A method for representing a multi-zone intrusion with binary bits

Abstract

The application provides a method for representing multi-region intrusion by binary bits, comprising the following steps: S1, positioning a target position; S2, judging a state relation between the target position and a multi-region; S3, setting a representation meaning of a plurality of binary bits; S4, mapping the state relation between the target and the multi-region to the plurality of binary bits; and S5, analyzing the binary bits. The application uniformly represents the state of the target and the multi-region by binary bits, and ingeniously represents the position relation between the target and the multi-region by the method of binary bits.

Description

Technical Field

[0001] This invention belongs to the field of intelligent video processing technology, and specifically relates to a method for representing multi-region intrusion using binary bits. Background Technology

[0002] With the rapid development of artificial intelligence technology, intelligent multi-object detection has been applied to countless households. Human detection, vehicle detection, pet detection, and other applications are becoming increasingly widespread. Furthermore, as the accuracy of target detection algorithms improves, other needs have emerged. For example, after identifying a target, obtaining its coordinates and determining whether it appears within a designated area is frequently required in fields such as electronic fences, intrusion detection in unsafe areas, smart cities, and security monitoring. Specific scenarios include: human detection followed by facial recognition. This saves manpower and resources in company time clocks and access control systems. Through facial registration, the corresponding person's identity information is stored in a database, and identification can be efficiently completed by matching the individuals. The same applies to license plate recognition in parking garages. Similarly, the maturity of human detection technology has brought higher demands to related fields that require human detection. Dangerous areas such as high-speed rail, subway, train, mine, factory, and chemical plant are off-limits to non-professionals and under certain circumstances. If we follow the traditional method of sending people to keep watch, the risk depends entirely on the staff, who must be highly focused and at all times. This places extremely high demands on people. Therefore, the area intrusion function was proposed after the maturity of human detection technology.

[0003] For a single area, there are corresponding methods to determine whether a humanoid target has intruded into that area, namely the intersection-union ratio (IUU) of the area. Currently, most intrusion detection methods on the market set up a single area within a video frame. For intrusion detection in a single area, most methods are based on positional relationships. If a target is found, all targets in that area are directly reported, while targets outside the area are filtered out. Multi-area intrusion detection has emerged in response to the increasing complexity of applicable scenarios in practical applications.

[0004] However, for intrusion detection across multiple regions, the complexity of managing intrusion relationships between targets and regions increases with the increase in region types, the complexity of region location relationships, and the diversification of target types. For multiple regions, building upon single-region intrusion methods, determining the relationship between a target and multiple defined regions has become a common challenge and focus in the field. Summary of the Invention

[0005] To address the aforementioned issues, this application aims to propose a binary-based method for representing the state of human targets and multiple regions in fields such as human detection, vehicle detection, and pet detection. This state allows for the determination of whether a target has intruded into certain specific regions, thus replacing the traditional intelligent method that relies on human observation. This method is not only suitable for representing intrusions of multiple targets and multiple regions but can also be used to represent the departure detection needs of multiple targets and multiple regions.

[0006] The location relationship between the target and multiple areas was determined. The location relationship between the target and different areas can be seen simply and clearly through binary bits. The intrusion status of a specific area is mapped to the corresponding bit, which helps to carry out subsequent intrusion alarm work related to bit operations.

[0007] Specifically, the present invention provides a method for representing multi-region intrusion using binary bits, the method comprising the following steps:

[0008] S1: Target location positioning:

[0009] S1: Use object detection algorithms to detect objects in the image and determine the object's category and location; S2: Determine the relationship between the object's location and the state of multiple regions:

[0010] Calculate the intersection-union ratio (IoU) between the target bounding box and multiple defined region bounding boxes;

[0011] When IoU > 0.5, the target is determined to be within the defined area;

[0012] By calculating the IoU between the target and multiple regions, the state relationship between the target and each region is determined;

[0013] The value of IoU is between 0 and 1, and it is calculated as follows:

[0014] IoU = (target bounding box ∩ set box) / (target bounding box ∪ set box);

[0015] S3: Defines the meaning of the binary bits:

[0016] Use an int64_t variable intrude, convert it to an 11-bit binary number, to represent the intrusion status of the target and multiple areas;

[0017] The meanings of binary bits are as follows:

[0018] L K J I H G F E CB A

[0019] binary bits L->E C->B A meaning Is there any intrusion into the regional perimeter? Perimeter Intrusion Types Intrusion flag present

[0020] Location information is as follows:

[0021] A is the least significant bit, and L is the most significant bit;

[0022] A: Is there an intrusion perimeter flag?

[0023] B, C: Flags for region type, where B: polygonal region type, C: rectangular region type;

[0024] E, F, G, H, I, J, K, L: Polygonal and rectangular regions. The maximum number of region-type boundary bits is 8. The order of E->L represents the 1st to 8th region-type boundaries, which is related to the order in which the boundaries are set. There is no limit to the number of polygonal and rectangular regions. 0 / 1 in the 8 positions indicates no / intrusion at the corresponding boundary. S4: Maps the state relationship between the target and multiple regions to binary bits.

[0025] Iterate through the state relationships between the target and multiple regions, and count the number and type of regions invaded by the target.

[0026] Based on the statistical results, assign values ​​to the binary bits:

[0027] If the target has invaded a certain type of area, the corresponding type flag B or C is set to 1;

[0028] Based on the regional order, set the regional perimeter bit E->L of the target intrusion to 1;

[0029] The state relationship between the target and multiple regions is encoded into an 11-bit binary number intrude through mapping;

[0030] S5: Parsing binary bits:

[0031] In step S3, the value of the variable intrude is analyzed. Since the previous four steps have mapped the intrusion status values ​​of the target and multiple areas to 11 binary bits inturde, it is sufficient to determine whether the intrusion status values ​​of the target and certain areas of interest are 1.

[0032] The target detection algorithms include, but are not limited to, the YOLO series, the RCNN series, or the SSD algorithm; the RCNN series is a representative algorithm based on region detection, the YOLO is a representative algorithm based on region extraction, and the SSD is an improvement based on the first two series.

[0033] In step S1, all targets of interest, i.e., objects, in the image are identified, and their categories and locations are determined. This step can be summarized as target detection. The preferred method is to use the YOLOv7 algorithm for humanoid targets.

[0034] In step S3, the number of binary bit region boundary bits E->L can be expanded or reduced according to actual needs.

[0035] In step S3, the region type flags B and C can be increased according to actual needs to support more types of regions.

[0036] In step S3, it is assumed that 01000010 indicates that the target has invaded the 2nd and 7th regional perimeters.

[0037] Step S4 further includes: based on step S2, the state relationship between the same target and different regions is known. Therefore, for a certain target, multiple regions are traversed, the state values ​​of the target and the corresponding regions are counted, and the number of times the same region is invaded is counted. Based on this number, the corresponding binary bits can be assigned values. Assuming there are 3 polygonal regions, and the target invades 2 of them, the polygonal region flag bit is recorded as 1. At the same time, the order of the three polygonal regions is also related to the traversal of multiple regions, so the order is one-to-one. Through this method, the state relationship between the target and multiple regions can be mapped to the binary bits of 11.

[0038] In step S5, assuming there are 8 areas, to determine whether a humanoid target has entered the first area, the judgment statement 1<<(4-1)&intrude&&class_id=human can be used. If the judgment statement is true, the target has invaded the first area; otherwise, it has not invaded. The 4 in the judgment statement is because the first area is located in the lower 4 bits of the 11 binary numbers.

[0039] In this method, the mapping order of binary bits corresponds one-to-one with the order of the set regions to ensure the accuracy of the parsing results.

[0040] The method is applicable to multi-region intrusion detection scenarios, including the ability to simultaneously support intrusion status representation of polygonal and rectangular regions.

[0041] Therefore, the advantages of this application are: the method of using binary bits to represent multi-region intrusion proposed in this invention can solve the situation of multi-region intrusion, and can clearly determine whether the target is an intruded region when the location relationship of multiple regions is complex and there are many types. Moreover, it is not restricted by the type of target, has universality and wide applicability, and is more conducive to the development of underlying hardware. Attached Figure Description

[0042] The accompanying drawings, which are provided to further illustrate the invention and form part of this application, are not intended to limit the scope of the invention.

[0043] Figure 1 This is a block diagram illustrating the principle of binary representation of multi-region intrusion.

[0044] Figure 2 This is a flowchart illustrating the method described in this application.

[0045] Figure 3 This is a schematic diagram of the code implementation of the region setting method in this application.

[0046] Figure 4 This is a schematic diagram of the code implementation for traversing the detection box and recording the intrusion status in the method of this application.

[0047] Figure 5 This is a schematic diagram of the code implementation in this application that maps the human detection box and the intrusion state of the region to binary bits. Detailed Implementation

[0048] To better understand the technical content and advantages of the present invention, the present invention will now be described in further detail with reference to the accompanying drawings.

[0049] This invention provides a method for characterizing target intrusion when monitoring self-defined multiple areas in video footage, that is, a means of determining whether a target has entered a specially defined multiple area.

[0050] This invention proposes a method for representing multi-region intrusions using binary bits. The method mainly consists of five steps, such as... Figure 1 As shown in the diagram, the principle block diagram mainly involves the five steps proposed in this paper. The location of the corresponding target falls under the scope of target detection; therefore, the accuracy of target detection is the primary prerequisite for intrusion judgment. Besides the IOU method used in this paper, other methods, such as distance, can be used to determine the relationship between the target location and the state of multiple regions. The meaning of the multiple binary bits can be set according to the different regions to be allocated. It is important to note that the configuration order must correspond one-to-one with the mapping order of the target's state relationship with multiple regions onto the multiple binary bits. Only the information on the finally parsed binary bits can be mapped to the multiple regions set. Using the method proposed in this paper, different types of regions can be set, or multiple different types can be set corresponding to different numbers of regions. All of this can be clearly understood through the binary bits, freeing it from the limitations of the number of regions and the limitations of region overlap.

[0051] Specifically, such as Figure 2 As shown, the main implementation steps of this method are as follows:

[0052] Step S1. Locate the position of the corresponding target.

[0053] The step of finding all targets (objects) of interest in an image and determining their category and location can be summarized as object detection. Since various objects have different appearances, shapes and poses, coupled with interference from factors such as imaging time and occlusion, object detection has been widely studied in the field of computer vision. To this day, object detection algorithms have emerged in an endless stream, such as the R-CNN series and the YOLO series. The R-CNN series is a representative algorithm based on region detection, while YOLO is a representative algorithm based on region extraction. In addition, there is the well-known SSD, which is an improvement based on the first two series, and its corresponding detection accuracy is getting higher and higher.

[0054] This paper utilizes the YOLOv7 algorithm, which is used for human-shaped target detection. Therefore, this paper takes the example of a human-shaped target invading multiple regions.

[0055] Step S2. Determine the relationship between the target location and the state of multiple regions.

[0056] In computer vision and data annotation, the Intersection over Union (IoU) ratio in object detection is a metric used to evaluate the similarity between two bounding boxes. The IoU value between the defined ground truth box and the predicted box represents their degree of overlap. IoU values ​​range from 0 to 1, and are calculated as follows:

[0057] IoU = (Target bounding box ∩ Set bounding box) / (Target bounding box ∪ Set bounding box)

[0058] This paper defines a target as being within a defined region when the IoU > 0.5. By calculating the IoU between the target and multiple regions, multiple state values ​​are obtained, thus determining the corresponding relationships. Step S3. Define the meaning of the binary bits.

[0059] This article defines an int64_t intrude; after conversion to binary, it uses 11 bits to represent the intrusion status of the target and multiple regions.

[0060] L K J I H G F E C B A

[0061] binary bits L->E C->B A meaning Is there any intrusion into the regional perimeter? Perimeter Intrusion Types Intrusion flag present

[0062] Location information is as follows:

[0063] A represents the least significant bit, and L represents the most significant bit.

[0064] A: Is there an intrusion perimeter marker?

[0065] B, C: Region type. B: Polygonal region type, C: Rectangular region type.

[0066] E, F, G, H, I, J, K, L: Polygonal and rectangular regions. The maximum number of region-type perimeter bits is 8. The order of E->L represents the 1st to 8th region-type perimeters respectively (related to the order in which the perimeters are set). There is no limit to the number of polygons and rectangles. 0 / 1 in the 8 positions indicates no / some intrusion into the corresponding perimeter. For example, 01000010 indicates that the target intruded into the 2nd and 7th region-type perimeters.

[0067] Step S4. Map the state relationship between the target and multiple regions to binary bits.

[0068] As shown in step S2, the state relationships between the same target and different regions are known. Therefore, for a given target, multiple regions are traversed, and the state values ​​of the target and corresponding regions are statistically analyzed. The number of times the same region is invaded is also counted. Based on this count, the corresponding binary bits can be assigned values. For example, if there are three polygonal regions and the target invades two, the polygonal region flag is recorded as 1. The order of the three polygonal regions is also related to the traversal of multiple regions; therefore, the order is one-to-one. This method maps the state relationships between the target and multiple regions to 11 binary bits.

[0069] Step S5. Parse binary bits

[0070] Analyze the value of the variable `intrude` in step S3. Since the previous four steps have mapped the intrusion status values ​​of the target and multiple areas to the 11-bit binary variable `intrude`, we only need to check if the intrusion status value of the target and certain areas of interest is 1. For example, if we set 8 areas and want to determine if a humanoid target has entered the first area, we can use the conditional statement `1 << (4-1) &intrude && class_id = human`. If the conditional statement is true, the target has invaded the first area; otherwise, it has not. The 4 in the conditional statement is because the first area is located in the lower 4 bits of the 11-bit binary variable.

[0071] In summary, this method represents multi-region intrusion using binary bits. Through steps such as target detection, IoU calculation, binary bit mapping, and parsing, it achieves efficient representation and judgment of the intrusion status of targets and multiple regions. This method supports various region types and numbers, offering flexibility and scalability.

[0072] The following are implementation examples of specific embodiment code, including:

[0073] 1. Regional settings:

[0074] Initialize the region parameter structure param;

[0075] Set the number of regions in param.permcnt;

[0076] Configure type and coordinate information for each region;

[0077] 2. Intrusion detection:

[0078] Use YOLOv7 to detect human targets and obtain the bounding box coordinates (e.g., [452,48,530,151]). Use the IOU (Intersection over Union) method to calculate the intersection of the detection box with each region.

[0079] Set intrusion status flags:

[0080] There is an intersection: perimeter = 1;

[0081] No intersection: perimeter = 0;

[0082] 3-State Coding:

[0083] Encode the intrusion status into a binary bit representation:

[0084] 1) Initialize variables;

[0085] 2) Traverse the detection box and region, and record the intrusion status;

[0086] 3) Binary encoding;

[0087] Furthermore, including:

[0088] First, the target is detected using the corresponding detection model. This paper uses YOLOv7 as an example. When the detection result is a human shape, the corresponding human bounding box coordinates [x0, y0, x1, y1] are [452, 48, 530, 151].

[0089] 1. If you set two regions (the number of regions is unlimited, but we'll use two regions as an example here), and the types are polygon and rectangle (other types are also supported, but we'll use polygon and rectangle as examples here for now), the region settings are as follows: Figure 3 As shown:

[0090] Set the number of regions: param.permcnt = 2;

[0091] First polygonal region:

[0092] int perm_index = 0;

[0093] param.perms[perm_index].pcnt=4;

[0094] The region type: param.perms[perm_index].mod = 0;

[0095] param.perms[perm_index].p=(IVSPoint*)malloc(param.perms[index].pcnt*sizeof(IVSPoint));

[0096] param.perms[perm_index].p[0].x=0;

[0097] param.perms[perm_index].p[0].y=0;

[0098] param.perms[perm_index].p[1].x=500;

[0099] param.perms[perm_index].p[1].y=0;

[0100] param.perms[perm_index].p[2].x=500;

[0101] param.perms[perm_index].p[2].y=383;

[0102] param.perms[perm_index].p[3].x=0;

[0103] param.perms[perm_index].p[3].y=383;

[0104] The second rectangular area:

[0105] perm_index += 1;

[0106] param.perms[perm_index].mod=1;

[0107] param.perms[perm_index].r.ul.x=500;

[0108] param.perms[perm_index].r.ul.y=1;

[0109] param.perms[perm_index].r.br.x=620;

[0110] param.perms[perm_index].r.br.y=383;

[0111] 2. Use the IOU method to determine if the human bounding box is within the corresponding region. For example, [452,48,530,151] intersects with two defined regions. Therefore, it is determined to have intruded into both regions, and the intrusion status value perimeter of the human bounding box in different regions is set to 1; otherwise, the perimeter is set to 0.

[0112] 3. Convert human detection and area intrusion status to binary.

[0113] 1) Initialize variables;

[0114] int64_t intrude_value=0,intrude_state=0; int polygon_num_intrude=0,

[0115] rectangle_num_intrude = 0; / / Record the number of intrusion zones

[0116] vector <int>polygon_state, rectangle_state; / / Records the intrusion status of each detection box and multiple defined regions.

[0117] 2) Traverse the detection box and region, and record the intrusion status, such as Figure 4 As shown:

[0118] Implement intrusion status determination for human detection bounding boxes and defined areas, including:

[0119] 1>: Traverse all detected human-shaped bounding boxes;

[0120] 2>: For each human-shaped detection box, traverse the pre-defined regions;

[0121] 3>: When the mode parameter mod is 0, perform a region type check:

[0122] If the current region is a polygonal region, then further determine the intrusion status value (perimeter) of the human-shaped detection box:

[0123] If the perimeter is equal to 1, then mark the intrusion status of the polygonal region as 1 and increment the polygonal region intrusion counter;

[0124] Otherwise, mark the intrusion status of the polygonal region as 0;

[0125] If the current region is a rectangular region, the same judgment logic as for polygonal regions will be used.

[0126] 3) Map the human detection bounding box and the intrusion status of the region to binary bits; such as... Figure 5 As shown:

[0127] Intrusion flag assignment based on regional intrusion statistics includes: 1>: Counting the number of intrusions in polygonal regions (polygon_num_intrude) and the number of intrusions in rectangular regions (rectangle_num_intrude);

[0128] 2>: If the total number of intrusions is greater than 0, set the least significant bit (bit A) of the intrusion flag (intrude_value) to indicate that a regional intrusion has occurred;

[0129] 3>: If the number of intrusions in the polygonal region is greater than 0, then set the second least significant bit (B bit) of the intrusion flag to indicate that there is an intrusion in the polygonal region;

[0130] 4>: Traverse the invasion status (polygon_state) of all polygon regions. If a region is invaded (state value is 1), set the corresponding flag bit in the corresponding high bit (starting from the 4th bit) of the invasion flag bit to identify the specific invaded polygon region.

[0131] Optionally, the starting offset of the high-order flag is determined by the bit layout of the intrusion flag; 5>: The intrusion status of the rectangular area is processed in the same way to complete the full assignment of the intrusion flag.

[0132] Encoding example:

[0133] The final binary representation of the human detection bounding box [452,48,530,151] and the two regions:

[0134] The intrusion status result for the polygonal region [(0,0),(500,0),(500,383),(0,383)] and the rectangle [(500,1),(620,383)] is 00000011111.

[0135] The chips involved in this technical solution can be applied to all series of Ingenic chips and similar chips.

[0136] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, various modifications and variations can be made to the embodiments of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.< / int>

Claims

1. A method for representing multi-region intrusion using binary bits, characterized in that, The method includes the following steps: S1: Target location positioning: Target detection algorithms are used to detect targets in images and determine their category and location. S2: Determine the relationship between the target location and the state of multiple regions: Calculate the intersection-union ratio (IoU) between the target bounding box and multiple defined region bounding boxes; When IoU > 0.5, the target is determined to be within the defined area; By calculating the IoU between the target and multiple regions, the state relationship between the target and each region is determined; The value of IoU is between 0 and 1, and it is calculated as follows: IoU = (target bounding box ∩ set box) / (target bounding box ∪ set box); S3: Defines the meaning of the binary bits: Use an int64_t variable intrude, convert it to an 11-bit binary number, to represent the intrusion status of the target and multiple areas; The meanings of binary bits are as follows: LKJIHGFECBA Location information is as follows: A is the least significant bit, and L is the most significant bit; A: Is there an intrusion perimeter flag? B, C: Flags for region type, where B: polygonal region type, C: rectangular region type; E, F, G, H, I, J, K, L: Polygonal and rectangular regions. The maximum number of region-type boundary bits is 8. The order of E->L represents the 1st to 8th region-type boundaries, which is related to the order in which the boundaries are set. There is no limit to the number of polygonal and rectangular regions. 0 / 1 in the 8 positions indicates no / intrusion at the corresponding boundary. S4: Maps the state relationship between the target and multiple regions to binary bits. Iterate through the state relationships between the target and multiple regions, and count the number and type of regions invaded by the target. Based on the statistical results, assign values ​​to the binary bits: If the target has invaded a certain type of area, the corresponding type flag B or C is set to 1; Based on the regional order, set the regional perimeter bit E->L of the target intrusion to 1; The state relationship between the target and multiple regions is encoded into an 11-bit binary number intrude through mapping; S5: Parsing binary bits: In step S3, the value of the variable intrude is analyzed. Since the previous four steps have mapped the intrusion status values ​​of the target and multiple areas to 11 binary bits inturde, it is sufficient to determine whether the intrusion status values ​​of the target and certain areas of interest are 1.

2. The method for representing multi-region intrusion using binary bits according to claim 1, characterized in that: The target detection algorithms include, but are not limited to, the YOLO series, the RCNN series, or the SSD algorithm; the RCNN series is a representative algorithm based on region detection, the YOLO is a representative algorithm based on region extraction, and the SSD is an improvement based on the first two series.

3. The method for representing multi-region intrusion using binary bits according to claim 2, characterized in that: In step S1, all targets of interest, i.e., objects, in the image are identified, and their categories and locations are determined. This step can be summarized as target detection. The preferred method is to use the YOLOv7 algorithm for humanoid targets.

4. The method for representing multi-region intrusion using binary bits according to claim 1, characterized in that: In step S3, the number of binary bit region boundary bits E->L can be expanded or reduced according to actual needs.

5. The method for representing multi-region intrusion using binary bits according to claim 1, characterized in that: In step S3, the region type flags B and C can be increased according to actual needs to support more types of regions.

6. The method for representing multi-region intrusion using binary bits according to claim 1, characterized in that: In step S3, it is assumed that 01000010 indicates that the target has invaded the 2nd and 7th regional perimeters.

7. The method for representing multi-region intrusion using binary bits according to claim 1, characterized in that: Step S4 further includes: based on step S2, the state relationship between the same target and different regions is known. Therefore, for a certain target, multiple regions are traversed, the state values ​​of the target and the corresponding regions are counted, and the number of times the same region is invaded is counted. Based on this number, the corresponding binary bits can be assigned values. Assuming there are 3 polygonal regions, and the target invades 2 of them, the polygonal region flag bit is recorded as 1. At the same time, the order of the three polygonal regions is also related to the traversal of multiple regions, so the order is one-to-one. Through this method, the state relationship between the target and multiple regions can be mapped to the binary bits of 11.

8. The method for representing multi-region intrusion using binary bits according to claim 1, characterized in that: In step S5, assuming there are 8 areas, to determine whether a humanoid target has entered the first area, the judgment statement 1<<(4-1)&intrude&&class_id=human can be used. If the judgment statement is true, the target has invaded the first area; otherwise, it has not invaded. The 4 in the judgment statement is because the first area is located in the lower 4 bits of the 11 binary numbers.

9. The method for representing multi-region intrusion using binary bits according to claim 1, characterized in that: In this method, the mapping order of binary bits corresponds one-to-one with the order of the set regions to ensure the accuracy of the parsing results.

10. The method for representing multi-region intrusion using binary bits according to claim 1, characterized in that: The method is applicable to multi-area intrusion detection scenarios. This includes the ability to simultaneously support intrusion status representation for both polygonal and rectangular regions.

Images