An artificial intelligence-based port violation behavior intelligent identification and analysis system

Abstract

This invention proposes an intelligent identification and analysis system for port violations based on artificial intelligence, belonging to the field of safety production technology. The system involves reading port images sent by a port image acquisition unit, preprocessing the port images, locating license plates based on the preprocessed images to obtain license plate location area images, segmenting the license plate location area images into characters, and performing character recognition on the segmented characters to obtain the final license plate recognition result. This approach can accurately identify license plate information while simplifying the image processing flow, freeing up computing resources, and balancing the accuracy and timeliness of license plate image recognition, thus providing a guarantee for the supervision of port violations.

Description

Technical Field

[0001] This invention belongs to the field of safety production technology, and in particular relates to an intelligent identification and analysis system for port violations based on artificial intelligence. Background Technology

[0002] Ports are crucial pillars for national economic development and international trade. The complex port environment presents numerous safety risks and hazards. Accidents can not only cause casualties and property damage but also severely impact local economic development. Therefore, port safety is paramount. Safe vehicle transport is a vital component of port safety and a critical link in port logistics. It is essential to identify and monitor port violations, particularly the entry and exit of unregistered vehicles into port loading and unloading areas. Computer vision, a key component of artificial intelligence, can be effectively applied to the monitoring, intelligent analysis, and identification of port violations, thereby reducing the workload of regulatory personnel and improving regulatory efficiency. In the prior art of patent number CN112069870A, an image decay model of the original image is constructed. Based on a multi-sided filtering algorithm, edge reconstruction, ghost matrix estimation, and latent image estimation are performed on the original image to obtain an enhanced ghost image. The point spread function of the image decay model is estimated using the original image and the enhanced ghost image. Adaptive deconvolution and the point spread function are used to evaluate the potential sharp image in the original ghost image. In the prior art of patent number CN114913518A, a license plate image is preprocessed based on brightness to obtain a first image. Brightness compensation is performed on the license plate image to avoid the influence of brightness on the recognition accuracy. The first image is then located to obtain the first image. The license plate area in the image is corrected to obtain a second image, which corrects the license plate area and avoids angular interference during license plate recognition. Feature extraction is performed on the second image to obtain target features. Character recognition is performed based on the target features to obtain the license plate recognition result of the license plate image. Then, license plate recognition is realized based on artificial intelligence. The above-mentioned existing patents do not focus on the illegal behavior of port vehicles. The inventor also found through research that there is little analysis of illegal behavior of port vehicles in the existing technology. Moreover, the above-mentioned license plate image processing method requires multiple image transformations and feature adjustments (such as brightness adjustment, image correction and enhancement, image decay adjustment), the image processing process is complex, and the computational resources are large. Summary of the Invention

[0003] To address the aforementioned problems in existing technologies, this invention proposes an intelligent identification and analysis system for port violations based on artificial intelligence. This system simplifies image processing, frees up computing resources, and enables intelligent identification and analysis of violations involving unregistered vehicles entering or leaving port loading and unloading areas.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: an intelligent identification and analysis system for port violations based on artificial intelligence, comprising a port image acquisition unit, a port image analysis unit, a data matching unit, and an early warning unit. The port image acquisition unit is connected to the port image analysis unit, the port image analysis unit is connected to the data matching unit, and the data matching unit is connected to the early warning unit. The port image analysis unit performs license plate recognition of vehicles in port images based on an improved image analysis method, sends the license plate recognition result to the data matching unit, and matches it with a preset license plate in the data matching unit to determine whether the license plate belongs to a registered license plate. When the license plate does not belong to a registered license plate, the data matching unit sends an early warning message to the early warning unit, and the early warning unit issues an early warning based on the early warning message.

[0005] Furthermore, the improved image analysis method includes the following steps: Step S1: Read the port image sent by the port image acquisition unit; Step S2: Perform data preprocessing on the port image to obtain a preprocessed image; Step S3: Perform license plate localization based on the preprocessed image to obtain a license plate localization area image; Step S4: Perform character segmentation on the license plate localization area image to obtain segmented characters; Step S5: Perform character recognition on the segmented characters to obtain the final license plate recognition result.

[0006] Further, the data preprocessing of the port image in step S2 specifically includes the following sub-steps: Step S21: Convert the port image into a grayscale image and filter the grayscale image; Step S22: Perform edge detection on the filtered grayscale image to obtain the preprocessed image.

[0007] Furthermore, the grayscale image is filtered in step S21, specifically as follows:

[0008] ;

[0009] Where I1(x, y) represents the filtered image, x represents the horizontal coordinate of the image, y represents the vertical coordinate of the image, F(i, j) represents the filtering window function, I(x+i, y+j) represents the image before filtering, i represents the horizontal coordinate of the filtering window function bias, j represents the vertical coordinate of the filtering window function bias, k represents the width of the filtering window, and σ represents the standard deviation of the filtering window function.

[0010] Further, step S3, which involves locating the license plate based on the preprocessed image and obtaining the license plate location region, specifically includes the following sub-steps: Step S31: Perform a closing operation on the preprocessed image using a rectangular structuring element; Step S32: Fill holes in the image using the imagefill function; Step S33: Obtain the attributes of all connected regions using the regionprop function; Step S34: Traverse all connected regions and select regions that meet the first condition as the license plate region; Step S35: Obtain bounding box information from the license plate region, and crop the license plate region based on the bounding box information to complete the license plate location.

[0011] Furthermore, the first condition is that the area is greater than 1000 pixels and the aspect ratio is greater than 2 and less than 5; the rectangular structure is a 30-by-30 dimension matrix.

[0012] Further, step S4 involves character segmentation of the license plate location area to obtain segmented characters, specifically including: Step S41: Binarizing the license plate location area image; Step S42: Eroding the binarized license plate location area image using a linear structuring element, and dilating the eroded image to obtain a dilated image; Step S43: Vertically projecting the dilated image to obtain a vertically projected image; Step S44: Setting a dynamic threshold to determine the start and end positions of the characters and performing character segmentation on the vertically projected image based on the dynamic threshold.

[0013] Further, the dynamic threshold in step S43 is the average pixel value of the vertically projected image multiplied by a first adjustment coefficient. The first adjustment coefficient is 0.5.

[0014] Furthermore, the warning unit is a voice warning unit or a flashing light warning unit.

[0015] The beneficial technical effects of this invention compared with the prior art are as follows:

[0016] (1) Set a filter window function to filter the grayscale image before filtering. The filtering effect under different image backgrounds can be adjusted by changing the width of the filter window function, thereby improving the versatility and timeliness of the filtering algorithm.

[0017] (2) Setting a dynamic threshold for character segmentation based on the pixels of the image after vertical projection can reduce missegmentation due to excessive width or narrowness and improve the accuracy of license plate image recognition. Attached Figure Description

[0018] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0019] Figure 1 This is a simplified flowchart of the improved image analysis method involved in this invention;

[0020] Figure 2 Simulation diagram for license plate recognition using the improved image analysis method involved in this invention. Detailed Implementation

[0021] 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.

[0022] The concepts involved in this application will first be described with reference to the accompanying drawings. It should be noted that the following descriptions of various concepts are only for the purpose of making the content of this application easier to understand and do not constitute a limitation on the scope of protection of this application; furthermore, the embodiments and features in the embodiments of this application can be combined with each other unless otherwise specified. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0023] Refer to the instruction manual. Figure 1This invention proposes an intelligent identification and analysis system for port violations based on artificial intelligence, comprising a port image acquisition unit, a port image analysis unit, a data matching unit, and an early warning unit. The port image acquisition unit is connected to the port image analysis unit, which is connected to the data matching unit, and the data matching unit is connected to the early warning unit. The early warning unit is preferably a voice early warning unit or a flashing light early warning unit. The port image analysis unit performs license plate recognition on vehicles in port images based on an improved image analysis method, sends the license plate recognition result to the data matching unit, and matches it with preset license plates in the data matching unit to determine whether the license plate belongs to a registered vehicle. When the license plate does not belong to a registered vehicle, the data matching unit sends an early warning message to the early warning unit, which then issues an early warning based on the warning message. The aforementioned port image acquisition unit mainly filters and enlarges frame images of port monitoring videos to obtain the required port image information; the aforementioned data matching unit contains an information dataset or information database for storing the license plate information of registered vehicles, which is used to match or pair with the license plate information identified by the port image analysis unit to determine whether the vehicle analyzed by the port image analysis unit is a registered vehicle, thereby deciding whether to issue an alert.

[0024] The improved image analysis method includes the following steps: Step S1: Read the port image sent by the port image acquisition unit; Step S2: Perform data preprocessing on the port image to obtain a preprocessed image; Step S3: Perform license plate localization based on the preprocessed image to obtain a license plate localization area image; Step S4: Perform character segmentation on the license plate localization area image to obtain segmented characters; Step S5: Perform character recognition on the segmented characters to obtain the final license plate recognition result.

[0025] Step S2 involves data preprocessing of the port image, specifically including the following sub-steps: Step S21: Convert the port image to a grayscale image and filter the grayscale image; Step S22: Perform edge detection on the filtered grayscale image to obtain the preprocessed image. It is understood that the above edge detection can be based on existing edge detection algorithms using Sobel, Canny, and Scharr operators.

[0026] The filtering process for the grayscale image in step S21 is specifically as follows:

[0027] ;

[0028] Where I1(x, y) represents the filtered image, x represents the horizontal coordinate of the image, y represents the vertical coordinate of the image, F(i, j) represents the filtering window function, I(x+i, y+j) represents the image before filtering, i represents the horizontal coordinate of the filtering window function bias, j represents the vertical coordinate of the filtering window function bias, k represents the filtering window width, and σ represents the standard deviation of the filtering window function (preferably greater than 1 and less than 1.5). It should be noted that the filtering window width k needs to be manually adjusted in the program. The filtering window width k can be set according to the different image backgrounds of the port environment during license plate recognition. When the environment is more complex (i.e., the background contains different colors and the light intensity is weak), a smaller filtering window width k is set to improve the filtering effect; when the environment is simpler (i.e., the background contains a few colors and the light intensity is strong), a larger filtering window width k is set to improve the filtering efficiency.

[0029] Step S3, which involves locating the license plate and obtaining the license plate location region based on the preprocessed image, specifically includes the following sub-steps: Step S31: Perform a closing operation on the preprocessed image using a rectangular structuring element; Step S32: Fill holes in the image using the imagefill function; Step S33: Obtain the attributes of all connected regions using the regionprop function; Step S34: Traverse all connected regions and select regions that meet the first condition as the license plate region; Step S35: Obtain bounding box information from the license plate region, and crop the license plate region based on the bounding box information to complete the license plate location. The imagefill and regionprop functions mentioned above are common mathematical functions in MATLAB, and the cropping process in step S35 can also use the common mathematical function imcrop in MATLAB. The principles and working processes of these common functions will not be elaborated here.

[0030] The first condition is that the area is greater than 1000 pixels and the aspect ratio is greater than 2 and less than 5 (this aspect ratio is the common size in license plate image recognition); the rectangular structure is a 30-by-30 dimension matrix.

[0031] Step S4 involves segmenting the license plate location area to obtain segmented characters. Specifically, this includes: Step S41: Binarizing the license plate location area image; Step S42: Eroding the binarized license plate location area image using a linear structuring element, and then dilating the eroded image to obtain a dilated image; Step S43: Vertically projecting the dilated image to obtain a vertically projected image; Step S44: Setting a dynamic threshold to determine the start and end positions of the characters, and segmenting the vertically projected image based on the dynamic threshold.

[0032] The dynamic threshold in step S43 is the average pixel value of the vertically projected image multiplied by a first adjustment coefficient, which is 0.5. It can be seen that the dynamic threshold setting is proportional to the pixel value of the vertically projected image. Setting a dynamic threshold can fine-tune the starting position of character segmentation based on the pixel quality of the vertically projected image, thus reducing missegmentation (incomplete characters after segmentation or characters containing parts of adjacent characters) and improving the accuracy of license plate image recognition.

[0033] Instruction manual attached Figure 2 Simulation images show license plate recognition using the improved image analysis method involved in this invention. Figure 2 (a) is the original license plate image, which is the enlarged image of the license plate obtained by the image acquisition unit (the image clarity is generally poor due to pixel limitations). Figure 2 (b) The results of license plate recognition using the improved image analysis method of this invention show that the improved image analysis method of this invention effectively identifies the main numbers and letters in the original license plate image, and the MATLAB program recognizes them well. Figure 2 The processing time for the license plate information is 0.8853 seconds. It is evident that, compared to existing technologies, this invention simplifies the image processing workflow, frees up computing resources, and balances the accuracy and timeliness of license plate image recognition. It enables rapid and efficient vehicle license plate recognition, providing a safeguard for the supervision of port violations, particularly for vehicles entering and exiting unregistered loading and unloading areas of ports.

[0034] The embodiments and / or implementation methods described above are merely preferred embodiments and / or implementation methods for implementing the technology of the present invention, and are not intended to limit the implementation methods of the technology of the present invention in any way. Any person skilled in the art can make some modifications or alterations to other equivalent embodiments without departing from the scope of the technical means disclosed in the content of the present invention, but they should still be regarded as the technology or embodiments that are substantially the same as the present invention.

[0035] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. The above descriptions are only preferred embodiments of this application. It should be noted that due to the limitations of written expression, while there are objectively infinite specific structures, those skilled in the art can make several improvements, modifications, or changes without departing from the principles of this application, and can also combine the above technical features in an appropriate manner. These improvements, modifications, changes, or combinations, or the direct application of the inventive concept and technical solution to other situations without modification, should all be considered within the scope of protection of this application.

Claims

1. An intelligent identification and analysis system for port violations based on artificial intelligence, comprising a port image acquisition unit, a port image analysis unit, a data matching unit, and an early warning unit, wherein the port image acquisition unit is connected to the port image analysis unit, the port image analysis unit is connected to the data matching unit, and the data matching unit is connected to the early warning unit, characterized in that, The port image analysis unit performs license plate recognition of vehicles in port images based on an improved image analysis method, sends the license plate recognition result to the data matching unit, and matches it with the preset license plates in the data matching unit to determine whether the license plate belongs to a registered license plate. When it does not belong to a registered license plate, the data matching unit sends a warning message to the warning unit, and the warning unit issues a warning based on the warning message. The improved image analysis method includes the following steps: Step S1: Read the port image sent by the port image acquisition unit; Step S2: Perform data preprocessing on the port image to obtain a preprocessed image; Step S3: Locate the license plate based on the preprocessed image and obtain a license plate location area image; Step S4: Perform character segmentation on the license plate location area image to obtain segmented characters, specifically: Step S41: Binarize the license plate location area image; Step S42: Erode the binarized license plate location area image using a linear structuring element, and dilate the eroded image to obtain a dilated image; Step S43: Perform vertical projection on the dilated image to obtain a vertically projected image; Step S44: Set a dynamic threshold to determine the start and end positions of the characters and perform character segmentation on the vertically projected image based on the dynamic threshold; Step S5: Perform character recognition on the segmented characters to obtain the final license plate recognition result. Step S2 involves data preprocessing of the port image, including step S21: converting the port image into a grayscale image and filtering the grayscale image; the filtering of the grayscale image specifically involves: ； Where I1(x, y) represents the filtered image, x represents the abscissa of the image, y represents the ordinate of the image, F(i, j) represents the filtering window function, I(x+i, y+j) represents the grayscale image before filtering, i represents the abscissa of the filtering window function bias, j represents the ordinate of the filtering window function bias, k represents the width of the filtering window, and σ represents the standard deviation of the filtering window function. The dynamic threshold in step S43 is the average pixel value of the vertically projected image multiplied by the first adjustment coefficient.

2. The intelligent identification and analysis system for port violations based on artificial intelligence according to claim 1, characterized in that, The data preprocessing of the port image in step S2 also includes: Step S22: Perform edge detection on the filtered grayscale image to obtain the preprocessed image.

3. The intelligent identification and analysis system for port violations based on artificial intelligence according to claim 1, characterized in that, Step S3, which involves locating the license plate based on the preprocessed image and obtaining the license plate location region, specifically includes the following sub-steps: Step S31: Perform a closing operation on the preprocessed image using a rectangular structuring element; Step S32: Use the imagefill function to fill the holes in the image; Step S33: Obtain the properties of all connected regions using the regionprop function; Step S34: Traverse all connected regions and select regions that meet the first condition as license plate regions: Step S35: Obtain the bounding box information from the license plate area, and crop out the license plate area based on the bounding box information to complete the license plate localization.

4. The intelligent identification and analysis system for port violations based on artificial intelligence according to claim 3, characterized in that, The first condition is that the area is greater than 1000 pixels and the aspect ratio is greater than 2 and less than 5; the rectangular structure is a 30-by-30 dimension matrix.

5. The intelligent identification and analysis system for port violations based on artificial intelligence according to claim 4, characterized in that, The first adjustment coefficient is 0.

5.

6. The intelligent identification and analysis system for port violations based on artificial intelligence according to claim 1, characterized in that, The warning unit is either a voice warning unit or a flashing light warning unit.

Images