative analysis model. The specific processing method is as follows:

[0054] A1. First, obtain the left-turn lane, straight lane, and right-turn lane on the specified road, and then divide the specified road into multiple analysis lanes according to the left-turn lane, straight lane, and right-turn lane;

[0055] A2. Subsequently, in the direction of vehicle travel on the designated road, multiple surveillance cameras are coded and marked starting from the end position of the direction of travel on the designated road, and the marking codes are marked in ascending order;

[0056] A3. Then, the monitoring images collected by multiple monitoring cameras in each analysis lane are obtained from the monitoring information 1, sorted and stitched according to the corresponding coding order to obtain the stitched image, and used as the background image of each analysis lane.

[0057] A4. Convert the background images of each analysis lane to grayscale, then use the grayscale background images as recognition samples and import them into the pre-trained comparative analysis model.

[0058] A5. Then, obtain the monitoring images collected by multiple monitoring cameras in each analysis lane from the monitoring information 2, sort them according to the corresponding coding order and stitch them together to obtain the stitched image, and use it as the real-time image.

[0059] A6. Convert the real-time images of each analysis lane to grayscale, and then import the grayscale-processed real-time images into the pre-trained comparative analysis model.

[0060] The distance between two adjacent analysis nodes is a fixed value, and the distance between multiple adjacent analysis lanes is the same.

[0061] The present invention adopts a method of multi-point acquisition and collective summarization to obtain panoramic images on a specified road, improving the accuracy of road data acquisition. At the same time, the images are preprocessed to facilitate the rapid analysis of the traffic congestion status at the same time point, improving the acquisition speed of the analysis results and facilitating the advance setting of the traffic light passing time, with good real-time performance.

[0062] A comparative analysis model, which includes a data analysis unit and a data processing unit, and the comparative analysis model is used to perform a comparative analysis on the recognition samples and real-time images through the data analysis unit, and obtain an analysis sequence list, and then calculate the absolute value of the difference between adjacent analysis nodes in the analysis sequence list. After that, the data analysis unit calculates the distance difference between two groups of adjacent analysis nodes in the analysis sequence list according to the distance values of the two groups of adjacent analysis nodes. Subsequently, the distance difference is compared with a preset distance comparison value, and according to the comparison result, the congestion distance of the vehicles on the corresponding analysis lane is calculated, and then the result is sent to the data control module; the comparative analysis method is as follows:

[0063] D1. Taking a section of specified road as an example, in the recognition samples and real-time images obtained on this specified road, the central position of the driving route of this analysis lane is used as the analysis route;

[0064] D2. Subsequently, the analysis route is divided into several equally spaced analysis nodes, and at the same time, the gray values of the corresponding analysis nodes are obtained from the recognition samples and real-time images according to the several analysis nodes;

[0065] And the gray values of the corresponding analysis nodes in the recognition samples and real-time images are respectively marked as Yi and Si, i = 1, 2,..., n, i represents the number of the analysis node, and n represents the number of analysis nodes;

[0066] D3. Subsequently, through Ci = |Si - Yi|, the gray value difference Ci of each analysis node is obtained;

[0067] D4. Then through the discrete value CL of the gray value difference is obtained, where Cp is the average value of all gray value differences when calculating the corresponding discrete value;

[0068] Subsequently, the calculated discrete value CL is compared with a preset discrete threshold CL0. If CL > CL0, it is considered that the discrete value CL of this group of data is too large. The corresponding Ci values are sequentially deleted in descending order of |Ci - Cp| and the remaining discrete values are calculated correspondingly until Ci < CL0. Subsequently, all the deleted Ci are obtained;

[0069] D5. Then, obtain the corresponding values of i from all the deleted Cis, sort the corresponding values of i in ascending order to obtain an analysis sequence list. After that, in the analysis sequence list, obtain the absolute value of the difference between two adjacent values of i, and denote it as Gj, where j = 1, 2,..., m. Here, j represents the j-th pair of adjacent i, and m represents the number of pairs of adjacent i in the sorted analysis sequence list.

[0070] D6. Subsequently, obtain the distance values between two adjacent analysis nodes, and the distance values between all adjacent pairs of analysis nodes in the analysis route are the same. Then, mark the distance value as W.

[0071] D7. Next, calculate the distance difference GCj between two adjacent values of i in the analysis sequence list through the formula GCj = Gj * W. Then, compare GCj with a preset comparison value GC0. The comparison method is as follows:

[0072] First, let the value of j be 1. If GCj > GC0, then when the value of j is 1, obtain the corresponding two pairs of adjacent i, and in the two pairs of adjacent i, obtain the pair of i values with the smallest i value. Subsequently, through the formula Z = i * W, obtain the congestion distance Z of the vehicle on this analysis lane.

[0073] If GCj < GC0, then increment the value of j by 1, and compare GCj with the preset comparison value GC0 again until the corresponding value of j satisfies GCj > GC0. Subsequently, when GCj > GC0, obtain the corresponding value of j, and based on the corresponding value of j, obtain the corresponding two pairs of adjacent i, and in the two pairs of adjacent i, obtain the pair of i values with the smallest i value. Then, follow the same steps as above to obtain the congestion distance of the vehicle on this analysis lane.

[0074] D8. Then, send the congestion distance on the corresponding analysis lane to the data control module.

[0075] The data control module is used to import the congestion distance of the vehicle on the corresponding analysis lane into a preset control list, and based on the congestion distance and the preset congestion distance interval in the control list, obtain the corresponding signal light control time. Subsequently, set the signal light control time as the green light duration on the corresponding analysis lane.

[0076] In this invention, a monitoring camera is used to obtain the monitoring images on a specified road. Subsequently, through analysis and processing, the congestion distance of the vehicle on the corresponding analysis lane is obtained. Then, in combination with the preset congestion distance interval in the control list, the corresponding signal light control time is obtained, and the passing time of the traffic lights at the intersection is reset, which can effectively optimize the road conditions and improve the vehicle passing speed.

[0077] As the second embodiment of this invention

[0078] The difference between this embodiment and embodiment one is that in this embodiment, the distance between two adjacent analysis nodes is a preset value, and the distance between multiple adjacent analysis lanes is the same;

[0079] Please see Figure 2 The present invention also provides a technical solution: a traffic monitoring method based on AI technology, which is implemented using the aforementioned AI-based traffic monitoring system, and includes the following steps:

[0080] Step 1: Information Collection

[0081] When the traffic light at the designated intersection is red, multiple monitoring images of the designated road are collected, as well as multiple monitoring images of the designated road in a vehicle-free environment.

[0082] Step 2: Information Preprocessing

[0083] The acquisition results from step one are processed by image stitching and image grayscale conversion to obtain background and real-time images, which are then imported into a pre-trained comparative analysis model.

[0084] Step 3: Image Analysis

[0085] First, the travel route of the corresponding analysis lane is obtained from the recognition samples and real-time images, and the center position of the travel route is used as the analysis route. Then, multiple analysis nodes are divided on the analysis route.

[0086] Then, the gray values ​​of the identification sample and the real-time image at the analysis node are obtained through the pre-trained contrast analysis model, and the gray value difference is calculated. Then, the gray value difference is calculated and analyzed to obtain the analysis sequence list.

[0087] Then, the absolute value of the difference between each adjacent analysis node in the analysis sequence list is calculated;

[0088] Step 4: Calculation and Processing

[0089] Based on the distance values ​​of two sets of adjacent analysis nodes, the distance difference between the two sets of adjacent analysis nodes in the analysis sequence list is calculated. Then, the distance difference is compared with the preset distance comparison value. Based on the comparison result, the congestion distance of vehicles on the corresponding analysis lane is calculated.

[0090] Step 5: Equipment Control

[0091] The congestion distance of vehicles in the corresponding analysis lane is imported into a preset control list. Based on the congestion distance and the preset congestion distance range in the control list, the corresponding traffic light control time is obtained. Then, the traffic light control time is set to the green light duration in the corresponding analysis lane.

[0092] Furthermore, any content not described in detail in this specification is existing technology known to those skilled in the art.

[0093] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A traffic monitoring system based on AI technology, characterized in that, include: The data acquisition module is used to collect monitoring information 1 and monitoring information 2 for a designated road and send them to the preprocessing module. The designated road is a road area between two traffic lights. Monitoring information 1 is the monitoring image collected at different positions in the corresponding lane when the traffic light at the intersection of the designated road is red. Monitoring information 2 is the monitoring image collected at different positions in the corresponding lane when there are no vehicles on the designated road. The preprocessing module is used to divide the specified road into multiple analysis lanes, and to perform image stitching and image grayscale processing on monitoring information one and monitoring information two to obtain background image and real-time image respectively, which are then imported into the pre-trained comparative analysis model. The comparative analysis model is communicatively connected to the preprocessing module and the data control module, and includes a data analysis unit and a data processing unit. The data analysis unit is used to obtain the route of the corresponding analysis lane from the identification samples and real-time images, and take the center position of the route as the analysis route. Then, it divides multiple analysis nodes on the analysis route and obtains the gray values ​​of the analysis nodes in the identification samples and real-time images. Then, it calculates the gray value difference by using the gray values ​​of the identification samples and real-time images at the analysis nodes. Then, it calculates and analyzes the gray value difference to obtain the analysis sequence list. Then, it calculates the absolute value of the difference between each adjacent analysis node in the analysis sequence list and then sends the result to the data processing unit. The distance between two adjacent analysis nodes is a preset value, and the distance between multiple adjacent analysis lanes is the same. The data processing unit is used to calculate the distance difference between two sets of adjacent analysis nodes in the analysis sequence list based on the distance values ​​of two sets of adjacent analysis nodes. Then, it compares the distance difference with a preset distance comparison value. Based on the comparison result, it calculates the congestion distance of vehicles in the corresponding analysis lane and then sends the congestion distance to the data control module. The specific processing method of the data processing unit is as follows: D21. Then obtain the distance values ​​between two sets of adjacent analysis nodes, and then mark the distance values ​​as W; D22. Next, the distance difference GCj between two adjacent i groups in the analysis sequence list is calculated using the formula GCj=Gj*W. Then, GCj is compared with the preset distance comparison value GC0. Based on the comparison result, the congestion distance of vehicles on the corresponding analysis lane is calculated. Here, Gj represents the absolute value of the difference between two adjacent data sets in the analysis sequence list; D23. Subsequently, the congestion distance on the corresponding analyzed lane is sent to the data control module; The data control module is used to import the congestion distance of vehicles in the corresponding analysis lane into a preset control list, and obtain the corresponding traffic light control time based on the congestion distance and the preset congestion distance range in the control list. Then, the traffic light control time is set to the green light duration in the corresponding analysis lane.

2. The traffic monitoring system based on AI technology according to claim 1, characterized in that: The surveillance images are captured by multiple surveillance cameras installed on the designated road, and the monitoring areas of two adjacent surveillance cameras are connected.

3. The traffic monitoring system based on AI technology according to claim 1, characterized in that: The specific processing method of the preprocessing module is as follows: A1. First, obtain the left-turn lane, straight-through lane, and right-turn lane on the specified road, and then divide the specified road into multiple analysis lanes according to the left-turn lane, straight-through lane, and right-turn lane; A2. Subsequently, obtain the surveillance images collected by multiple surveillance cameras in each analysis lane from the surveillance information 1, splice them to obtain a spliced image, and use it as the background image of each analysis lane; A3. Perform image grayscale processing on the background images of each analysis lane, and then use the background images after image grayscale processing as recognition samples and import them into a pre-trained comparative analysis model; A4. Then, obtain the surveillance images collected by multiple surveillance cameras in each analysis lane from the surveillance information 2, splice them to obtain a spliced image, and use it as the real-time image of each analysis lane; A5. Perform image grayscale processing on the real-time images of each analysis lane, and then import the real-time images after image grayscale processing into the pre-trained comparative analysis model.

4. A traffic monitoring system based on AI technology according to claim 3, characterized in that: In steps A2 and A4, the method for obtaining the spliced image is as follows: B1. According to the driving direction of vehicles on the specified road, encode and mark multiple surveillance cameras starting from the end position of the driving direction of the specified road, and the marking codes are marked in ascending order; B2. Subsequently, obtain the surveillance images collected by multiple surveillance cameras in each analysis lane from the surveillance information 1 and the surveillance information 2, sort and splice them according to the corresponding coding order to obtain the corresponding spliced image.

5. A traffic monitoring system based on AI technology according to claim 1, characterized in that: The specific analysis method of the data analysis unit is as follows: D11. In the recognition sample and the real-time image, use the central position of the driving route of the corresponding analysis lane as the analysis route; D12. Subsequently, divide the analysis route into several equally spaced analysis nodes, and at the same time obtain the grayscale values of the corresponding analysis nodes from the recognition sample and the real-time image according to the several analysis nodes; Mark the grayscale values of the corresponding analysis nodes in the recognition sample and the real-time image as Yi and Si respectively, i = 1, 2,..., n, i represents the number of the analysis node, and n represents the number of analysis nodes; D13. Subsequently, obtain the grayscale difference Ci of each analysis node through Ci = |Si - Yi|; D14, then through The discrete value CL of the grayscale difference is obtained, where Cp is the average value of all grayscale differences when calculating the corresponding discrete value; Then compare the calculated discrete value CL with the preset discrete threshold CL0. If CL > CL0, it is considered that the discrete value CL of this set of data is too large. Delete the corresponding Ci values in descending order of |Ci - Cp| and calculate the remaining discrete values correspondingly until Ci < CL0, and then obtain all the deleted Ci; D15. Then obtain the corresponding i values from all the deleted Ci, and sort the corresponding i values in ascending order to obtain an analysis sequence list; D16. Then, in the analysis sequence list, obtain the absolute value of the difference between two adjacent i values and denote it as Gj, j = 1, 2,..., m, j represents the number of the adjacent i, and m represents the number of two adjacent i in the sorted analysis sequence list.

6. A traffic monitoring system based on AI technology according to claim 1, characterized in that: In step D22, the comparison method between GCj and GC0 is as follows: First, set the value of j to 1. If GCj > GC0, when the value of j is 1, obtain the corresponding two groups of adjacent i, and among the two groups of adjacent i, obtain the group of i values with the smallest i value; Subsequently, through the formula Z = i * W, obtain the congestion distance Z of the vehicles on the analysis lane; If GCj < GC0, increase the value of j by 1, and then compare GCj with the preset comparison value GC0 until the corresponding value of j satisfies GCj > GC0. Subsequently, when GCj > GC0, obtain the corresponding value of j, and then according to the corresponding value of j, obtain the corresponding two groups of adjacent i, and among the two groups of adjacent i, obtain the group of i values with the smallest i value. After that, the following steps are the same as above to obtain the congestion distance of the vehicles on the analysis lane.

7. An image analysis method based on AI technology, characterized in that, This method is implemented by an image analysis system based on AI technology described in any one of claims 1-6.

Images