Extraction and rapid screening system for image features of green passing vehicles on expressway

Abstract

The invention relates to the technical field of image analysis, in particular to an expressway green passing vehicle image feature extraction and rapid screening system. The expressway green passing vehicle image feature extraction and rapid screening system comprises an image acquisition unit, an image extraction unit and an image screening unit. According to the method, the image acquisition, extraction and screening processes are automated, the detection efficiency is greatly improved, the vehicle waiting time is greatly shortened, smooth passing of the expressway is guaranteed, the method based on morphological gradient and distance transformation is adopted, the complex scene that cargos are disorderly stacked and mutually shielded in a green passing vehicle container of the expressway can be dealt with, and the traffic safety of the expressway is guaranteed. In the image fusion process, the adjacent sub-regions with similar features are merged based on the feature similarity, redundant information in the image is reduced, in addition, non-green passing standard goods can be screened out, and subjectivity and uncertainty of manual judgment are avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of image analysis, and particularly to a system for extracting image features and rapidly screening green-pass vehicles on expressways. Background Art

[0002] In order to reduce the transportation costs of fresh goods such as agricultural products, improve transportation efficiency, and promote the circulation of agricultural products, the state has introduced the green-pass policy for expressways. Vehicles transporting specific fresh agricultural products are given preferential measures such as toll exemptions to ensure people's livelihood and promote agricultural development.

[0003] Traditional detection methods mainly rely on manual inspection. Staff need to conduct manual inspections on the vehicle cargo compartments at toll stations or specific inspection points, and judge whether the goods meet the green-pass standards by visual observation. This method has many drawbacks. First of all, the efficiency of manual inspection is extremely low. When the traffic flow is large, it is easy to cause vehicle congestion and reduce the passing efficiency of expressways. Secondly, manual inspection has subjectivity and uncertainty. The judgment criteria of different staff may vary, which is prone to misjudgment. It may allow vehicles that do not meet the standards to enjoy green-pass preferences, resulting in toll losses, or it may also misstop vehicles that meet the standards, causing unnecessary losses to vehicle owners.

[0004] In recent years, image processing technology has been more and more widely applied in the transportation field, and remarkable achievements have been made in aspects such as license plate recognition and traffic flow monitoring. However, for the specific application scenario of identifying the goods of green-pass vehicles on expressways, the application of existing image processing technology still faces many challenges. For example, the types and shapes of goods transported by green-pass vehicles on expressways are diverse, and their placement methods in the cargo compartments are disorderly, and there may be situations such as mutual occlusion and stacking. This makes the scanned images of the goods in the cargo compartments highly complex, and traditional image processing algorithms are difficult to accurately extract features and analyze them.

[0005] Based on this, the present invention provides a system for extracting image features and rapidly screening green-pass vehicles on expressways to solve the above-mentioned technical problems. Summary of the Invention

[0006] The object of the present invention is to provide a system for extracting image features and quickly screening green-pass vehicles on highways, which automates the processes of image acquisition, extraction, and screening, greatly improves the detection efficiency, significantly reduces the vehicle waiting time, and ensures the smooth passage of highways. By using the method based on morphological gradient and distance transformation, it can cope with the complex scenarios where the goods in the cargo compartments of green-pass vehicles on highways are stacked messily and mutually blocked. During the image fusion process, adjacent and feature-similar sub-regions are merged based on feature similarity, reducing the redundant information in the images. In addition, it can also screen out the goods that do not meet the green-pass standards, avoiding the subjectivity and uncertainty of manual judgment.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] The present invention provides a system for extracting image features and quickly screening green-pass vehicles on highways, including an image acquisition unit, an image extraction unit, and an image screening unit, where:

[0009] The image acquisition unit is used to scan and image the cargo compartment of a green-pass vehicle on a highway, obtain the scanned image of the goods in the cargo compartment, and collect and obtain a group of scanned images of the same green-pass vehicle on the highway;

[0010] The image extraction unit is used to divide the regions of each scanned image, extract the features of each segmented sub-region, fuse the segmented scanned images based on the extracted features, and extract the features of the fused image, obtain and summarize the features. The image extraction unit is connected to the image acquisition unit;

[0011] The image screening unit obtains the matching recognition results of each scanned image according to the pre-stored goods recognition features, and is used to screen out the goods that do not meet the green-pass standards. The image screening unit is connected to the image extraction unit.

[0012] The present invention is further configured as follows: The image acquisition unit includes an X-ray scanning module, an image collection module, and a first communication module, where:

[0013] The X-ray scanning module uses X-rays to scan and image the cargo compartment to obtain the scanned image of the goods in the cargo compartment;

[0014] The image collection module is used to number the scanned images of the same green-pass vehicle on the highway and collect them into a corresponding group of scanned images. The image collection module is connected to the X-ray scanning module;

[0015] The first communication module is used to realize the information interaction between the image acquisition unit and the image extraction unit. The first communication module is connected to the image collection module.

[0016] A further setting of the present invention is that: the image extraction unit includes a second communication module, a region division module, and a region fusion module, where:

[0017] The second communication module is used to realize information interaction between the image extraction unit and the image acquisition unit and the image screening unit;

[0018] The region division module is used to divide each scanned image into regions, and the region division module is connected to the second communication module;

[0019] The region fusion module extracts features from each sub-region obtained by segmentation according to the segmented scanned image, and fuses the segmented scanned images based on the extracted features. The region fusion module is connected to the region division module.

[0020] A further setting of the present invention is that: the process of dividing each scanned image into regions is as follows:

[0021] Calculate the morphological gradient of the scanned image In the formula, represents the dilation operation on the original grayscale image I, represents the erosion operation on the original grayscale image I;

[0022] Perform distance transformation on the morphological gradient image to obtain the distance from each pixel to the nearest background point In the formula, (x, y) is the pixel coordinate for which the distance transformation value is currently being calculated, (s, t) is the coordinate of the background pixel, and Background is the set of background pixels;

[0023] Use the Otsu method to determine the threshold T to binarize the distance transformation image to obtain the labeled image L;

[0024] For each pixel (x, y) in the labeled image L, check the pixel values in its neighborhood. If the pixel value is less than all the pixel values in its neighborhood, mark it as a local minimum point and assign it a unique label value;

[0025] Take each local minimum point with a different label value in the labeled image L as a seed point and put it into a priority queue. Among them, the priority of each seed point in the queue is determined by the value in the corresponding distance transformation image. The smaller the distance value, the higher the priority;

[0026] Take out the seed point with the highest priority from the priority queue, check the neighborhood pixels of the seed point. If the neighborhood pixels are not labeled, mark them with the same label value as the seed point;

[0027] Repeat the above process. When the expansions in different marked areas meet, obtain a dividing line, and divide the scanned image with the dividing line to obtain a segmented scanned image.

[0028] The present invention is further configured as follows: The process of fusing the segmented scanned images is as follows:

[0029] Obtain the gray-scale and texture features of each sub-region;

[0030] Respectively compare the gray-scale and texture features of each sub-region to obtain a gray-scale similarity and a texture similarity, and compare them with a preset gray-scale similarity threshold and texture similarity threshold. If the gray-scale similarity is greater than the gray-scale similarity threshold and if the texture similarity is greater than the texture similarity threshold, then merge the corresponding sub-regions; otherwise, do not merge, to obtain a fused large-region image.

[0031] The present invention is further configured as follows: The image extraction unit further includes a feature extraction module and a feature summarization module, wherein:

[0032] The feature extraction module is used to extract features from the large-region image, and the feature extraction module is connected to the region fusion module;

[0033] The feature summarization module obtains a first feature set of the corresponding scanned image according to the obtained large-region image features, and summarizes a second feature set of different scanned images. The feature summarization module is connected to both the second communication module and the feature extraction module.

[0034] The present invention is further configured as follows: The process of extracting features from the large-region image is as follows:

[0035] Adopt the Canny edge detection algorithm to extract the contour of the large-region image;

[0036] According to the extracted contour, obtain the area, perimeter, and aspect ratio of the contour. Among them, the calculation formula for the area is where n is the number of points constituting the closed contour, (x i , y i ) is the coordinate of the i-th point on the contour, and the calculation formula for the perimeter is The calculation formula for the aspect ratio is where ω is the width of the circumscribed rectangle of the contour, and h is the height of the circumscribed rectangle of the contour.

[0037] The present invention is further configured as follows: The image screening unit includes a third communication module, a feature matching module, and a database module, wherein:

[0038] The third communication module is used to realize information interaction between the image screening unit and the image extraction unit;

[0039] The feature matching module is used to match the features in the second feature set with the pre-stored goods recognition features to obtain the matching recognition results of each scanned image. The feature matching module is connected to the third communication module;

[0040] The database module is used to store the preset recognition features and store the received data information. The database module is connected to both the third communication module and the feature matching module.

[0041] A further setting of the present invention is that: the image screening unit further includes an image screening module, wherein the image screening module is used to screen out goods that do not meet the green passage standard. The image screening module is connected to both the feature matching module and the database module.

[0042] A further setting of the present invention is that: the process of screening out goods that do not meet the green passage standard is as follows:

[0043] According to the matching recognition results of each scanned image, obtain the type and quantity information of the goods;

[0044] Compare the type and quantity information of the goods obtained from each scanned image. If it is within the preset error threshold, calculate the average type and quantity information. Otherwise, rescan the cargo box;

[0045] When it is within the preset error threshold, it further includes comparing the type and quantity information of the goods with the pre-stored green passage standard to screen out the goods information that does not meet the green passage standard.

[0046] Compared with the prior art, the beneficial effects of the present invention are:

[0047] The present invention automates the processes of image acquisition, extraction, and screening, greatly improving the detection efficiency, significantly reducing the vehicle waiting time, and ensuring the smooth passage of the highway. When dividing regions, the method based on morphological gradient and distance transformation is adopted, which can handle the complex scenarios where the goods in the cargo boxes of green passage vehicles on the highway are randomly stacked and mutually blocked, and accurately distinguish different goods or different parts of the same good, greatly improving the accuracy of region division. In the process of image fusion, adjacent and feature-similar sub-regions are merged based on feature similarity, reducing the redundant information in the image. At the same time, the merged large region contains more pixel information, improving the representativeness and stability of the features. In addition, according to the pre-stored goods recognition features, the summarized features are matched to obtain the matching recognition results of each scanned image, and the goods that do not meet the green passage standard can be screened out by comparing the type and quantity information of the goods with the green passage standard, avoiding the subjectivity and uncertainty of manual judgment and improving the accuracy and fairness of screening. Description of the Drawings

[0048] Figure 1 This is the system diagram of the system for extracting and quickly screening the image features of green-pass vehicles on highways according to the present invention.

[0049] Figure 2 This is the system diagram in the image acquisition unit of the system for extracting and quickly screening the image features of green-pass vehicles on highways according to the present invention.

[0050] Figure 3 This is the system diagram in the image extraction unit of the system for extracting and quickly screening the image features of green-pass vehicles on highways according to the present invention.

[0051] Figure 4 This is the system diagram in the image screening unit of the system for extracting and quickly screening the image features of green-pass vehicles on highways according to the present invention.

[0052] Explanation of the reference numerals in the attached drawings:

[0053] 100, image acquisition unit; 110, X-ray scanning module; 120, image aggregation module; 130, first communication module; 200, image extraction unit; 210, second communication module; 220, region division module; 230, region fusion module; 240, feature extraction module; 250, feature aggregation module; 300, image screening unit; 310, third communication module; 320, feature matching module; 330, database module; 340, image screening module. Specific embodiments

[0054] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0055] Embodiment:

[0056] As Figures 1-4As shown in the figure, this embodiment provides a system for extracting image features and quickly screening green passing vehicles on expressways, including an image acquisition unit 100, an image extraction unit 200, and an image screening unit 300, where: The image acquisition unit 100 is used to scan and image the cargo box of green passing vehicles on expressways, obtain the scanned images of the goods in the cargo box, and collect and obtain a group of scanned images of the same green passing vehicle on the expressway; The image extraction unit 200 is used to divide the regions of each scanned image, extract features from each segmented sub-region, fuse the segmented scanned images based on the extracted features, and extract features from the fused image, obtain and summarize the features. The image extraction unit 200 is connected to the image acquisition unit 100; The image screening unit 300 obtains the matching recognition results of each scanned image according to the pre-stored goods recognition features, and is used to screen out goods that do not meet the green passing standards. The image screening unit 300 is connected to the image extraction unit 200.

[0057] In this embodiment, it should be noted that the image acquisition unit 100 uses X-rays to scan the goods in the cargo box at different angles, and then obtains scanned images at different angles. The multiple scanned images of the same truck are numbered and uploaded to the image extraction unit 200. The image extraction unit 200 receives the group of scanned images from the image acquisition unit 100, divides the regions of each scanned image, that is, performs region division by calculating morphological gradients, distance transforms, Otsu methods, etc., which can effectively distinguish different goods or different parts of the same good, and can accurately find the segmentation line even in complex image scenes, realizing reasonable region division. Then, fuse the segmented scanned images, and extract features from the fused image, obtain and summarize the features. The feature extraction of the sub-regions can deeply mine the information contained in each sub-region. Among them, the gray feature reflects the brightness distribution of the sub-region, and the texture feature reflects the surface texture information of the sub-region, which can effectively improve the effect of distinguishing different types. Through image fusion, similar sub-regions can be merged into a large region, reducing the redundant information in the image, making the image more concise and facilitating subsequent analysis and processing. By extracting features from the fused image, features can be better captured, improving the accuracy and reliability of goods recognition. The image extraction unit 200 will upload the processed image information to the image screening unit 300. The image screening unit 300 obtains the type and quantity information of the goods according to the matching recognition results of each scanned image, compares the type and quantity information of the goods obtained from each scanned image, and then compares this information with the pre-stored green passing standards. If it is found that there are non-agricultural products that do not meet the green passing standards in a certain batch of goods, such as some industrial products are carried, then the information of these goods that do not meet the green passing standards is screened out.

[0058] In the present invention, the image acquisition unit 100 includes an X-ray scanning module 110, an image aggregation module 120, and a first communication module 130, where: The X-ray scanning module 110 uses X-rays to scan and image a cargo box to obtain a scanned image of the goods inside the cargo box; The image aggregation module 120 is used to number the scanned images of the same green-pass vehicles on the highway and aggregate them into corresponding scanned image groups. The image aggregation module 120 is connected to the X-ray scanning module 110; The first communication module 130 is used to realize the information interaction between the image acquisition unit 100 and the image extraction unit 200. The first communication module 130 is connected to the image aggregation module 120.

[0059] In this embodiment, it should be noted that the X-ray scanning module 110 uses X-rays to scan and image a freight car cargo box. For example, the cargo box is loaded with fruits such as apples and bananas, as well as some vegetables. The X-rays penetrate the cargo box and the goods to form a scanned image of the goods, and the scanned image is transmitted to the image aggregation module 120. The image aggregation module 120 numbers multiple scanned images of the same freight car, such as S001, S002, etc., and aggregates them into corresponding scanned image groups. As an example, assuming that 5 scanned images at different angles are obtained in this scan, the image aggregation module 120 organizes them together to form a corresponding scanned image group, and the obtained scanned image group is uploaded to the image extraction unit 200 through the first communication module 130.

[0060] In the present invention, the image extraction unit 200 includes a second communication module 210, a region division module 220, and a region fusion module 230, where: The second communication module 210 is used to realize the information interaction between the image extraction unit 200 and the image acquisition unit 100 and the image screening unit 300; The region division module 220 is used to divide each scanned image. The region division module 220 is connected to the second communication module 210; The region fusion module 230 extracts features from each sub-region obtained by segmentation according to the segmented scanned images, and fuses the segmented scanned images based on the extracted features. The region fusion module 230 is connected to the region division module 220.

[0061] Among them, the process of dividing each scanned image is as follows:

[0062] Calculate the morphological gradient of the scanned image In the formula, represents the dilation operation on the original grayscale image I, represents the erosion operation on the original grayscale image I.

[0063] Perform distance transformation on the morphological gradient image to obtain the distance from each pixel to the nearest background point Wherein, (x, y) are the pixel coordinates for which the distance transformation value is currently being calculated, (s, t) are the coordinates of the background pixels, and Background is the set of background pixels.

[0064] The Otsu method is used to determine the threshold T to binarize the distance transformation image, obtaining the labeled image L.

[0065] For each pixel (x, y) in the labeled image L, check the pixel values in its neighborhood. If the pixel value is less than all the pixel values in its neighborhood, then mark it as a local minimum point and assign a unique label value.

[0066] Take each local minimum point with a different label value in the labeled image L as a seed point and put it into a priority queue. Among them, the priority of each seed point in the queue is determined by the value in the corresponding distance transformation image. The smaller the distance value, the higher the priority.

[0067] Take out the seed point with the highest priority from the priority queue, check the neighborhood pixels of the seed point. If the neighborhood pixels are not labeled, then label them with the same label value as the seed point.

[0068] Repeat the above process. When the expansions of different labeled regions meet, obtain the dividing line, and divide the scanned image with the dividing line to obtain the segmented scanned image.

[0069] In addition, the process of fusing the segmented scanned image is as follows:

[0070] Obtain the gray-scale and texture features of each sub-region.

[0071] Respectively compare the gray-scale and texture features of each sub-region to obtain the gray-scale similarity and texture similarity, and compare them with the preset gray-scale similarity threshold and texture similarity threshold. If the gray-scale similarity is greater than the gray-scale similarity threshold, and if the texture similarity is greater than the texture similarity threshold, then merge the corresponding sub-regions, otherwise, do not merge, to obtain the fused large-region image.

[0072] In this embodiment, it should be noted that the information of the image acquisition unit 100 is received through the second communication module 210 and uploaded to the region division module 220. The region division module 220 divides each scanned image into regions and transmits the divided scanned image to the region fusion module 230. The region fusion module 230 extracts the features of each segmented sub-region and, based on the extracted features, fuses the segmented scanned images. As an example, assume that the gray-scale similarity threshold is 0.8 and the texture similarity threshold is 0.7. If the gray-scale similarity between two sub-regions is greater than 0.8 and the texture similarity is greater than 0.7, then merge them, otherwise do not merge, to obtain the fused large-region image.

[0073] In the present invention, the image extraction unit 200 further includes a feature extraction module 240 and a feature summarization module 250, where: the feature extraction module 240 is used to extract features from the large-area image, and the feature extraction module 240 is connected to the region fusion module 230; the feature summarization module 250 obtains a first feature set of the corresponding scanned image according to the obtained large-area image features, and summarizes a second feature set of different scanned images. The feature summarization module 250 is connected to both the second communication module 210 and the feature extraction module 240.

[0074] Among them, the process of extracting features from the large-area image is as follows:

[0075] The Canny edge detection algorithm is used to extract the contour of the large-area image.

[0076] According to the extracted contour, the area, perimeter, and aspect ratio of the contour are obtained. Among them, the calculation formula for the area is In the formula, n is the number of points forming the closed contour, (x i , y i ) is the coordinate of the i-th point on the contour. The calculation formula for the perimeter is The calculation formula for the aspect ratio is In the formula, ω is the width of the circumscribed rectangle of the contour, and h is the height of the circumscribed rectangle of the contour.

[0077] In this embodiment, it should be noted that the feature extraction module 240 receives the information of the region fusion module 230, obtains the feature information of the large-area image, and transmits it to the feature summarization module 250. After being summarized by the feature summarization module 250, it is uploaded to the image screening unit 300 through the second communication module 210. As an example, assume that the contour of a watermelon is composed of 100 points, and the coordinates of these points are (x 1 , y 1 ), (x 2 , y 2 ), …, (x 100 , y 100 ). According to the area calculation formula Substitute the coordinate values for calculation to obtain the area of the watermelon contour; for the above watermelon contour, according to the perimeter calculation formula, calculate the distance between adjacent points in turn and sum them to obtain the perimeter of the watermelon contour; measure the width 30 and height 20 of the circumscribed rectangle of the watermelon contour, and according to the aspect ratio calculation formula, calculate the aspect ratio of the watermelon contour to be 1.5.

[0078] In the present invention, the image screening unit 300 includes a third communication module 310, a feature matching module 320, and a database module 330, where: the third communication module 310 is used to implement information interaction between the image screening unit 300 and the image extraction unit 200; the feature matching module 320 is used to match the features in the second feature set with the pre-stored goods recognition features to obtain the matching recognition results of each scanned image, and the feature matching module 320 is connected to the third communication module 310; the database module 330 is used to store the preset recognition features and store the received data information, and the database module 330 is connected to both the third communication module 310 and the feature matching module 320.

[0079] In addition, the image screening unit 300 further includes an image screening module 340, where the image screening module 340 is used to screen out goods that do not meet the green passage standard, and the image screening module 340 is connected to both the feature matching module 320 and the database module 330.

[0080] Among them, the process of screening out goods that do not meet the green passage standard is as follows:

[0081] According to the matching recognition results of each scanned image, obtain the type and quantity information of the goods.

[0082] Compare the type and quantity information of the goods obtained from each scanned image. If it is within the preset error threshold, calculate the average type and quantity information; otherwise, rescan the cargo box.

[0083] When it is within the preset error threshold, it further includes comparing the type and quantity information of the goods with the pre-stored green passage standard to screen out the goods information that does not meet the green passage standard.

[0084] In this embodiment, it should be noted that the information of the image extraction unit 200 is received through the provided third communication module 310 and transmitted to the feature matching module 320 to obtain the matching recognition results of each scanned image. Then, the image screening module 340 screens out the goods that do not meet the green passage standard. As an example, compare the type and quantity information of the goods obtained from each scanned image. Assume that the preset error threshold is 5%. If the difference in the quantity and type information of the goods in each scanned image is within 5%, calculate the average type and quantity information, and then compare this information with the pre-stored green passage standard. If it is found that there are non-agricultural products that do not meet the green passage standard in a certain batch of goods, such as some industrial products being carried, then screen out the goods information that does not meet the green passage standard.

[0085] In the description of this specification, the descriptions referring to the terms "one embodiment", "example", "specific example", etc. mean that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0086] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific embodiments described. Obviously, many modifications and variations can be made according to the content of this specification. These embodiments are selected and specifically described in this specification in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can well understand and utilize the present invention. The present invention is only limited by the claims and their full scope and equivalents.

Claims

1. The highway green pass vehicle image feature extraction and rapid screening system is characterized by: The invention comprises an image acquisition unit (100), an image extraction unit (200) and an image screening unit (300), wherein: The image acquisition unit (100) is used to scan and image the cargo box of a highway green pass vehicle, obtain a scanned image of the cargo in the cargo box, and collect and obtain a scanned image group of the same highway green pass vehicle; The image extraction unit (200) is used to divide each scanned image into regions, extract features from each segmented sub-region, fuse the segmented scanned images based on the extracted features, and extract features from the fused image to obtain and summarize features. The image extraction unit (200) is connected to the image acquisition unit (100); The image screening unit (300) obtains matching identification results of each scanned image according to pre-stored cargo identification features, and is used to screen out cargo that does not meet green pass standards. The image screening unit (300) is connected to the image extraction unit (200).

2. The highway green pass vehicle image feature extraction and rapid screening system according to claim 1 is characterized in that: The image acquisition unit (100) comprises an X-ray scanning module (110), an image collection module (120) and a first communication module (130), wherein: The X-ray scanning module (110) uses X-rays to scan and image the cargo box to obtain a scanned image of the cargo in the cargo box; The image collection module (120) is used to number the scanned images of green-pass vehicles on the same expressway and collect them into corresponding scanned image groups. The image collection module (120) is connected to the X-ray scanning module (110); The first communication module (130) is used to realize information interaction between the image acquisition unit (100) and the image extraction unit (200), and the first communication module (130) is connected to the image collection module (120).

3. The highway green pass vehicle image feature extraction and rapid screening system according to claim 1 is characterized in that: The image extraction unit (200) comprises a second communication module (210), a region division module (220) and a region fusion module (230), wherein: The second communication module (210) is used to realize information interaction between the image extraction unit (200) and the image acquisition unit (100) and the image screening unit (300); The region division module (220) is used to divide each scanned image into regions, and the region division module (220) is connected to the second communication module (210); The region fusion module (230) extracts features from each segmented sub-region according to the segmented scanned image, and fuses the segmented scanned images based on the extracted features. The region fusion module (230) is connected to the region division module (220).

4. The highway green pass vehicle image feature extraction and rapid screening system according to claim 3 is characterized in that: The process of dividing each scanned image into regions is as follows: Compute morphological gradients of scanned images In the formula, Indicates the expansion operation of the original grayscale image I. Indicates that the original grayscale image I is corroded; Perform distance transformation on the morphological gradient image to obtain the distance from each pixel to the nearest background point Where (x, y) is the pixel coordinates of the current distance transform value being calculated, (s, t) is the coordinates of the background pixel, and Background is the background pixel set; The Otsu method is used to determine the threshold T to binarize the distance transform image and obtain the marked image L; For each pixel (x, y) in the labeled image L, check the pixel values ​​in its neighborhood. If the pixel value is less than all the pixel values ​​in its neighborhood, mark it as a local minimum point and assign a unique label value. Each local minimum point with different marking values ​​in the marked image L is taken as a seed point and put into a priority queue, where the priority of each seed point in the queue is determined by the value in its corresponding distance transformation image. The smaller the distance value, the higher the priority; Take the seed point with the highest priority from the priority queue, check the neighboring pixels of the seed point, and if the neighboring pixels are not marked, mark them with the same mark value as the seed point; The above process is repeated, and when the expansions of different marked areas meet, a dividing line is obtained, and the scanned image is divided by the dividing line to obtain a segmented scanned image.

5. The highway green pass vehicle image feature extraction and rapid screening system according to claim 3 is characterized in that: The process of fusing the segmented scanned images is as follows: Obtain the grayscale and texture features of each sub-region; The grayscale and texture features of each sub-region are compared for similarity respectively to obtain grayscale similarity and texture similarity, and compared with the preset grayscale similarity threshold and texture similarity threshold. If the grayscale similarity is greater than the grayscale similarity threshold, and if the texture similarity is greater than the texture similarity threshold, the corresponding sub-regions are merged, otherwise, they are not merged to obtain a fused large-region image.

6. The highway green pass vehicle image feature extraction and rapid screening system according to claim 5 is characterized in that: The image extraction unit (200) further comprises a feature extraction module (240) and a feature aggregation module (250), wherein: The feature extraction module (240) is used to extract features from a large-area image, and the feature extraction module (240) is connected to the area fusion module (230); The feature aggregation module (250) acquires a first feature set corresponding to the scanned image and a second feature set for summarizing different scanned images based on the acquired large-area image features. The feature aggregation module (250) is connected to both the second communication module (210) and the feature extraction module (240).

7. The highway green pass vehicle image feature extraction and rapid screening system according to claim 6 is characterized in that: The process of extracting features from large area images is as follows: The Canny edge detection algorithm is used to extract the contours of large area images; According to the extracted contour, the area, perimeter and aspect ratio of the contour are obtained, where the area calculation formula is: Where n is the number of points that form a closed contour, (x i ,y i ) is the coordinate of the i-th point on the contour, and the calculation formula for the perimeter is The aspect ratio is calculated as Where ω is the width of the outline's circumscribed rectangle, and h is the height of the outline's circumscribed rectangle.

8. The highway green pass vehicle image feature extraction and rapid screening system according to claim 6 is characterized in that: The image screening unit (300) comprises a third communication module (310), a feature matching module (320) and a database module (330), wherein: The third communication module (310) is used to realize information interaction between the image screening unit (300) and the image extraction unit (200); The feature matching module (320) is used to match the features in the second feature set with the pre-stored cargo identification features to obtain matching identification results of each scanned image, and the feature matching module (320) is connected to the third communication module (310); The database module (330) is used to store preset identification features and store received data information, and the database module (330) is connected to both the third communication module (310) and the feature matching module (320).

9. The highway green pass vehicle image feature extraction and rapid screening system according to claim 8 is characterized in that: The image screening unit (300) further comprises an image screening module (340), wherein the image screening module (340) is used to screen out goods that do not meet the green pass standard, and the image screening module (340) is connected to both the feature matching module (320) and the database module (330).

10. The highway green pass vehicle image feature extraction and rapid screening system according to claim 9 is characterized in that: The process of screening out non-green pass standard goods is as follows: According to the matching and recognition results of each scanned image, the type and quantity information of the goods are obtained; Compare the type and quantity information of the goods obtained from each scanned image. If they are within the preset error threshold, calculate the average type and quantity information. Otherwise, rescan the cargo box. When the error is within the preset threshold, it also includes comparing the type and quantity of the goods with the pre-stored green pass standards to screen out goods that do not meet the green pass standards.

Images