AI algorithm blind spot warning method

Abstract

The present invention relates to an AI algorithm blind spot warning method. The present invention includes the following steps: S1, image acquisition and input; S2, image detection of a target object in a road blind spot area requiring warning; S3, image tracking of the target object in the warning area; S4, drawing a line parallel to the warning area to determine direction L1; S5, drawing a warning area to determine area m; S6, determining the moving direction L2 of the target object; S7, calculating the angle θ between L1 and L2, and determining whether it is greater than a threshold C; S8, if the target object crosses the warning blind spot area, determining whether it is moving away from area m of the current warning area or approaching from area m of the current warning area: by comparing the position change of the target object in several consecutive frames, it can be determined whether it is approaching or moving away; if the threshold for the target object to approach or move away in the warning area reaches the warning value, a warning signal is issued; S9, the result is output.

Description

Technical Field

[0001] The present invention relates to an AI algorithm blind spot warning method, belonging to the field of intelligent transportation technology. Background Art

[0002] Road safety is crucial in urban traffic scenarios. Current road systems often create blind spots due to greenery, barriers, buildings, and other obstructions, making it difficult for traffic participants to detect approaching pedestrians or vehicles. When objects moving simultaneously toward a blind spot from different directions intersect and collide, increasing the risk of accidents. Summary of the Invention

[0003] The purpose of the present invention is to overcome the above-mentioned deficiencies in the prior art and to provide a blind spot warning method with an AI algorithm having a rational structural design. By using computer vision technology to detect and track pedestrian and vehicle images, the method can provide real-time warnings of dangerous situations in which traffic accidents may occur due to road blind spots.

[0004] The technical solution adopted by the present invention to solve the above problem is: the AI ​​algorithm blind spot warning method is characterized by comprising the following steps:

[0005] S1. Image acquisition and input:

[0006] S2. Image detection of targets (pedestrians, vehicles) in the blind spot area requiring warning: Use background subtraction or deep learning methods in computer vision to detect people and vehicles in the warning area. Appropriate algorithms and logic can be selected based on the characteristics of the warning area.

[0007] S3. Image tracking of targets (pedestrians, vehicles) in the warning area: After detecting the target (pedestrian, vehicle), it is tracked using the tracking algorithms in computer vision, such as Kalman filtering and Mean-Shift algorithm, so that the position of the target (pedestrian, vehicle) can be determined frame by frame;

[0008] S4. Draw a line parallel to the warning area to determine direction L1: Determine the extension line of the warning area as direction L1, and use a line detection algorithm in image processing, such as Hough transform, to determine the position and direction of the warning area;

[0009] S5. Draw a warning area to determine area m: Determine an area m parallel to the warning area. The area m is used to limit the detection area for pedestrians and vehicles. An edge detection algorithm in image processing, such as the Canny algorithm, can be used to determine the scope of the warning area.

[0010] S6. Based on the results of S2 and S3 recognition and tracking, the target's moving direction L2 can be determined: by comparing the position difference of the target obtained by tracking in adjacent frames, its moving direction L2 can be calculated. The vector dot multiplication formula can be used: L2 = atan2(v\_y, v\_x).

[0011] Where v_x and v_y are the x-component and y-component of the target's velocity vector, respectively;

[0012] S7. Calculate the angle θ between L1 and L2 and determine whether it is greater than a threshold C, where C = θ = 90°: Calculate the angle d between L1 and L2 using the vector angle formula, d = acos(cos(L1) * cos(L2) + sin(L1) * sin(L2)). If d is greater than the threshold C, determine that the target has crossed the warning area.

[0013] S8. If the target object crosses the warning blind area, determine whether it is moving away from or approaching the current warning area within area m:

[0014] By comparing the position changes of the target in several consecutive frames, it can be determined whether it is approaching or moving away;

[0015] If the target object approaches or moves away from the warning area and the threshold reaches the warning value, a warning signal is issued;

[0016] S9. Result output: The system sends the image of the detected target object and the warning signal to the remote monitoring center or mobile phone through the network so that timely response measures can be taken.

[0017] Furthermore, in said S6, ,

[0018] Where a and b are two vectors, v is their dot product, and d is the angle between them.

[0019] Furthermore, in S7, the direction vectors of L1 and L2 can generally be expressed as:

[0020] L1 = (k1, k2), L2 = (k3, k4),

[0021] Then the angle d between them can be calculated using the vector angle formula:

[0022] d = arccos(cos(L1, L2))

[0023] = arccos( (k1 * k3 + k2 * k4) / (sqrt(k1^2 + k2^2) * sqrt(k3^2 + k4^2)) ).

[0024] Furthermore, the AI ​​algorithm blind spot warning method is implemented through an AI algorithm blind spot warning system, which includes an AI algorithm camera for monitoring the movement direction of the target object and determining whether the target object is in the designated area m, and a roadside warning light for warning traffic participants. The roadside warning light is wirelessly or linearly connected to the AI ​​algorithm camera.

[0025] Furthermore, the AI ​​algorithm blind spot warning system also includes an underground warning light for warning traffic participants, and the underground warning light is wirelessly or linearly connected to the AI ​​algorithm camera.

[0026] Furthermore, the AI ​​algorithm blind spot warning system also includes a display screen for prompting traffic participants and displaying text information, and the display screen is wirelessly or linearly connected to the AI ​​algorithm camera.

[0027] Furthermore, the AI ​​algorithm blind spot warning system also includes a projection light for warning traffic participants, and the projection light is wirelessly or linearly connected to the AI ​​algorithm camera.

[0028] Furthermore, the AI ​​algorithm blind spot warning system also includes a laser projection device for warning traffic participants, and the laser projection device is wirelessly or linearly connected to the AI ​​algorithm camera.

[0029] Compared with the existing technology, the present invention has the following advantages: the AI ​​algorithm blind spot warning method uses computer vision technology to detect and track pedestrians and vehicles, and determines whether they cross the warning blind spot, and issues warning signals in time, which can effectively reduce the risk of traffic accidents caused by road blind spots. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 4 is a flow chart of the AI ​​algorithm blind spot warning method according to an embodiment of the present invention.

[0031] Figure 2 2 is a schematic diagram of a zebra crossing warning system according to an embodiment of the present invention.

[0032] Figure 3 1 is a schematic diagram of an intersection warning according to an embodiment of the present invention.

[0033] Figure 4 This is a schematic diagram of a T-junction warning according to an embodiment of the present invention.

[0034] Figure 5 This is a schematic diagram of a vehicle right turn blind spot warning according to an embodiment of the present invention.

[0035] Figure 6 yes Figure 5 The diagram of the blind spot warning for right turns of vehicles with Zone 2 removed. Zone 4 is surrounded by the blue dotted line.

[0036] Figure 7 yes Figure 5 The diagram of the blind spot warning for right turns of vehicles with Zone 1 removed. The red dotted line in the figure is Zone 2.

[0037] Figure 8 2 is a schematic diagram of a traffic light mode at an intersection according to an embodiment of the present invention.

[0038] Figure 9 3 is a schematic diagram of a model of an AI algorithm blind spot warning method according to an embodiment of the present invention.

[0039] In the picture: AI algorithm camera 1, roadside warning light 2, underground warning light 3, traffic light 4, crossbeam 5. DETAILED DESCRIPTION

[0040] The present invention will be further described in detail below with reference to the accompanying drawings and through examples. The following examples are intended to explain the present invention but the present invention is not limited to the following examples.

[0041] Example.

[0042] See also Figures 1 to 9 As shown, it should be noted that the structures, proportions, sizes, etc. illustrated in the drawings of this specification are only used to match the contents disclosed in the specification for people familiar with this technology to understand and read, and are not used to limit the limiting conditions for the implementation of the present invention. Therefore, they have no substantial technical significance. Any modification of the structure, change in the proportional relationship or adjustment of the size should still fall within the scope of the technical content disclosed by the present invention without affecting the efficacy and purpose that can be achieved by the present invention. At the same time, if there are references to terms such as "upper", "lower", "left", "right", "middle" and "one" in this specification, they are only for the convenience of description and are not used to limit the scope of the implementation of the present invention. Changes or adjustments in their relative relationships should also be regarded as the scope of the implementation of the present invention without substantially changing the technical content.

[0043] The AI ​​algorithm blind spot warning method in this embodiment (such as Figure 1 As shown), follow these steps:

[0044] S1. Image acquisition and input.

[0045] S2. Image detection of targets (pedestrians, vehicles) in the warning area: Use background subtraction methods or deep learning methods in computer vision to detect people and vehicles in the warning area. You can choose the appropriate logic and target category based on the actual situation of the warning area.

[0046] S3. Image tracking of targets (pedestrians, vehicles) in the warning area: After detecting the target (pedestrian, vehicle), it is tracked using the tracking algorithms in computer vision, Kalman filtering and Mean-Shift algorithm, so that the position of the target (pedestrian, vehicle) can be determined frame by frame.

[0047] S4. Draw a line parallel to the warning area to determine the direction L1: Determine the extension line of the warning area as the direction L1, and use a line detection algorithm in image processing, such as Hough transform, to determine the position and direction of the warning area.

[0048] S5. Draw a warning area to determine area m: Determine an area m parallel to the warning area. The area m is used to limit the detection area of ​​pedestrians and vehicles. The edge detection algorithm in image processing, such as the Canny algorithm, can be used to determine the scope of the warning area.

[0049] S6. Based on the results of S2 and S3 recognition and tracking, the target's moving direction L2 can be determined: by comparing the position difference of the target obtained by tracking in adjacent frames, its moving direction L2 can be calculated. The vector dot multiplication formula can be used: L2 = atan2(v\_y, v\_x).

[0050] Where v_x and v_y are the x-component and y-component of the target's velocity vector, respectively;

[0051] ,

[0052] Where a and b are two vectors, v is their dot product, and d is the angle between them.

[0053] S7. Calculate the angle θ between L1 and L2 and determine whether it is greater than a threshold C: Calculate the angle d between L1 and L2 using the vector angle formula: d = acos(cos(L1) * cos(L2) + sin(L1) * sin(L2)). If d is greater than the threshold C, it is determined that the target has crossed the warning area.

[0054] The direction vectors of L1 and L2 can usually be expressed as:

[0055] L1 = (k1, k2), k1 is the speed in the X direction, k2 is the speed in the Y direction,

[0056] L2 = (k3, k4), k3 is the speed in the X direction, k4 is the speed in the Y direction,

[0057] Then the angle d between them can be calculated using the vector angle formula:

[0058] d = arccos(cos(L1, L2))

[0059] = arccos( (k1 * k3 + k2 * k4) / (sqrt(k1^2 + k2^2) * sqrt(k3^2 + k4^2)) ).

[0060] S8. If the target object crosses the warning blind area, determine whether it is moving away from or approaching the current warning area within area m:

[0061] By comparing the position changes of the target in several consecutive frames, it can be determined whether it is approaching or moving away;

[0062] If the target object approaches or moves away from the warning area and the threshold reaches the warning value, a warning signal is issued;

[0063] S9. Result output: The system sends the detected target image and warning signal to the remote monitoring center or mobile phone via the network so that timely response measures can be taken.

[0064] The AI ​​algorithm blind spot warning method in this embodiment is implemented through an AI algorithm blind spot warning system, which includes an AI algorithm camera 1 for monitoring the movement direction of the target object and determining whether the target object is in the designated area m, roadside warning lights 2 and underground warning lights 3 for warning traffic participants, and a display screen for prompting traffic participants and displaying text information. The roadside warning lights 2, underground warning lights 3 and display screen are all wirelessly connected or linearly connected to the AI ​​algorithm camera 1.

[0065] like Figure 2 As shown in the figure, the zebra crossing warning logic is as follows:

[0066] 1. When a pedestrian enters area 0: roadside warning light 2A and roadside warning light 2B light up. When the pedestrian leaves area 0 and enters area 1: roadside warning light 2A goes out and roadside warning light 2B continues to light. When the pedestrian leaves area 1: roadside warning light 2B also goes out.

[0067] 2. When a pedestrian enters area 2: roadside warning light 2A and roadside warning light 2B light up. When the pedestrian leaves area 2 and enters area 3: roadside warning light 2B goes out and roadside warning light 2A continues to light. When the pedestrian leaves area 3: roadside warning light 2A also goes out.

[0068] 3. When a pedestrian enters area 4: Roadside warning light 2C and roadside warning light 2D light up. When a pedestrian leaves area 4 and enters area 5: Roadside warning light 2C goes out, and roadside warning light 2D continues to light up.

[0069] When the pedestrian leaves area 5: the roadside warning light 2D also goes out.

[0070] 4. When a pedestrian enters area 6, roadside warning lights 2C and 2D light up. When the pedestrian leaves area 6 and enters area 7, roadside warning lights 2D go out, while roadside warning lights 2C remain on. When the pedestrian leaves area 7, roadside warning lights 2C also go out.

[0071] like Figure 3 As shown, the intersection warning logic is as follows, and the left and right side markings are determined by the direction of travel:

[0072] 1. AI algorithm camera 1 in the X-axis direction: When a target in area A approaches the intersection, the roadside warning lights 2 on the left side of area B and the right side of area D are activated. When a target in area C approaches the intersection, the roadside warning lights 2 on the right side of area B and the left side of area D are activated, and the buried warning lights 3 are also lit.

[0073] 2. AI algorithm camera 1 in the Y-axis direction: When a target in area B approaches the intersection, the roadside warning lights 2 on the right side of area A and the left side of area C are activated. When a target in area D approaches the intersection, the roadside warning lights 2 on the left side of area A and the right side of area C are activated, and the underground warning lights 3 are also lit.

[0074] That is to say, the AI ​​algorithm camera 1 arranged at the far end of the X and Y axis directions monitors the travel direction information of the target objects (pedestrians, vehicles, etc.) in the four designated areas A, B, C, and D, and uses the AI ​​algorithm to determine whether there is an intersection risk; if a target object appears at the intersection in the X and Y axis directions at the same time, the system will issue a warning signal, and the underground warning light 3 and the roadside warning light 2 will issue a warning signal; if a target object appears in only one of the areas and moves towards the intersection, the system will not alarm, and if a target object appears in two areas on the same axis, the system will not alarm.

[0075] like Figure 4As shown, the T-junction pre-warning logic is as follows: when a target object in area A approaches the intersection, the roadside warning lights 2 on both sides of area A and the left side of the branch road are activated; when a target object in area B approaches the intersection, the roadside warning lights 2 on both sides of area B and the right side of the branch road are activated; when a target object in the branch road approaches the intersection, all roadside warning lights 2 on both sides of areas A and B are activated, and the underground warning lights 3 are also activated.

[0076] An AI algorithm camera 1 is installed at the far end of the non-motorized vehicle lane to delineate the area that needs to be monitored and the movement direction of the target objects (pedestrians, vehicles, etc.) in the set area. The AI ​​algorithm is used to determine whether there is an intersection risk. If a vehicle in the main lane needs to turn into a branch road, or a target object on the branch road needs to turn into the non-motorized vehicle lane or the main road, and there is a target object in the non-motorized vehicle lane approaching the intersection, the system will issue a warning signal, and the underground warning light 3 and the roadside warning light 2 will issue a warning signal. If there are no targets in areas A and B, or the target object's movement direction is away from or away from the intersection, the system will not alarm.

[0077] like Figure 5-7 As shown, the vehicle right turn blind spot warning plan is as follows:

[0078] An AI algorithm camera 1 is installed at one end of the turning lane in the direction of vehicle front travel. The turning path is demarcated as area 1 for determining whether the vehicle has turning behavior. At the same time, area 2 is set in the turning area in front of area 1 in the direction of vehicle front travel, partially overlapping with area 1. Areas 3 and 4 are set to the left and right of area 1. The AI ​​algorithm camera 1 determines the travel direction of areas 3 and 4. The AI ​​algorithm camera 1, display screen, flashing light, directional speaker, directional antenna and other devices are installed on the cross frame 5.

[0079] The vehicle right turn blind spot warning logic is as follows:

[0080] If area 1 determines that a vehicle is turning, and there is a target object in area 2, and areas 3 and 4 are approaching area 1, as long as one of the thresholds in areas 2, 3, and 4 meets the alarm condition, the system will determine that there is a cross risk and issue an early warning through the alarm device. If only the threshold in area 1 is met, and none of the thresholds in areas 2, 3, and 4, either individually or in any combination, meet the early warning condition, the system will not issue an early warning. If only areas 2, 3, and 4, or any individual or in any combination of these three areas, exist, and there is no alarm threshold in area 1, the system will not issue an alarm.

[0081] The components of the vehicle right turn blind spot warning alarm device are as follows:

[0082] 1. A horizontal frame is installed at a high position at the front of the vehicle in the turning lane, and an AI device, such as an AI algorithm camera 1, is installed;

[0083] a. Install large LED display screens mainly to display text information, such as danger, parking and other text information;

[0084] b. Install high-brightness flashing lights;

[0085] c. Install ultrasonic directional speakers for voice reminders. The ultrasonic directional radar arranges the ultrasonic sound units according to the area where the sound needs to be transmitted, such as an angle of 25 degrees;

[0086] d. It is equipped with a directional image transmission antenna and uses hotspot transmission. When the vehicle passes through area 1 and the system determines that there is a turning behavior, the hotspot signal is automatically connected to the display in the vehicle. The real-time image of the monitored area can be wirelessly transmitted to the display installed in the vehicle, providing traffic participants with real-time images of the vehicle's surroundings. The display will be automatically turned off when the hotspot signal is lost after leaving area 1.

[0087] like Figure 8 As shown, the traffic light mode at the intersection has the following logic:

[0088] An AI algorithm camera 1 is installed in each of the X and Y axis directions. The far end of the AI ​​algorithm camera 1 in the X axis direction is area 1, which is set to approach, and the near end is area 2, which is set to far away. The far end of the AI ​​algorithm camera 1 in the Y axis direction is area 3, which is set to approach, and the near end is area 4, which is set to far away. Four traffic lights 4 are set at the XY axis intersection in the normal mode.

[0089] When the XY axes each trigger one or two thresholds, all yellow lights or red lights will flash to remind traffic participants of possible intersection risks and take measures to avoid them in advance; if only one of the XY axes is in the right direction and this axis triggers both the approach and distance thresholds at the same time, all green lights will flash to remind traffic participants in the two directions relative to this axis to pay attention to the situation in the opposite direction; if only one of the XY axes is in the right direction and this axis only triggers one of the approach and distance thresholds, all green lights will be on steadily; when no threshold is triggered on any of the XY axes, all green lights will be on steadily.

[0090] In addition, it should be noted that the shapes and names of the parts and components of the specific embodiments described in this specification may be different, and the above content described in this specification is only an example of the structure of the present invention. Any equivalent changes or simple changes made based on the structure, features and principles described in the patent concept of the present invention are included in the protection scope of the patent of the present invention. Those skilled in the art of the technology to which the present invention belongs can make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the structure of the present invention or exceed the scope defined by the claims, they should all fall within the protection scope of the present invention.

Claims

1. An AI algorithm blind spot warning method, characterized by: The steps include: S1, image acquisition and input; S2. Image detection: Use background subtraction methods or deep learning methods in computer vision to detect people and vehicles in the blind spots of the road that require warning; S3, Image Tracking: After detecting the target, it is tracked using the tracking algorithms in computer vision, such as Kalman filtering and Mean-Shift algorithm; S4. Determine direction L1: Determine the extension line of the blind spot area of ​​the road that requires warning as direction L1; S5. Determine region m: Determine a region m parallel to the blind spot area of ​​the road that requires warning. The region m is used to limit the detection area for pedestrians and vehicles. S6. Determine the target's moving direction L2 based on the results of S2 and S3 recognition and tracking: By comparing the position difference of the target obtained by tracking in adjacent frames, calculate its moving direction L2, L2 = atan2(v_y, v_x). Where v_x and v_y are the x-component and y-component of the target's velocity vector, respectively. In said S6, , Where a and b are two vectors, v is their dot product, and d is the angle between them; S7. Calculate the angle θ between L1 and L2 and determine whether it is greater than a threshold C: Calculate the angle d between L1 and L2 using the vector angle formula, d = acos(cos(L1) * cos(L2) + sin(L1) * sin(L2)). If d is greater than the threshold C, it is determined that the target object has crossed the warning blind area. In S7, the direction vectors of L1 and L2 are expressed as: L1 = (k1, k2), L2 = (k3, k4), The angle d between them is calculated using the vector angle formula: d = arccos(cos(L1, L2)) = arccos( (k1 * k3 + k2 * k4) / (sqrt(k1^2 + k2^2) * sqrt(k3^2 + k4^2)) ); S8. If the target object crosses the warning blind area, determine whether it is moving away from or approaching the current warning area within area m: By comparing the position changes of the target in several consecutive frames, it can be judged whether it is approaching or moving away; If the target object approaches or moves away from the warning area and the threshold reaches the warning value, a warning signal is issued; S9. Result output: The system sends the detected target image and warning signal through the network so that timely response measures can be taken.

2. The AI ​​algorithm blind spot warning method according to claim 1, characterized in that: The AI ​​algorithm blind spot warning method is implemented through an AI algorithm blind spot warning system, wherein the AI ​​algorithm blind spot warning system comprises an AI algorithm camera (1) for monitoring the movement direction of a target object and determining whether the target object is in a defined area m, and a roadside warning light (2) for warning traffic participants, wherein the roadside warning light (2) is wirelessly connected or linearly connected to the AI ​​algorithm camera (1).

3. The AI ​​algorithm blind spot warning method according to claim 2, characterized in that: The AI ​​algorithm blind spot warning system further comprises an underground warning light (3) for warning traffic participants, wherein the underground warning light (3) is wirelessly or linearly connected to the AI ​​algorithm camera (1).

4. The AI ​​algorithm blind spot warning method according to claim 2, characterized in that: The AI ​​algorithm blind spot warning system also includes a display screen for prompting traffic participants and displaying text information, and the display screen is wirelessly or linearly connected to the AI ​​algorithm camera (1).

Images