Blind spot collision warning method and system

Abstract

The present application relates to the field of driving control technology, and discloses a blind spot collision warning method and system, which includes: taking images of the vehicle's blind spot through multiple cameras installed on the outside of the vehicle; synthesizing the images taken by multiple cameras to form a panoramic image of the vehicle's blind spot; analyzing the panoramic image of the vehicle's blind spot and identifying obstacles located in the vehicle's blind spot; detecting the properties of the obstacle, and calculating the possibility of the vehicle colliding with the obstacle based on the properties of the obstacle; and sending a warning signal to the driver of the vehicle based on the possibility of the vehicle colliding with the obstacle. According to the present invention, the images from multiple cameras are synthesized to form a panoramic image of the vehicle's surroundings, and the possibility of a collision between the vehicle and the obstacle is calculated based on the detection results of the obstacle's properties, showing great potential in improving driving safety and comfort.

Description

Technical Field

[0001] The present application relates to the field of driving control technology, and more specifically, to a blind spot collision warning method and system. Background Art

[0002] With the development of the automobile industry and the increasing complexity of urban traffic conditions, vehicle safety issues have received more and more attention. Especially during driving, due to the limitations of the vehicle structure, there are many areas that the driver cannot directly observe, which are called "blind spots". The existence of blind spots brings great safety hazards to drivers, especially during operations such as changing lanes, reversing and turning. Obstacles in the blind spots (such as pedestrians, bicycles or other vehicles) may cause serious traffic accidents.

[0003] In the existing technology, there are already some solutions that attempt to solve the problem of vehicle blind spots. For example: traditional rearview mirrors and wide-angle mirrors can help drivers expand their field of vision, but there are still visual blind spots; ultrasonic sensors detect obstacles behind the vehicle and sound an alarm, but this system can only detect obstacles within a certain distance and cannot provide specific obstacle information; some vehicles are equipped with a single rear camera to help the driver observe the environment behind the vehicle when reversing. Although this method can provide intuitive visual information, it has limited coverage and cannot provide a panoramic view. It can be seen that the above technologies have improved driving safety to a certain extent, but they still have shortcomings.

[0004] Therefore, there is an urgent need for a more comprehensive, accurate and real-time vehicle blind spot collision warning method to improve driving safety and reduce the occurrence of traffic accidents. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed to provide a blind spot collision warning method and system to achieve vehicle blind spot collision warning with wide coverage, comprehensive information, strong real-time performance and low false alarm rate.

[0006] In a first aspect, the present application provides a blind spot collision warning method, comprising: capturing images of a vehicle's blind spot through multiple cameras installed on the outside of the vehicle; synthesizing the images captured by the multiple cameras to form a panoramic image of the vehicle's blind spot; analyzing the panoramic image of the vehicle's blind spot and identifying obstacles located in the vehicle's blind spot; detecting properties of the obstacle and calculating the possibility of a collision between the vehicle and the obstacle based on the properties of the obstacle; and issuing a warning signal to the driver of the vehicle based on the possibility of a collision between the vehicle and the obstacle.

[0007] Optionally, the aforementioned blind spot collision warning method, synthesizing the images captured by the multiple cameras includes: detecting the speed and acceleration of the vehicle in the direction of the axes of the multiple cameras; calculating the mean speed, speed variance, acceleration mean, and acceleration variance of the vehicle in the direction of the axes of the multiple cameras; synthesizing the images captured by the multiple cameras into a panoramic image of the vehicle blind spot, wherein for any point in the panoramic image of the vehicle blind spot, The pixel value at ,in, is the number of the multiple cameras, is an image captured by the i-th camera among the multiple cameras, , is the velocity and acceleration in the axis direction of the i-th camera, , is the velocity and acceleration in the axis direction of the j-th camera, , , , are the velocity mean, velocity variance, acceleration mean, and acceleration variance in the axis directions of the multiple cameras.

[0008] Optionally, the aforementioned blind spot collision warning method analyzes the panoramic image of the vehicle blind spot and identifies obstacles located in the vehicle blind spot, including: extracting feature maps of the panoramic image of the vehicle blind spot at multiple scales through a preset convolutional neural network; fusing feature maps corresponding to different scales to obtain a multi-scale feature map, wherein the feature value of any point in the multi-scale feature map ,in, is the number of the multiple scales, , is a feature map of the panoramic image of the vehicle blind spot extracted at the kth and mth scales among the multiple scales, , is the deviation and peak value of the feature map of the panoramic image of the vehicle blind spot extracted at the k-th scale, , is the deviation and peak value of the feature map of the panoramic image of the vehicle blind spot extracted at the m-th scale, , is a preset weight coefficient; and according to the multi-scale feature map, obstacles in the blind spot of the vehicle are identified.

[0009] Optionally, the aforementioned blind spot collision warning method, before extracting the feature map of the panoramic image of the vehicle blind spot at multiple scales through a preset convolutional neural network, further includes: identifying the noise area of ​​the panoramic image of the vehicle blind spot; detecting the edge of the noise area; identifying the type of the noise area; when the noise area is a Gaussian noise type, using Gaussian filtering to process the non-edge part of the noise area; when the noise area is a pulse noise type, using median filtering to process the non-edge part of the noise area.

[0010] Optionally, the aforementioned blind spot collision warning method, before extracting the feature map of the panoramic image of the vehicle blind spot at multiple scales through a preset convolutional neural network, further includes: dividing the panoramic image of the vehicle blind spot into blocks; performing histogram equalization processing on each small block obtained by the block division; merging each small block that has undergone histogram equalization processing to regain the panoramic image; and adjusting the brightness distribution of the merged panoramic image according to a preset strategy.

[0011] Optionally, the aforementioned blind spot collision warning method adjusts the gamma value of the merged panoramic image according to a preset strategy to adjust the brightness distribution of the panoramic image of the vehicle blind spot, including: counting the brightness value of each point in the panoramic image of the vehicle blind spot; calculating the local contrast and local brightness of each point in the panoramic image of the vehicle blind spot according to the brightness value of each point in the panoramic image of the vehicle blind spot; calculating the average local contrast and average local brightness of the panoramic image of the vehicle blind spot; adjusting the brightness value of each point in the panoramic image of the vehicle blind spot, wherein the adjusted brightness value of the qth point , where C is the preset normalization factor, is the initial gamma value of the qth point, is the preset reference gamma value, , is the preset weight coefficient, , is the local contrast and local brightness of the qth point, , are the average local contrast and average local brightness of the panoramic image of the vehicle blind spot.

[0012] Optionally, the aforementioned blind spot collision warning method, detecting the properties of the obstacle, includes: detecting the speed of the obstacle; and calculating the possibility of the vehicle colliding with the obstacle according to the properties of the obstacle, including: obtaining the speed of the vehicle; detecting the distance between the vehicle and the obstacle; and calculating the possibility of the vehicle colliding with the obstacle. ,in is the speed of the vehicle, is the speed of the obstacle, is the distance between the vehicle and the obstacle, t is the known system reaction time of the vehicle, , is the preset weight coefficient.

[0013] Optionally, the aforementioned blind spot collision warning method issues a warning signal to the driver of the vehicle according to the possibility of the vehicle colliding with the obstacle, including: capturing an image of the driver through a camera inside the vehicle; analyzing the driver's concentration level based on the image of the driver; when the driver's concentration level exceeds a preset level, if the possibility of the vehicle colliding with the obstacle is higher than a first threshold, issuing the warning signal; when the driver's concentration level is lower than the preset level, if the possibility of the vehicle colliding with the obstacle is higher than a second threshold, issuing the warning signal, the second threshold being lower than the first threshold.

[0014] In the second aspect, the present application provides a blind spot collision warning system, comprising: a shooting module, which captures images of the vehicle's blind spot through multiple cameras installed on the outside of the vehicle; a synthesis module, which synthesizes the images captured by the multiple cameras to form a panoramic image of the vehicle's blind spot; an analysis module, which analyzes the panoramic image of the vehicle's blind spot and identifies obstacles located in the vehicle's blind spot; a collision calculation module, which detects the properties of the obstacle and calculates the possibility of the vehicle colliding with the obstacle based on the properties of the obstacle; and a warning module, which sends a warning signal to the driver of the vehicle based on the possibility of the vehicle colliding with the obstacle.

[0015] The above one or more technical solutions of the present application have at least one or more of the following beneficial effects:

[0016] According to the technical solution of the present invention, a blind spot collision warning method is proposed which not only solves many problems existing in traditional solutions, but also synthesizes images from multiple cameras to form a panoramic image of the vehicle's surroundings, and calculates the possibility of a collision between the vehicle and the obstacle based on the detection results of the obstacle's attributes (such as size, speed, etc.), showing great potential in improving driving safety and comfort. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0018] Figure 1 is a flow chart of a blind spot collision warning method according to an embodiment of the present application;

[0019] Figure 2 is another flow chart of a blind spot collision warning method according to an embodiment of the present application;

[0020] Figure 3 is another flow chart of a blind spot collision warning method according to an embodiment of the present application;

[0021] Figure 4 is a partial flow chart of a blind spot collision warning method according to an embodiment of the present application;

[0022] Figure 5 is another partial flow chart of a blind spot collision warning method according to an embodiment of the present application;

[0023] Figure 6 4 is a block diagram of a blind spot collision warning system according to an embodiment of the present application. DETAILED DESCRIPTION

[0024] Some embodiments of the present application are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0025] like Figure 1 As shown, an embodiment of the present application provides a blind spot collision warning method, comprising:

[0026] Step S110, images of the vehicle's blind spots are captured by multiple cameras installed outside the vehicle.

[0027] In this embodiment, by using multiple cameras installed on the outside of the vehicle, the environmental information around the vehicle can be captured from different angles, effectively expanding the monitoring range of the blind spot. Compared with the traditional single camera or rearview mirror, this method can provide a more comprehensive field of view and reduce visual blind spots.

[0028] Step S120, synthesizing the images captured by multiple cameras to form a panoramic image of the vehicle blind spot.

[0029] In this embodiment, the images from multiple cameras are synthesized to form a panoramic image of the vehicle's surroundings, which not only helps the driver obtain more intuitive visual information, but also provides a basis for subsequent image analysis, ensuring the accuracy of obstacle recognition.

[0030] Step S130 , analyzing the panoramic image of the vehicle blind spot and identifying obstacles in the vehicle blind spot.

[0031] In this embodiment, by deeply analyzing the formed panoramic image, obstacles in the vehicle blind spot can be accurately identified. Combined with advanced image recognition algorithms, this method can maintain a high recognition rate in various complex scenarios (such as at night and in bad weather conditions).

[0032] Step S140, detecting the properties of the obstacle, and calculating the possibility of collision between the vehicle and the obstacle according to the properties of the obstacle.

[0033] In this embodiment, based on the detection results of obstacle attributes (such as size, speed, etc.), the method can calculate the possibility of collision between the vehicle and the obstacle. This quantitative risk assessment method is more scientific and reasonable than the method that simply relies on distance judgment, and helps to take more effective risk avoidance measures.

[0034] Step S150, sending a warning signal to the driver of the vehicle according to the possibility of collision between the vehicle and the obstacle.

[0035] According to the technical solution of this embodiment, a blind spot collision warning method is proposed which not only solves many problems existing in traditional solutions, but also synthesizes images from multiple cameras to form a panoramic image of the vehicle's surroundings, and calculates the possibility of a collision between the vehicle and the obstacle based on the detection results of the obstacle's attributes (such as size, speed, etc.), showing great potential in improving driving safety and comfort.

[0036] like Figure 2 As shown, an embodiment of the present application provides another blind spot collision warning method. Compared with the above-mentioned embodiment, the blind spot collision warning method of this embodiment, step S120 includes:

[0037] Step S210, detecting the speed and acceleration of the vehicle in the directions of multiple camera axes.

[0038] In this embodiment, the vehicle speed and acceleration are taken into consideration on the basis of the traditional image synthesis, so that the synthesized panoramic image is more stable and can maintain good visual continuity and consistency even when the vehicle moves fast or accelerates / decelerates suddenly.

[0039] Step S220, calculating the mean velocity, velocity variance, acceleration mean, and acceleration variance of the vehicle in the directions of multiple camera axes.

[0040] In this embodiment, image fusion is performed by adjusting the contribution of each camera using the vehicle's speed and acceleration information, which can better adapt to scene changes under the vehicle's motion state and reduce image distortion caused by the vehicle's own movement, thereby improving the accuracy of obstacle recognition.

[0041] Step S230, combining the images captured by the multiple cameras into a panoramic image of the vehicle blind spot, wherein for any point in the panoramic image of the vehicle blind spot The pixel value at ,in, is the number of cameras, is the image captured by the i-th camera among multiple cameras, , is the velocity and acceleration in the axis direction of the i-th camera, , is the velocity and acceleration in the direction of the j-th camera axis, , , , It is the velocity mean, velocity variance, acceleration mean, and acceleration variance in the axis directions of multiple cameras.

[0042] According to the technical solution of this embodiment, taking into account the impact of the actual operation of the vehicle on image synthesis, a method of adjusting weights based on speed and acceleration is adopted, which can effectively reduce false detections caused by changes in the external environment, and also avoid the situation where certain important obstacles are ignored.

[0043] like Figure 3 As shown, an embodiment of the present application provides another blind spot collision warning method. Compared with the above-mentioned embodiment, the blind spot collision warning method of this embodiment, step S130 includes:

[0044] Step S310, extracting feature maps of the panoramic image of the vehicle blind spot at multiple scales through a preset convolutional neural network.

[0045] In this embodiment, by using a convolutional neural network (CNN) at multiple scales to extract feature maps of the panoramic image of the vehicle blind spot, detailed information at different scales can be captured. This helps the system better identify obstacles of various sizes and shapes, especially in complex scenes.

[0046] Step S320: The feature maps corresponding to different scales are fused to obtain a multi-scale feature map. The feature value of any point in the multi-scale feature map is ,in, is the number of multiple scales, , is the feature map of the panoramic image of the vehicle blind spot extracted at the kth and mth scales among multiple scales, , is the bias and peak of the feature map of the panoramic image of the vehicle blind spot extracted at the kth scale, , is the bias and peak of the feature map of the panoramic image of the vehicle blind spot extracted at the mth scale, , is the preset weight coefficient.

[0047] In this embodiment, a multi-scale feature map is generated by weighted fusion of feature maps of different scales. The adaptive weighting method in the formula ensures that the final feature map can fully utilize the advantages of each scale.

[0048] Step S330: identifying obstacles in the blind spot of the vehicle according to the multi-scale feature map.

[0049] According to the technical solution of this embodiment, multi-scale feature fusion helps to reduce false detection and missed detection caused by insufficient single-scale feature extraction, and improves the overall recognition performance of the system. In addition, the multi-scale feature extraction and fusion mechanism can better adapt to different lighting conditions, weather conditions and background interference, and improve the robustness of the system in various complex environments.

[0050] An embodiment of the present application provides another blind spot collision warning method. Compared with the above-mentioned embodiment, the blind spot collision warning method of this embodiment, before step S310, further includes:

[0051] (1) Identify the noisy areas in the panoramic image of the vehicle’s blind spot.

[0052] (2) Detect the edges of the noise area.

[0053] In this embodiment, the image obtained after noise processing has clearer boundaries, which helps the convolutional neural network (CNN) to more accurately extract features of the vehicle blind spot, especially for obstacles of small size or rich details.

[0054] (3) Identify the type of noise area.

[0055] (4) When the noise area is of Gaussian noise type, Gaussian filtering is used to process the non-edge part of the noise area.

[0056] (5) When the noise area is of impulse noise type, use median filtering to process the non-edge part of the noise area.

[0057] According to the technical solution of this embodiment, by identifying and processing the noise area in a targeted manner, the Gaussian noise and impulse noise in the image can be effectively reduced, thereby improving the overall image quality. High-quality input images are a prerequisite for accurate feature extraction. Appropriate noise processing as a preprocessing step can help speed up the subsequent feature extraction and analysis process and improve the response time of the entire system. Through selective noise processing, better image purification effects can be achieved without sacrificing too many computing resources, which is particularly important for the real-time requirements of vehicle-mounted systems.

[0058] like Figure 4 As shown, an embodiment of the present application provides another blind spot collision warning method. Compared with the above-mentioned embodiment, the blind spot collision warning method of this embodiment, before step S310, further includes:

[0059] Step S410, dividing the panoramic image of the vehicle blind spot into blocks.

[0060] Step S420: performing histogram equalization processing on each small block obtained by block division.

[0061] In this embodiment, by dividing the panoramic image of the vehicle blind spot into blocks and performing histogram equalization processing on each small block, the local contrast of the image can be significantly improved, which helps to highlight the details in the image, especially in the case of uneven lighting.

[0062] Step S430: merge each small block processed by histogram equalization to obtain a panoramic image again.

[0063] In this embodiment, the panoramic image is re-obtained by merging small blocks that have been processed by histogram equalization, thereby ensuring the overall consistency of the image and avoiding the boundary effect caused by local processing.

[0064] Step S440, adjusting the brightness distribution of the merged panoramic image according to a preset strategy, specifically includes:

[0065] (1) Count the brightness value of each point in the panoramic image of the vehicle's blind spot.

[0066] (2) According to the brightness value of each point in the panoramic image of the vehicle blind spot, the local contrast and local brightness of each point in the panoramic image of the vehicle blind spot are calculated.

[0067] In this embodiment, the brightness value is dynamically adjusted according to the local contrast and local brightness of each point. This adaptive method can better preserve the detail information of the image and enhance the visibility of the dark area.

[0068] (3) Calculate the average local contrast and average local brightness of the panoramic image of the vehicle’s blind spot.

[0069] (4) Adjust the brightness value of each point in the panoramic image of the vehicle blind spot, where the adjusted brightness value of the qth point is , where C is the preset normalization factor, is the initial gamma value of the qth point, is the preset reference gamma value, , is the preset weight coefficient, , is the local contrast and local brightness of the qth point, , are the average local contrast and average local brightness of the panoramic image of the vehicle blind spot.

[0070] In this embodiment, by calculating the average local contrast and the average local brightness and adjusting the formula parameters, overexposure or underexposure in the image can be effectively avoided, making the image more natural and clear.

[0071] According to the technical solution of this embodiment, the block histogram equalization processing and brightness distribution adjustment method proposed in this embodiment not only improves the quality of input data from the source, but also lays a solid foundation for subsequent feature extraction and obstacle recognition, and finally realizes a more efficient, accurate and reliable blind spot collision warning system. These improvements not only improve the overall performance of the system, but also significantly enhance the user experience and safety.

[0072] like Figure 5 As shown, an embodiment of the present application provides another blind spot collision warning method. Compared with the above-mentioned embodiment, the blind spot collision warning method of this embodiment, step S140 includes:

[0073] Step S510, detecting the speed of the obstacle.

[0074] Step S520, obtaining the speed of the vehicle.

[0075] Step S530, detecting the distance between the vehicle and the obstacle.

[0076] In this embodiment, by introducing the obstacle speed, the vehicle speed, and the distance between the vehicle and the obstacle, the relative motion state can be evaluated more accurately, thereby more accurately predicting the potential collision risk.

[0077] Step S540, calculating the possibility of collision between the vehicle and the obstacle ,in, is the speed of the vehicle, is the speed of the obstacle, is the distance between the vehicle and the obstacle, t is the known system reaction time of the vehicle, , is the preset weight coefficient.

[0078] According to the technical solution of this embodiment, the collision probability calculation can be completed in a very short time using a concise mathematical model, ensuring the real-time response capability of the system. Based on the rapid calculation of real-time data, the system can issue a warning signal in time before the potential danger is about to occur, providing the driver with sufficient reaction time.

[0079] An embodiment of the present application provides another blind spot collision warning method. Compared with the above-mentioned embodiment, the blind spot collision warning method of this embodiment, step S150 includes:

[0080] (1) The driver’s image is captured by a camera inside the vehicle.

[0081] (2) Analyze the driver's concentration level based on the driver's image.

[0082] In this embodiment, by analyzing the driver's concentration level, the system can dynamically adjust the threshold for issuing a warning signal, making the alarm more intelligent and personalized.

[0083] (3) When the driver's concentration exceeds a preset level and the probability of the vehicle colliding with an obstacle is higher than a first threshold, a warning signal is issued.

[0084] (4) When the driver's concentration level is lower than a preset level, a warning signal is issued if the possibility of the vehicle colliding with an obstacle is higher than a second threshold, and the second threshold is lower than the first threshold.

[0085] According to the technical solution of this embodiment, by using the vehicle's internal camera to capture and analyze the driver's image in real time, the system can quickly and accurately determine the driver's concentration level, thereby improving the system's response speed. Combined with the driver's status and external environment information, the system provides multi-dimensional safety protection and improves overall reliability and effectiveness.

[0086] like Figure 6 As shown, an embodiment of the present application provides a blind spot collision warning system, comprising:

[0087] The shooting module 610 shoots images of the vehicle's blind spots through multiple cameras installed on the outside of the vehicle.

[0088] In this embodiment, by using multiple cameras installed on the outside of the vehicle, the environmental information around the vehicle can be captured from different angles, effectively expanding the monitoring range of the blind spot. Compared with the traditional single camera or rearview mirror, this method can provide a more comprehensive field of view and reduce visual blind spots.

[0089] The synthesis module 620 synthesizes the images taken by multiple cameras to form a panoramic image of the vehicle's blind spot.

[0090] In this embodiment, the images from multiple cameras are synthesized to form a panoramic image of the vehicle's surroundings, which not only helps the driver obtain more intuitive visual information, but also provides a basis for subsequent image analysis, ensuring the accuracy of obstacle recognition.

[0091] The analysis module 630 analyzes the panoramic image of the vehicle's blind spot and identifies obstacles located in the vehicle's blind spot.

[0092] In this embodiment, by deeply analyzing the formed panoramic image, obstacles in the vehicle blind spot can be accurately identified. Combined with advanced image recognition algorithms, this method can maintain a high recognition rate in various complex scenarios (such as at night and in bad weather conditions).

[0093] The collision calculation module 640 detects the properties of the obstacle and calculates the possibility of the vehicle colliding with the obstacle according to the properties of the obstacle.

[0094] In this embodiment, based on the detection results of obstacle attributes (such as size, speed, etc.), the method can calculate the possibility of collision between the vehicle and the obstacle. This quantitative risk assessment method is more scientific and reasonable than the method that simply relies on distance judgment, and helps to take more effective risk avoidance measures.

[0095] The warning module 650 sends a warning signal to the driver of the vehicle according to the possibility of collision between the vehicle and the obstacle.

[0096] According to the technical solution of this embodiment, a blind spot collision warning method is proposed which not only solves many problems existing in traditional solutions, but also synthesizes images from multiple cameras to form a panoramic image of the vehicle's surroundings, and calculates the possibility of a collision between the vehicle and the obstacle based on the detection results of the obstacle's attributes (such as size, speed, etc.), showing great potential in improving driving safety and comfort.

[0097] The basic principles of the present application are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present application. In addition, the specific details disclosed above are only for the purpose of illustration and ease of understanding, not for limitation, and the above details do not limit the present application to being implemented by adopting the above specific details.

[0098] The block diagrams of the devices, apparatuses, equipment, and systems involved in this application are only illustrative examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagram. As will be appreciated by those skilled in the art, these devices, apparatuses, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including", "comprising", "having", etc. are open words, referring to "including but not limited to", and can be used interchangeably with them. The words "or" and "and" used here refer to the words "and / or" and can be used interchangeably with them, unless the context clearly indicates otherwise. The words "such as" used here refer to the phrase "such as but not limited to", and can be used interchangeably with them.

[0099] It should also be noted that in the apparatus, device and method of the present application, each component or each step can be decomposed and / or recombined. Such decomposition and / or recombination should be regarded as equivalent solutions of the present application.

[0100] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present application. Therefore, the present application is not intended to be limited to the aspects shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

[0101] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present application to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.

Claims

1. A blind spot collision warning method, characterized in that: include: images of the blind spots of the vehicle captured by multiple cameras installed outside the vehicle; Synthesizing the images captured by the multiple cameras to form a panoramic image of the vehicle blind spot; Analyzing the panoramic image of the vehicle blind spot and identifying obstacles located in the vehicle blind spot; Detecting properties of the obstacle, and calculating the possibility of collision between the vehicle and the obstacle according to the properties of the obstacle; According to the probability of the vehicle colliding with the obstacle, a warning signal is issued to the driver of the vehicle, The synthesizing of the images captured by the multiple cameras includes: Detecting the speed and acceleration of the vehicle in the directions of the multiple camera axes; Calculating a velocity mean, a velocity variance, an acceleration mean, and an acceleration variance of the vehicle in the directions of the plurality of camera axes; The images captured by the multiple cameras are combined into a panoramic image of the vehicle blind spot, wherein for any point in the panoramic image of the vehicle blind spot The pixel value at ,in, is the number of the multiple cameras, is an image captured by the i-th camera among the multiple cameras, , is the velocity and acceleration in the axis direction of the i-th camera, , is the velocity and acceleration in the direction of the j-th camera axis, , , , are the velocity mean, velocity variance, acceleration mean, and acceleration variance in the axis directions of the multiple cameras, The step of analyzing the panoramic image of the vehicle blind spot and identifying obstacles in the vehicle blind spot includes: Extracting feature maps of the panoramic image of the vehicle blind spot at multiple scales through a preset convolutional neural network; The feature maps corresponding to different scales are fused to obtain a multi-scale feature map. The feature value of any point in the multi-scale feature map is ,in, is the number of the multiple scales, , is a feature map of the panoramic image of the vehicle blind spot extracted at the kth and mth scales among the multiple scales, , is the deviation and peak value of the feature map of the panoramic image of the vehicle blind spot extracted at the k-th scale, , is the deviation and peak value of the feature map of the panoramic image of the vehicle blind spot extracted at the m-th scale, , is the preset weight coefficient; According to the multi-scale feature map, obstacles in the blind spot of the vehicle are identified, Wherein, detecting the attribute of the obstacle includes: detecting a speed of the obstacle; and Calculating the possibility of the vehicle colliding with the obstacle according to the property of the obstacle includes: obtaining the speed of the vehicle; detecting a distance between the vehicle and the obstacle; Calculate the probability of collision between the vehicle and the obstacle , in, is the speed of the vehicle, is the speed of the obstacle, is the distance between the vehicle and the obstacle, t is the known system reaction time of the vehicle, , is the preset weight coefficient.

2. The blind spot collision warning method according to claim 1, characterized in that: Before extracting the feature map of the panoramic image of the vehicle blind spot at multiple scales through a preset convolutional neural network, the method further includes: Identifying a noise region of the panoramic image of the vehicle blind spot; Detecting the edge of the noise area; identifying the type of the noise area; When the noise region is of Gaussian noise type, using Gaussian filtering to process the non-edge portion of the noise region; When the noise region is of impulse noise type, a median filter is used to process a non-edge portion of the noise region.

3. The blind spot collision warning method according to claim 1, characterized in that: Before extracting the feature map of the panoramic image of the vehicle blind spot at multiple scales through a preset convolutional neural network, the method further includes: Dividing the panoramic image of the vehicle blind spot into blocks; Perform histogram equalization on each small block obtained by block division; Merging each small block processed by histogram equalization to obtain the panoramic image again; The brightness distribution of the merged panoramic image is adjusted according to a preset strategy.

4. The blind spot collision warning method according to claim 3, characterized in that: The gamma value of the merged panoramic image is adjusted according to a preset strategy to adjust the brightness distribution of the panoramic image of the vehicle blind spot, including: Counting the brightness value of each point in the panoramic image of the vehicle blind spot; Calculating the local contrast and local brightness of each point in the panoramic image of the vehicle blind spot according to the brightness value of each point in the panoramic image of the vehicle blind spot; Calculating an average local contrast and an average local brightness of the panoramic image of the vehicle blind spot; The brightness value of each point in the panoramic image of the vehicle blind spot is adjusted, wherein the brightness value of the qth point after adjustment is , where C is the preset normalization factor, is the initial gamma value of the qth point, is the preset reference gamma value, , is the preset weight coefficient, , is the local contrast and local brightness of the qth point, , are the average local contrast and average local brightness of the panoramic image of the vehicle blind spot.

5. The blind spot collision warning method according to claim 1, characterized in that: According to the probability of the vehicle colliding with the obstacle, a warning signal is issued to the driver of the vehicle, including: capturing an image of the driver by a camera inside the vehicle; analyzing the driver's concentration level based on the image of the driver; When the driver's concentration level exceeds a preset level, if the possibility of the vehicle colliding with the obstacle is higher than a first threshold, issuing the warning signal; When the driver's concentration level is lower than the preset level, if the possibility of the vehicle colliding with the obstacle is higher than a second threshold, the warning signal is issued, and the second threshold is lower than the first threshold.

6. A blind spot collision warning system, the blind spot collision warning system implementing the blind spot collision warning method according to any one of claims 1 to 5, characterized in that: include: A shooting module, which shoots images of the blind spots of the vehicle through multiple cameras installed outside the vehicle; A synthesis module synthesizes the images taken by the multiple cameras to form a panoramic image of the vehicle blind spot; An analysis module is used to analyze the panoramic image of the blind spot of the vehicle and identify obstacles located in the blind spot of the vehicle; A collision calculation module, detecting properties of the obstacle and calculating the possibility of collision between the vehicle and the obstacle according to the properties of the obstacle; The warning module sends a warning signal to the driver of the vehicle according to the possibility of the vehicle colliding with the obstacle.

Images