Collision avoidance system based on machine body vision

Abstract

This invention discloses a collision avoidance system based on machine vehicle vision, including an information acquisition unit, a data analysis unit, and an information output unit. This invention relates to the field of vehicle collision avoidance technology, solving the technical problem of collisions with fixed obstacles during reversing in underground parking garages. This invention segments the acquired images and radar radiation range to measure the distance between the vehicle body and the obstacle, comparing it with the actual measured distance to determine if the radar has made a misjudgment. For misjudgments, a warning signal is generated and transmitted to remind the operator to check. Secondly, the system uses the acquired image information for modeling and analyzes the modeled data to generate a virtual parking space. The system then adjusts the vehicle's position according to the virtual parking space to avoid collisions with obstacles on either side of the parking space during reversing.

Description

Technical Field

[0001] This invention relates to the field of vehicle collision avoidance technology, specifically to a collision avoidance system based on machine vision. Background Technology

[0002] Automotive collision avoidance systems are mainly used to assist drivers in avoiding high-speed and low-speed rear-end collisions, unintentional lane departures at high speeds, and collisions with pedestrians or other vehicles, which are major traffic accidents. At the same time, automotive collision avoidance systems can also be applied to reversing and parking vehicles, as bumps and collisions during the reversing process are difficult to avoid.

[0003] According to patent application number CN201910206570.9, the patent includes a buffer device and a collision monitoring module. The buffer device includes a front crossbeam, an energy-absorbing box, a reinforcing rod, and an airbag assembly. There are two energy-absorbing boxes and two reinforcing rods. One end of each energy-absorbing box is connected to the vehicle's longitudinal beam, and the other end is connected to the front crossbeam. One end of each reinforcing rod is connected to the vehicle's longitudinal beam, and the other end is connected to the middle of the front crossbeam. The patent uses a collision monitoring module to monitor collisions and thereby activate other components to buffer and protect the driver.

[0004] Some vehicles may encounter collisions while reversing in underground parking garages due to varying garage layouts and the presence of support pillars in front of parking spaces. Additionally, some cars equipped with automatic parking systems may experience collisions due to radar detection errors. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a collision avoidance system based on machine vehicle vision, which solves the problem of collisions with fixed obstacles when reversing in underground parking garages.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a collision avoidance system based on machine vehicle vision, comprising:

[0007] The information acquisition unit is used to collect basic data of the target object, which is a vehicle that is reversing. The basic data includes the environmental conditions around the vehicle during parking, the longitudinal length of the parking space, the lateral length of the parking space, the radar monitoring range, and the radar monitoring duration. The environmental conditions around the vehicle are obtained by real-time monitoring through the all-round monitoring image installed on the vehicle and generating corresponding image information.

[0008] The longitudinal and lateral lengths of the parking space are obtained by acquiring big data. In this application, a small car parking space is used as an example, with the dimensions of 2.5~2.7×5~6 meters.

[0009] Method for obtaining radar monitoring duration: The radar monitoring signal is whether the system receives the signal transmitted back from the radar, and the acquired basic data is transmitted to the image segmentation unit;

[0010] The image segmentation unit acquires the transmitted basic data and segments it. Specifically, the acquired image is divided into nine equal parts and labeled as Ci. Similarly, the radar radiation range is divided into nine equal parts and labeled as Di. Ci and Di correspond to each other, i=1, ...,9. The segmented data is then transmitted to the data analysis unit, where the image is processed based on the monitoring signal information acquired by the radar, and the acquired image and the monitoring signal range are processed to the same size.

[0011] The data analysis unit acquires the transmitted segmented data and analyzes the image information and radar reception information within the corresponding areas. The specific analysis methods are as follows:

[0012] The specific methods for image information analysis are as follows:

[0013] Acquire the duration of radar signal reception within the corresponding area and label it as T. D j, where j=1, ...,9, and areas with the same reception time are marked as identified areas, while areas with different reception times are marked as unidentified areas. No processing is performed on unidentified areas. It is assumed that all obstacles exist within identified areas, and there are no obstacles within unidentified areas. The width of obstacles is calculated in the following way:

[0014] Obtain the width of a single evenly divided image and the number of images containing obstacles, and perform 1:10 modeling of the obstacle dimensions through 3D modeling, and calculate the width LZ of the obstacle;

[0015] The distance to the vehicle obstacle is calculated by acquiring the reception duration. The calculated distance is marked as P1 and P2. The radar distance measurement here is a relatively mature calculation method in the existing technology, so it will not be elaborated on further.

[0016] The recognition areas are merged, and information data within the merged area is obtained. This information data includes the width LZ of the obstacle, and the distance between the vehicle and the obstacle is calculated. The specific calculation method is as follows:

[0017] Establish a Cartesian coordinate system with the obstacle as the origin, the vehicle's position as the X-axis, and the obstacle's height as the Y-axis. Label the vehicle A(X1, Y1), where X1 represents the distance from this point to the origin, i.e., the distance between the vehicle and the obstacle.

[0018] The distance is compared with the calculated distance L1. When L1=X1, the system determines that the distance calculated by the radar is the same as the actual measured distance and does not perform any processing. Otherwise, the system determines that the distance calculated by the radar is different from the actual measured distance and generates an abnormal signal.

[0019] Data analysis involves analyzing the acquired abnormal transmission signals. The specific analysis methods are as follows:

[0020] Q1: Obtain the longitudinal length Z and lateral length H of the parking space, and establish a rectangular coordinate system with the lateral length H as the X-axis and the longitudinal length Z as the Y-axis;

[0021] Q2: Obtain the distances P1 and P2 to obstacles on both sides of the vehicle, and compare P1 and P2.

[0022] If P1-P2≤10, the system determines that the distance between the two sides does not exceed the preset range and the system does not take any action. Otherwise, the system determines that the distance between the two sides exceeds the preset range, generates an alarm signal, and transmits the signal to the distance adjustment unit.

[0023] The distance adjustment unit analyzes the acquired and transmitted alarm signals, and the specific analysis method is as follows:

[0024] W1: Obtain the overall width of the target object, and simultaneously obtain the distances P1 and P2 on both sides of the target object, and calculate and generate virtual parking spaces. The specific method for generating virtual parking spaces is as follows:

[0025] W11: Obtain the obstacle length LZ and the preset range value, and input them into the calculation formula to calculate the maximum capacity value LZ+10;

[0026] W12: Substitute the obtained maximum capacity value LZ+10 and the lateral length H into the calculation formula to obtain the lateral length HX of the virtual parking space: ;

[0027] W2: Substitute HX into the calculation formula to obtain the safe range value I: It generates a safety range signal and transmits the adjustment signal to the information output unit.

[0028] As a further aspect of the present invention: the signal warning unit and the data analysis unit are bidirectionally electrically connected, while the image segmentation unit is unidirectionally electrically connected to the radar monitoring unit and the data analysis unit.

[0029] As a further aspect of the present invention: the acquired image and radar radiation range are segmented to measure the distance between the vehicle body and the obstacle, and compared with the actual measured distance to determine whether the radar has made a misjudgment. If a misjudgment occurs, a warning signal is generated and transmitted to remind the operator to check.

[0030] Beneficial effects

[0031] This invention provides a collision avoidance system based on machine vehicle vision. Compared with existing technologies, it has the following advantages:

[0032] This invention is achieved through:

[0033] 1. The acquired images and radar radiation range are segmented to measure the distance between the vehicle body and obstacles, and compared with the actual measured distance to determine whether the radar has made a false judgment. If a false judgment is made, a warning signal is generated and transmitted to remind the operator to check.

[0034] 2. Next, the information obtained from the images is used for modeling and processing. The data from the modeling is then analyzed to generate virtual parking spaces. The virtual parking spaces are then used to adjust the positions accordingly, thereby avoiding the impact of obstacles on the sides of the parking space on the vehicle body during reversing. Attached Figure Description

[0035] Figure 1 This is a system block diagram of the present invention. Detailed Implementation

[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] Please see Figure 1 This application provides a collision avoidance system based on machine vehicle vision, including:

[0038] The information acquisition unit is used to collect basic data of the target object, which is a vehicle that is reversing. The basic data includes the environmental conditions around the vehicle during parking, the longitudinal length of the parking space, the lateral length of the parking space, the radar monitoring range, and the radar monitoring duration. The environmental conditions around the vehicle are obtained by real-time monitoring through the all-round monitoring image installed on the vehicle and generating corresponding image information.

[0039] The longitudinal and lateral lengths of the parking space are obtained by acquiring big data. In this application, a small car parking space is used as an example, with the dimensions of 2.5~2.7×5~6 meters.

[0040] Method for obtaining radar monitoring duration: The radar monitoring signal is whether the system receives the signal transmitted back from the radar, and the acquired basic data is transmitted to the image segmentation unit;

[0041] The image segmentation unit acquires the transmitted basic data and segments the basic data. The specific segmentation method is as follows: the acquired image is divided into nine equal parts and labeled as Ci. Similarly, the radar radiation range is divided into nine equal parts and labeled as Di. Ci and Di correspond to each other, i=1, ...,9. The segmented data is then transmitted to the data analysis unit.

[0042] The data analysis unit acquires the transmitted segmented data and analyzes the image information and radar reception information within the corresponding areas. The specific analysis methods are as follows:

[0043] The specific methods for image information analysis are as follows:

[0044] Acquire the duration of radar signal reception within the corresponding area and label it as T. D j, where j=1, ...,9, and areas with the same reception time are marked as identified areas, while areas with different reception times are marked as unidentified areas. No processing is performed on unidentified areas. It is assumed that all obstacles exist within identified areas, and there are no obstacles within unidentified areas. The width of obstacles is calculated in the following way:

[0045] Obtain the width of a single evenly divided image and the number of images containing obstacles, and perform 1:10 modeling of the obstacle dimensions through 3D modeling, and calculate the width LZ of the obstacle;

[0046] The distance to the vehicle obstacle is calculated by acquiring the reception duration. The calculated distance is marked as P1 and P2. The radar distance measurement here is a relatively mature calculation method in the existing technology, so it will not be elaborated on further.

[0047] The recognition areas are merged, and information data within the merged area is obtained. This information data includes the width LZ of the obstacle, and the distance between the vehicle and the obstacle is calculated. The specific calculation method is as follows:

[0048] A rectangular coordinate system is established with the obstacle as the origin, the vehicle's position as the X-axis, and the obstacle's height as the Y-axis. The vehicle is marked A(X1, Y1), where X1 represents the distance from the origin, i.e., the distance between the vehicle and the obstacle. This distance is compared with the measured distance L1. When L1 = X1, the system determines that the distance calculated by the radar is the same as the actual measured distance and does not perform any processing. Otherwise, the system determines that the distance calculated by the radar is different from the actual measured distance and generates an abnormal signal.

[0049] Data analysis involves analyzing the acquired abnormal transmission signals. The specific analysis methods are as follows:

[0050] Q1: Obtain the longitudinal length Z and lateral length H of the parking space, and establish a rectangular coordinate system with the lateral length H as the X-axis and the longitudinal length Z as the Y-axis;

[0051] Q2: Obtain the distances P1 and P2 to obstacles on both sides of the vehicle, and compare P1 and P2.

[0052] If P1-P2≤10, the system determines that the distance between the two sides does not exceed the preset range and the system does not take any action. Otherwise, the system determines that the distance between the two sides exceeds the preset range, generates an alarm signal, and transmits the signal to the distance adjustment unit.

[0053] The distance adjustment unit analyzes the acquired and transmitted alarm signals, and the specific analysis method is as follows:

[0054] W1: Obtain the overall width of the target object, and simultaneously obtain the distances P1 and P2 on both sides of the target object, and calculate and generate virtual parking spaces. The specific method for generating virtual parking spaces is as follows:

[0055] W11: Obtain the obstacle length LZ and the preset range value, and input them into the calculation formula to calculate the maximum capacity value LZ+10;

[0056] W12: Substitute the obtained maximum capacity value LZ+10 and the lateral length H into the calculation formula to obtain the lateral length HX of the virtual parking space: ;

[0057] W2: Substitute HX into the calculation formula to obtain the safe range value I: It generates a safety range signal and transmits the adjustment signal to the information output unit.

[0058] The information output unit receives the adjustment signal and displays the adjustment information directly to the operator through the display device.

[0059] The signal warning unit and the data analysis unit are connected bidirectionally, while the image segmentation unit is connected unidirectionally to the radar monitoring unit and the data analysis unit.

[0060] The acquired images and radar radiation range are segmented to measure the distance between the vehicle body and obstacles. The distance is then compared with the actual measured distance to determine if the radar has made a false judgment. If a false judgment is made, a warning signal is generated and transmitted to remind the operator to check.

[0061] Secondly, the information from the acquired images is used for modeling and processing. The data from the modeling is then analyzed to generate virtual parking spaces. Based on these virtual parking spaces, the corresponding positions are adjusted to avoid collisions with the vehicle body caused by obstacles on both sides of the parking space during reversing.

[0062] The working principle of this invention is as follows: The information acquisition unit first acquires the basic data of the target object and provides feedback based on the radar signal reception time. Simultaneously, it generates a monitoring image and transmits it to the image segmentation unit. The image segmentation unit divides the image into nine equal parts, and similarly processes the radar reflected signal. Based on the correspondence between the segmented image information and the radar reflected signal, the image is segmented into recognized and unrecognized areas according to different reception times. The unrecognized areas are not processed. Next, the recognized areas are processed by merging them and analyzing the internal information. A corresponding model is generated through 3D modeling. A Cartesian coordinate system is established based on the generated model, and the distance between the obstacle and the vehicle body is calculated. It is then determined whether the distance measured by the radar matches the actual distance. If they do not match, the smallest value is used as the standard in subsequent calculations. A Cartesian coordinate system is then established, and a virtual parking space is generated based on the calculated distance value. The maximum capacity is calculated, and adjustments are made accordingly to avoid collisions during reversing.

[0063] Furthermore, any content not described in detail in this specification is existing technology known to those skilled in the art.

[0064] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A collision avoidance system based on machine vehicle vision, characterized in that, include: The information acquisition unit is used to collect basic data of the target object. The basic data includes the environmental conditions around the vehicle during parking, the longitudinal length of the parking space, the lateral length of the parking space, the radar monitoring range, and the radar monitoring duration. The acquired basic data is then transmitted to the image segmentation unit. The image segmentation unit acquires the basic data transmitted, divides the acquired image into nine equal parts and labels them as Ci, and similarly divides the radar radiation range into nine equal parts and labels them as Di, with Ci and Di corresponding to each other, i=1, ...,9, and transmits the segmented data to the data analysis unit. The data analysis unit acquires the transmitted segmented data, obtains the duration of radar signal reception in the corresponding area, and marks it as T. D j, where j=1, ...,9, and areas with the same reception time are marked as identified areas, while areas with different reception times are marked as unidentified areas. No processing is performed on unidentified areas, and the width LZ of the obstacle is calculated. The identification areas are merged to obtain a merged area. Information data within the merged area is obtained and the vehicle obstacle distance L1 is calculated. When L1=X1, the system determines that the distance calculated by the radar is the same as the actual measured distance and does not perform any processing. X1 is a preset value. Otherwise, the system determines that the distance calculated by the radar is different from the actual measured distance and generates an abnormal signal. At the same time, the abnormal signal is transmitted to the information output unit. The system analyzes the acquired abnormal signals and calculates the distances P1 and P2 between the vehicle and the obstacles on both sides of the vehicle. If P1-P2≤10, the system determines that the distance between the two sides does not exceed the preset range and does not perform any processing. Otherwise, the system determines that the distance between the two sides exceeds the preset range, generates an alarm signal, and transmits the signal to the distance adjustment unit. The distance adjustment unit receives the transmitted alarm signal, performs distance analysis, calculates and generates a virtual parking space, adjusts the vehicle according to the safe range value of the virtual parking space, and generates a corresponding adjustment signal to transmit to the information output unit. The width of the obstacle is calculated in the following manner: Obtain the width of a single evenly divided image and the number of images containing obstacles, and perform 1:10 modeling of the obstacle dimensions through 3D modeling, and calculate the width LZ of the obstacle; The reception duration is used to calculate the distance to the vehicle obstacle, and the distances to both sides of the vehicle are calculated and marked as P1 and P2. The recognition areas are merged, and information data within the merged area is obtained. This information data includes the width LZ of the obstacle, and the distance between the vehicle and the obstacle is calculated. The specific calculation method is as follows: Establish a rectangular coordinate system with the obstacle as the origin, the vehicle's position as the X-axis, and the obstacle's height as the Y-axis. Mark the vehicle as A(X1, Y1), where X1 represents the distance from the origin, i.e., the distance between the vehicle and the obstacle, and compare it with the measured distance L1. Q1: Obtain the longitudinal length Z and lateral length H of the parking space, and establish a rectangular coordinate system with the lateral length H as the X-axis and the longitudinal length Z as the Y-axis; Q2: Obtain the distances P1 and P2 between obstacles on both sides of the vehicle body, and compare P1 and P2; The distance adjustment unit analyzes the alarm signal. W1: It obtains the overall width of the target object, and simultaneously obtains the distances P1 and P2 on both sides of the target object, and calculates and generates virtual parking spaces. The specific method for generating virtual parking spaces is as follows: W11: Obtain the obstacle length LZ and the preset range value, and input them into the calculation formula to calculate the maximum capacity value LZ+10; W12: Substitute the obtained maximum capacity value LZ+10 and the lateral length H into the calculation formula to obtain the lateral length HX of the virtual parking space: ; W2: Substitute HX into the calculation formula to obtain the safe range value I: It generates a safety range signal and transmits the adjustment signal to the information output unit. The information output unit receives the adjustment signal and feeds it back directly to the operator through the display device.

2. The collision avoidance system based on machine vehicle vision according to claim 1, characterized in that, The information output unit acquires the adjustment information transmitted by the distance adjustment unit and displays the adjustment information directly to the operator through the display device.

Images