Image recognition security processing method and system for security screen

Abstract

The invention discloses an image recognition security and protection processing method and system for a security and protection screen, relates to the technical field of security and protection screens, solves the problem that original object contour feature processing is rough, and can effectively lock a dynamic object by analyzing pixel point differences between adjacent frame images acquired by a high-definition camera. When the dynamic features are confirmed, difference regions and contour regions are accurately divided through point location analysis and gradient calculation of the system, and the feature region with the largest area is selected from a plurality of feature regions to serve as the dynamic features, so that the recognition accuracy is ensured, and the misjudgment rate is reduced; in each contour partition, feature difference values of different internal connecting lines are compared, and the connecting line with the minimum difference value is selected as an average refining connecting line, so that refining processing of the dynamic feature edge contour is realized, the extracted feature contour is more accurate, and the real form of an object can be better reflected.

Description

Technical Field

[0001] The present invention relates to the technical field of security screen nets, and specifically to an image recognition security processing method and system for security screen nets. Background Art

[0002] In today's society, the importance of security prevention has become increasingly prominent. As a key facility for ensuring the security of specific areas, security screen nets are widely used in scenarios with extremely high security requirements such as military bases, important industrial sites, and high-end residential communities;

[0003] Traditional security systems process object features rather coarsely and cannot perform refined processing on object contours, resulting in inaccurate extracted features and difficulty in meeting scenarios with extremely high security requirements; when faced with objects with complex shapes, the features extracted by traditional methods are difficult to reflect the true form of the objects, causing difficulties in subsequent analysis and judgment; in this application, dynamic features are combined with a two-dimensional coordinate system. The center point is determined by calculating the mean coordinates, the contour is divided into sections based on the length of the feature connection lines, and an average quantity is selected within each section to refine the connection lines, thereby refining the edge contours of the dynamic features, being able to obtain more accurate object features, and performing refined processing on the edge contours of the object features, making the overall contour of the object features more prominent and achieving better recognition accuracy. Summary of the Invention

[0004] In view of the deficiencies of the prior art, the present invention provides an image recognition security processing method and system for security screen nets, which solves the problem of relatively coarse processing of the original object contour features.

[0005] To achieve the above objectives, the present invention is realized through the following technical solutions: An image recognition security processing method for security screen nets includes the following steps:

[0006] Step 1: Based on the laser detectors installed in the security screen net, identify whether an item has passed by, and based on the recognition result, confirm the image recognition signal. The specific method is as follows:

[0007] Monitor the detection signals associated with the laser detectors in real time. When the signal value associated with the detection signal does not belong to the preset interval, it indicates that an item has passed by, and an image recognition signal is directly generated. The endpoint values of the preset interval are both preset values;

[0008] Step 2: Based on the generated image recognition signal, control a high-definition camera to obtain the images of the area where the security screen net is located, and lock the dynamic features existing in multiple frames of images by analyzing the pixel differences between adjacent frames of images. The specific method is as follows:

[0009] Sort the images obtained from different frames according to the time sequence to confirm the image set;

[0010] Randomly select a point in a single image and mark it as the zero point, and confirm the position characteristics of other points based on the position differences between other points and the zero point. And so on, confirm the position characteristics of several points associated with different frames of images in the image set;

[0011] And calibrate the pixel values associated with different frames of images in the image set as X i-k , where i represents different frames of images, and k represents the points associated with different position characteristics. Confirm the pixel difference of the points with the same position characteristics in adjacent frames of images, and use XC j-k =|X j-k -X (j+1)-k | to confirm the pixel difference XC j-k , where j ∈ i, assuming i = 1, 2,..., n, then 1 ≤ j ≤ n - 1;

[0012] Mark the points that satisfy: XC j-k ≥ Y1 as the differential points, and mark the area covered by several differential points in the corresponding frame of image as the differential area. Then, perform contour calibration on the outer area of the differential area: Based on the different pixel values associated with different pixel points in the frame of image, use the Sobel algorithm to calibrate the horizontal gradient H k and the vertical gradient S k of each pixel point, and use: to confirm the comprehensive gradient ZH k associated with the corresponding pixel point. Mark the pixel points that satisfy ZH k > Y2 as the gradient pixel points, and mark the internal area associated with several consecutive gradient pixel points as the contour area. Otherwise, do not perform any calibration. Both Y1 and Y2 are preset values;

[0013] Based on the confirmed contour area and differential area in the corresponding frame of image, confirm the contour area to which the differential area belongs, and record this contour area as the characteristic area of this frame of image;

[0014] Based on the different characteristic areas determined in different frames of images, select the characteristic area with the largest area and mark it as the dynamic characteristic of this image set;

[0015] Step 3: Based on the dynamic characteristics associated with the corresponding image set, first confirm the contour partitions from the edge contours associated with the dynamic characteristics, then confirm the average refinement connections from different contour partitions, and according to the different average refinement connections associated with different contour partitions, perform refinement processing on the edge contours of this dynamic characteristic. The specific method is as follows:

[0016] Combine the calibrated dynamic features with a two-dimensional coordinate system to identify the different two-dimensional coordinates associated with different pixel points within the dynamic features. Then, perform a mean processing on several groups of two-dimensional coordinates to determine the mean coordinates, and mark the determined mean coordinates within the dynamic features, which is denoted as the center point of this dynamic feature;

[0017] Identify the edge points associated with the outermost edge of the dynamic feature, connect the center point with different edge points to lock the point connections. Then, based on several groups of gradient pixel points identified outside the dynamic feature, determine the gradient regions covered by these several groups of gradient pixel points, and further identify the part of the point connections located within the gradient regions, which is denoted as the feature connection corresponding to the gradient region, and determine the connection length associated with each feature connection;

[0018] Perform feature identification on adjacent feature connections to lock the clustering feature connection set: randomly select a group of feature connections and use the clockwise direction as the processing direction. Adopt: Feature difference = |Connection length of the previous group of feature connections - Connection length of the next group of feature connections|. If the feature difference ≤ Y3, where Y3 is a preset value, then mark the adjacent two groups of feature connections as the clustering feature connection set; otherwise, mark the corresponding feature connections as dividing lines;

[0019] Based on the positions of different dividing lines within the gradient region, divide the corresponding gradient region into multiple different contour partitions;

[0020] The specific method for calibrating different average refinement connections associated with different contour partitions is as follows:

[0021] Randomly select a group of pixel points from the pixel points associated with each different vertical row within the contour partition, which is denoted as the selected points. Then, perform a connection process on several selected points to determine a group of internal connections. Sum the pixel values corresponding to the multiple groups of pixel points associated with the upper part of the internal connection to determine the upper feature sum Z1. Then, sum the pixel values corresponding to the multiple groups of pixel points associated with the lower part of the internal connection to determine the lower feature sum Z2. Adopt: CZ = |Z1 - Z2| to determine the feature difference CZ;

[0022] Execute multiple different point selection processes. The internal connections associated with each selection process are different, and determine the different feature differences CZ associated with different internal connections. From the multiple different feature differences CZ, select the minimum value, and denote the internal connection associated with the minimum value as the average refinement connection of this contour partition. Connect the average refinement connections associated with each contour partition as the edge contour line after the refinement processing of this dynamic feature, and use this edge contour line as the feature contour of this dynamic feature;

[0023] Step 4: Perform early warning verification on the dynamic features after the contour refinement process. Compare this dynamic feature with the pre-stored feature contours in the model library to confirm whether there is a comparison result. If there is, perform early warning processing; if not, no processing is required. The specific method is as follows:

[0024] Compare the dynamic feature with the pre-stored feature contours in the model library one by one: Make the center point of the dynamic feature coincide with the center point of the feature contour, and control the rotation of the dynamic feature to identify whether the coincidence rate between its dynamic feature and the feature contour reaches 90%. If it reaches, it means there is a comparison result, and directly control the alarm to perform early warning processing to alert external operators; if it does not reach, no processing is performed.

[0025] Preferably, an image recognition security processing system for a security curtain net includes:

[0026] A feature confirmation end, which identifies whether there are items passing through based on the laser detectors set inside the security curtain net, and confirms the image recognition signal based on the recognition result;

[0027] A dynamic feature locking end, which, based on the generated image recognition signal, controls a high-definition camera to obtain the images of the area where the security curtain net is located, and locks the dynamic features existing in multiple frames of images by analyzing the pixel point differences between adjacent frames of images;

[0028] A contour refinement processing end, which, based on the dynamic features associated with the corresponding image set, first confirms the contour partitions from the edge contours associated with the dynamic features, then confirms the average refinement connections from different contour partitions, and refines the edge contour of this dynamic feature according to the different average refinement connections associated with different contour partitions;

[0029] An early warning processing end, which performs early warning verification on the dynamic features after the contour refinement process, compares this dynamic feature with the pre-stored feature contours in the model library to confirm whether there is a comparison result. If there is, perform early warning processing; if not, no processing is required.

[0030] The present invention provides an image recognition security processing method and system for a security curtain net. Compared with the prior art, it has the following beneficial effects:

[0031] By analyzing the pixel point differences between adjacent frames of images obtained by the high-definition camera, the present invention can effectively lock dynamic objects; when confirming dynamic features, through the point position analysis and gradient calculation of the system, the difference area and contour area are accurately divided, and the one with the largest area is selected from multiple feature areas as the dynamic feature to ensure the accuracy of recognition and reduce the false judgment rate;

[0032] Within each contour partition, by comparing the characteristic differences of different internal connections, the connection with the smallest difference is selected as the average quantity refinement connection, so as to achieve the refined processing of the dynamic feature edge contour, making the extracted feature contour more accurate and better reflecting the true shape of the object;

[0033] Compare the processed dynamic features with the pre-stored feature contours in the model library, and quickly and accurately identify possible security threats through the judgment of the center point coincidence and the contour coincidence rate. Brief Description of the Drawings

[0034] Figure 1 It is a schematic flow chart of the method of the present invention;

[0035] Figure 2 It is a schematic diagram of the principle framework of the present invention. Detailed Embodiment

[0036] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in 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. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0037] First Embodiment

[0038] Please refer to Figure 1 , this application provides an image recognition security processing method for a security curtain net, including the following steps:

[0039] Step 1: Based on the laser detectors set within the security curtain net, identify whether there are items passing by, and based on the recognition result, confirm the image recognition signal. Specifically, the laser detector includes a laser head assembly, a battery, and corresponding legs. The laser head assembly can emit laser to the relevant monitored area. The scanning angle range of the laser head assembly is 0-90°, the scanning length is 30m, and the height is 20m, forming a laser curtain wall in the front. When there is an object passing through this laser curtain wall, the laser signal detected by the laser detector changes, so it represents that there are items passing by, and directly generates an image recognition signal for image processing. Among them, the relevant operation method is: monitor the detection signal associated with the laser detector in real time. When the signal value associated with the detection signal does not belong to the preset interval (that is, the corresponding laser signal has intensity changes or numerical fluctuations), it represents that there are items passing by, and directly generates an image recognition signal. The end point values of the preset interval are all preset values, which are determined in advance by relevant personnel according to experience;

[0040] Step 2: Based on the generated image recognition signals, control the high-definition camera to obtain the images of the security screen network area, and lock the dynamic features existing in multiple frames of images by analyzing the pixel differences between adjacent frames of images. Specifically, the camera and the corresponding laser detector are placed at the same location, so the camera can effectively obtain the specific images in front. During the acquisition process, the images associated with each frame are different. Therefore, the corresponding dynamic objects can be confirmed based on the images of the corresponding adjacent frames to determine the corresponding dynamic features. The specific method for locking dynamic features is as follows:

[0041] Sort the images obtained from different frames according to the time sequence to confirm the image set;

[0042] Randomly select a point in a single image and mark it as the zero point, and confirm the position features of other points based on the position differences between other points and the zero point. By analogy, confirm the position features of several points associated with different frames of images in the image set;

[0043] And calibrate the pixel values of different frames of images in the image set as X i-k , where i represents different frames of images, and k represents the points associated with different position features. Confirm the pixel difference of the points with the same position features in adjacent frames of images, and use XC j-k =|X j-k -X (j+1)-k | to confirm the pixel difference XC j-k , where j ∈ i, assuming i = 1, 2, ……, n, then 1 ≤ j ≤ n - 1. When calibrating the subsequent differential points, they can be directly calibrated in the images of the adjacent front and rear frames. Then the images that exist in both the front and the back are the corresponding intermediate images. Based on the position features of the points associated with the intermediate image and the front and rear images, the corresponding differential points can be determined;

[0044] Calibrate the points that satisfy: XC j-k ≥ Y1 as differential points, and calibrate the area covered by several differential points in the corresponding frame of image as the differential area. Then calibrate the contour of the outer area of the differential area: Based on the different pixel values associated with different pixels in the frame of image, use the Sobel algorithm to calibrate the horizontal gradient H k and the vertical gradient S k of each pixel (since the method of confirming gradient data through algorithms is relatively common in the prior art, so it will not be elaborated here too much. By confirming the grid and related weights associated with the corresponding pixel, the comprehensive gradient of the corresponding pixel is determined), and use: to confirm the comprehensive gradient ZH k associated with the corresponding pixel. Calibrate the points that satisfy ZH kPixels greater than Y2 are designated as gradient pixels. The internal area associated with several consecutive gradient pixels is denoted as the contour area (for the confirmation of its gradient pixels, if all the pixels associated with the periphery of an object feature are gradient pixels, then the intermediate area included in such gradient pixels is the corresponding contour area). Otherwise, no calibration is performed, and both Y1 and Y2 are preset values, and their specific values are determined by the operator according to experience;

[0045] Based on the confirmed contour area and difference area within the corresponding frame image, determine the contour area to which the difference area belongs, and record this contour area as the feature area of this frame image. The difference area is the pixel position difference area generated by the corresponding dynamic object during movement. This difference area is generally generated by the change of the corresponding pixels when the object moves, so the corresponding difference area can be confirmed, and there is a specific contour area to which the difference area belongs;

[0046] Based on the different feature areas determined within different frame images, select the feature area with the largest area and denote it as the dynamic feature of this image set (different feature areas have different area values. From multiple area values, determine the feature area associated with the largest area value, and denote the feature area with the largest area value as the dynamic feature of the corresponding image set);

[0047] Step 3: Based on the dynamic feature associated with the corresponding image set, first confirm the contour partitions from the edge contour associated with the dynamic feature, then confirm the average refinement connections from different contour partitions, and based on the different average refinement connections associated with different contour partitions, perform refinement processing on the edge contour of this dynamic feature. The specific refinement processing steps are as follows:

[0048] 1. Confirm the contour partitions from the edge contour associated with the dynamic feature:

[0049] Combine the calibrated dynamic feature with the two-dimensional coordinate system to confirm the different two-dimensional coordinates associated with different pixels within the dynamic feature, then perform mean processing on several groups of two-dimensional coordinates to confirm the mean coordinate, and mark the confirmed mean coordinate within the dynamic feature, denoted as the center point of this dynamic feature;

[0050] Confirm the edge points associated with the outermost edge of the dynamic feature, connect the center point with different edge points to lock the point connections. Then, based on several groups of gradient pixels confirmed outside the dynamic feature, confirm the gradient area covered by several groups of gradient pixels, and then confirm the part of the point connections located within the gradient area, denoted as the feature connection of the corresponding gradient area, and determine the connection length associated with each feature connection;

[0051] Perform feature confirmation on adjacent feature connection lines and lock the set of clustered feature connection lines: Randomly select a set of feature connection lines and take the clockwise direction as the processing direction. Use: Feature difference = |Length of the previous set of feature connection lines - Length of the next set of feature connection lines|. If the feature difference ≤ Y3, then label the adjacent two sets of feature connection lines as the set of clustered feature connection lines (if the subsequent changes in connection line length still fall within the corresponding numerical range, then they are still labeled into the same set of clustered feature connection lines). Y3 is a preset value, and its specific value is determined by the operator according to experience. Otherwise, label the corresponding feature connection line as a dividing line;

[0052] Based on the positions of different dividing lines within the gradient region, divide the corresponding gradient region into multiple different contour partitions (that is, the region associated between adjacent dividing lines, and the corresponding region is the divided contour partition). Within the overall edge contour of the corresponding dynamic feature, there are different determined feature connection lines. Then, perform numerical confirmation on the lengths of different feature connection lines. Based on the confirmed specific numerical values, conduct feature division. From the specific length values obtained from the division, lock the corresponding dividing lines. Based on the corresponding dividing lines, perform region division, thereby locking different contour partitions, which is convenient for subsequent specific confirmation of the average refinement connection lines;

[0053] II. The specific method for calibrating different average refinement connection lines associated with different contour partitions is as follows:

[0054] When sorting its pixel points, it is sorted in a grid form. Therefore, there are multiple groups of related pixel points in each vertical direction. Randomly select a group of pixel points from the pixel points associated with each different vertical row within the contour partition and record them as the selected point positions. Then, perform connection processing on several selected point positions to confirm a set of internal connection lines. Sum the pixel values of the multiple groups of pixel points associated above the internal connection line to confirm the upper feature sum Z1. Then, sum the pixel values of the multiple groups of pixel points associated below the internal connection line to confirm the lower feature sum Z2. Use: CZ = |Z1 - Z2| to confirm the feature difference CZ;

[0055] Execute multiple different point selection processes. The internal connection lines associated with each selection process are all different, and confirm the different feature differences CZ associated with different internal connection lines. From the multiple different feature differences CZ, select the minimum value. Denote the internal connection line associated with the minimum value as the average refinement connection line of this contour partition. Connect the average refinement connection lines associated within each contour partition as the edge contour line after the refinement processing of this dynamic feature, and use this edge contour line as the feature contour of this dynamic feature;

[0056] Specifically, the corresponding overall edge contour is refined into multiple different contour partitions. Then, different feature connection lines associated with different contour partitions are numerically analyzed and confirmed. The specific connection lines with relatively consistent parameter features can be locked and used as a set of connection lines. Then, the dividing lines that do not belong to the corresponding connection line set are the best dividing features. The relevant confirmation of the edge contour partition is carried out, and specific adjustments are made to achieve a better contour confirmation processing effect. Subsequently, the connection lines within the contour are refined to confirm the specific feature contour of the corresponding dynamic features, so as to achieve a better feature refinement processing process.

[0057] Step 4: Perform early warning verification on the dynamic features after the contour refinement process. Compare this dynamic feature with the feature contours pre-stored in the model library to confirm whether there is a comparison result. If there is, perform early warning processing. If not, no processing is required. The specific method for performing early warning verification is as follows:

[0058] Compare the dynamic feature with the feature contours pre-stored in the model library one by one: Make the center point of the dynamic feature coincide with the center point of the feature contour, and control the rotation of the dynamic feature to identify whether the coincidence rate between its dynamic feature and the feature contour reaches 90%. If it reaches, it means there is a comparison result, and directly control the alarm to perform early warning processing to alert external operators. If it does not reach, no processing is performed.

[0059] Specifically, after the center point coincidence process is completed, identify the compliance of the coincidence rate between the corresponding contours. If it meets the standard, it means that such features belong to human features or animal features, and then directly perform early warning. If it does not meet the standard, it means that the object that invaded may be other items such as floating objects, and then no manual intervention is required.

[0060] Second Embodiment

[0061] Combined with Figure 2 , an image recognition security processing system for a security curtain net, includes:

[0062] A feature confirmation end, which identifies whether there are items passing through based on the laser detectors set within the security curtain net, and confirms the image recognition signal based on the recognition result;

[0063] A dynamic feature locking end, which controls a high-definition camera to obtain the images of the area where the security curtain net is located based on the generated image recognition signal, and locks the dynamic features existing in multiple frames of images by analyzing the pixel point differences between adjacent frames of images;

[0064] The contour refinement processing end, based on the dynamic features associated with the corresponding image set, first confirms the contour partitions from within the edge contours associated with the dynamic features, then confirms the average refinement connection lines from different contour partitions, and refines the edge contours of this dynamic feature according to the different average refinement connection lines associated with different contour partitions;

[0065] The early warning processing end performs an early warning verification on the dynamic feature after the contour refinement processing, compares this dynamic feature with the feature contours pre-stored in the model library to confirm whether there is a comparison result. If there is, early warning processing is performed; if not, no processing is required.

[0066] Some of the data in the above formula are numerically calculated after removing their dimensions, and the content not described in detail in this specification belongs to the prior art well-known to those skilled in the art.

[0067] The above embodiments are only used to illustrate the technical method of the present invention and not to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical method of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present invention.

Claims

1. An image recognition security processing method for a security curtain net, characterized in that, Including the following steps: Step 1: Based on the laser detectors set within the security curtain net, identify whether there are items passing by, and based on the identification result, confirm the image recognition signal; Step 2: Based on the generated image recognition signal, control the high-definition camera to obtain the images of the area where the security curtain net is located, and by analyzing the pixel differences between adjacent frame images, lock the dynamic features existing in multiple frame images; Step 3: Based on the dynamic features associated with the corresponding image set, first confirm the contour partitions from within the edge contours associated with the dynamic features, then confirm the average quantity refinement connections from different contour partitions, and based on the different average quantity refinement connections associated with different contour partitions, refine the edge contours of this dynamic feature; Step 4: Conduct early warning verification on the dynamic feature after the contour refinement process, compare this dynamic feature with the pre-stored feature contours in the model library to confirm whether there is a comparison result. If there is, conduct early warning processing; if not, no processing is required.

2. The image recognition security processing method for a security screen net according to claim 1, wherein, In the above Step 1, the specific method for confirming the image recognition signal is: Monitor the detection signal associated with the laser detector in real time. When the signal value associated with the detection signal does not belong to the preset interval, it means that there are items passing by, and directly generate an image recognition signal. The endpoint values of the preset interval are both preset values.

3. The image recognition security processing method for a security screen net according to claim 1, wherein, In the above Step 2, the specific method for locking the dynamic feature is: Sort the images belonging to different frames according to the time sequence relationship to confirm the image set; Randomly select a point position within a single image and mark it as the zero point, and confirm the position features of other points based on the position differences between other points and the zero point. By analogy, confirm the position features of several points associated with different frame images within the image set; And calibrate the pixel values associated with different frame images in the image set as X i-k , where i represents different frame images, and k represents the points associated with different position features. Confirm the pixel difference of the points of the same position feature in adjacent frame images, and use XC j-k = |X j-k - X (j+1)-k | to confirm the pixel difference XC j-k , where j ∈ i, assuming i = 1, 2,..., n, then 1 ≤ j ≤ n - 1; It will satisfy: XC j-k Points where ≥ Y1 are marked as differential points, and the area covered by several differential points within the corresponding frame image is marked as a differential area. Then, the contour of the outer area of the differential area is marked: Based on the different pixel values associated with different pixel points within the frame image, the Sobel algorithm is used to calculate the horizontal gradient H k and the vertical gradient S k for each pixel point, and the following formula is used: to confirm the comprehensive gradient ZH k associated with the corresponding pixel point. Pixel points where ZH k > Y2 are marked as gradient pixel points, and the internal area associated with several consecutive gradient pixel points is denoted as the contour area. Both Y1 and Y2 are preset values; Based on the confirmed contour area and difference area within the corresponding frame image, confirm the contour area to which the difference area belongs, and record this contour area as the feature area of this frame image; Based on the different feature areas determined within different frame images, select the feature area with the largest area and mark it as the dynamic feature of this image set.

4. The image recognition security processing method for a security screen network according to claim 3, wherein, Does not meet ZH k Pixels less than or equal to Y2 are not calibrated at all.

5. A method for image recognition security processing of a security screen net according to claim 1, characterized in that, In the above Step 3, the specific method for confirming the contour partitions from within the edge contours associated with the dynamic feature is: Combine the calibrated dynamic feature with the two-dimensional coordinate system to confirm the different two-dimensional coordinates associated with different pixel points within the dynamic feature, then perform mean value processing on several groups of two-dimensional coordinates to confirm the mean coordinate, and mark the confirmed mean coordinate within the dynamic feature, which is recorded as the center point of this dynamic feature; Confirm the edge point positions associated with the outermost edge of the dynamic feature, connect the center point with different edge point positions to lock the point position connection line. Then, based on several groups of gradient pixel points confirmed outside the dynamic feature, confirm the gradient areas covered by several groups of gradient pixel points, and then confirm the part of the connection line within the gradient area among the point position connection lines, which is recorded as the feature connection line of the corresponding gradient area, and determine the connection line length associated with each feature connection line; Perform feature confirmation on adjacent feature connection lines to lock the set of clustered feature connection lines: Randomly select a set of feature connection lines and use the clockwise direction as the processing direction. Use the formula: Feature difference = |Length of the previous set of feature connection lines - Length of the next set of feature connection lines|. If the feature difference ≤ Y3, where Y3 is a preset value, then label the adjacent two sets of feature connection lines as the set of clustered feature connection lines; otherwise, label the corresponding feature connection lines as dividing lines. Based on the positions of different dividing lines within the gradient region, divide the corresponding gradient region into multiple different contour partitions.

6. The image recognition security processing method for a security screen net according to claim 5, characterized in that, In step 3, the specific method for calibrating different average quantity refinement connection lines associated with different contour partitions is as follows: Randomly select a set of pixel points from the pixel points associated with each different vertical row within the contour partition and denote them as the selected points. Then, connect several selected points to confirm a set of internal connection lines. Sum the pixel values corresponding to the multiple sets of pixel points associated with the upper part of the internal connection line to confirm the upper feature sum Z1. Then, sum the pixel values corresponding to the multiple sets of pixel points associated with the lower part of the internal connection line to confirm the lower feature sum Z2. Use the formula: CZ = |Z1 - Z2| to confirm the feature difference CZ. Execute multiple different point selection processes. The internal connection lines associated with each selection process are all different, and confirm the different feature differences CZ associated with different internal connection lines. Select the minimum value from the multiple different feature differences CZ. Denote the internal connection line associated with the minimum value as the average quantity refinement connection line of this contour partition. Connect the average quantity refinement connection lines associated with each contour partition as the edge contour line after this dynamic feature refinement process, and use this edge contour line as the feature contour of this dynamic feature.

7. A method for image recognition and security processing of a security curtain net according to claim 1, characterized in that, In step 4, the specific method for warning verification of the dynamic feature is as follows: Compare the dynamic feature with the feature contours pre-stored in the model library one by one: Make the center point of the dynamic feature coincide with the center point of the feature contour, and control the dynamic feature to rotate to identify whether the coincidence rate between the dynamic feature and the feature contour reaches 90%. If it reaches, it means there is a comparison result, and directly control the alarm to perform a warning process to alert external operators; if it does not reach, no processing is performed.

8. An image recognition security processing system for a security curtain net, the system operates according to an image recognition security processing method for a security curtain net described in any one of claims 1-7, characterized in that, Include: A feature confirmation terminal, which identifies whether there is an item passing through based on the laser detectors set within the security screen network, and confirms an image recognition signal based on the recognition result. A dynamic feature locking terminal, which controls a high-definition camera to acquire the images of the area where the security screen network is located based on the generated image recognition signal, and locks the dynamic features existing within multiple frames of images by analyzing the pixel point differences between adjacent frames of images. A contour refinement processing terminal, which, based on the dynamic features associated with the corresponding image set, first confirms the contour partitions from the edge contours associated with the dynamic features, then confirms the average quantity refinement connection lines from different contour partitions, and refines the edge contour of this dynamic feature according to the different average quantity refinement connection lines associated with different contour partitions. The early warning processing terminal conducts early warning verification on the dynamic features after the contour refinement process, compares this dynamic feature with the feature contours pre-stored in the model library to confirm whether there is a comparison result. If there is, early warning processing is carried out; if not, no processing is required.

Images