Method and system for blind zone braking of commercial vehicles

Abstract

The present application belongs to the technical field of vehicle blind area braking, and discloses a method and system for blind area braking of a commercial vehicle. The method comprises the following steps: acquiring running data of the commercial vehicle to determine a predicted turning blind area and a rear blind area, detecting ground features in the turning blind area and predicting a turning probability of a target object; for the target object in the turning blind area, analyzing its avoidance possibility in combination with the ground features to reduce a false alarm rate; for the target object in the rear blind area, predicting its intrusion risk based on path constraints and speed trend to reduce a missed alarm rate. When the turning probability exceeds a preset threshold, a hierarchical alarm or braking is triggered, which not only avoids misjudgment caused by environmental interference or active avoidance of the target, but also ensures accurate capture of hidden risks, thereby realizing high-precision, low-missed-alarm and active safety protection in a complex traffic scene.

Description

Technical Field

[0001] This invention relates to the field of vehicle blind spot braking technology, specifically to methods, systems, storage media, and electronic terminals and systems for blind spot braking of commercial vehicles. Background Technology

[0002] Commercial vehicles, including freight trucks, buses, and other types of operator vehicles, play an irreplaceable role in economic activities and public transportation. However, due to their high and wide cab design, drivers have large blind spots, with the right-side blind spot being the most significant. This poses a serious threat to road traffic safety and poses a great danger to the lives and property of pedestrians and other road users.

[0003] Existing technologies primarily employ two approaches: rule-based constraints and auxiliary devices. Regarding rule-based constraints, some traffic management departments stipulate that drivers of commercial vehicles must stop for a period of time when making a right turn. The aim is to allow drivers more time to check the right-side blind spot and reduce the likelihood of accidents caused by blind spots. However, in practice, this significantly reduces road efficiency, easily causing traffic congestion. Furthermore, different drivers vary greatly in their adherence to this regulation, failing to effectively prevent right-side blind spot accidents at their root.

[0004] Regarding blind spot assist devices, some commercial vehicles are equipped with blind spot cameras, which use cameras to obtain blind spot vision and provide feedback to the driver to help them understand the surrounding situation of the commercial vehicle. However, in adverse weather conditions, such as heavy rain, heavy snow, and dense fog, the camera lens is easily covered by rain and snow, or the image becomes blurred due to light refraction and scattering, making it impossible to effectively obtain blind spot vision. In scenarios with sudden changes in light, it is difficult to accurately identify the movement trajectory, speed, and position information of target objects in the blind spot, failing to provide reliable decision support for the driver, and even more so, failing to accurately predict the risk of collision between the commercial vehicle and target objects, or to trigger braking in time in emergency situations to ensure the driving safety of the commercial vehicle.

[0005] Therefore, how to address the deficiencies and shortcomings of existing technologies through effective blind spot braking methods for commercial vehicles has become an important issue that researchers in this field urgently need to solve. Summary of the Invention

[0006] The purpose of this invention is to address the above-mentioned problems by providing a method, system, storage medium, and electronic terminal for blind spot braking of commercial vehicles.

[0007] The technical solution of the present invention is as follows: a method for blind spot braking of a commercial vehicle, comprising the following steps: acquiring first operating data of the commercial vehicle, determining a predicted steering blind spot and a direct rear blind spot of the commercial vehicle based at least on the first operating data, and detecting a first ground feature of a first target area in the predicted steering blind spot; acquiring second operating data of a first target object located within the predicted steering blind spot, and predicting a first change-of-direction probability of the first target object moving away from the predicted steering blind spot based on the second operating data and the first ground feature; acquiring third operating data of a second target object located in the direct rear blind spot, determining a second target area and its second ground feature based on the third operating data, and predicting a second change-of-direction probability of the second target object entering the predicted steering blind spot; if the first change-of-direction probability is lower than a first preset value and / or the second change-of-direction probability is higher, then outputting an alarm signal.

[0008] As an improvement to an embodiment of the present invention, it specifically includes: based on the second running data and the first ground features, predicting the first change-of-direction probability of the first target object moving away from the predicted turning blind zone using a pre-trained first model.

[0009] As an improvement to an embodiment of the present invention, it specifically includes: determining a second target area and its second ground features based on the third operating data; and determining the second change-of-direction probability of the second target object entering the predicted turning blind zone based on the second target area, its second ground features, and the speed change of the second target object.

[0010] As an improvement to an embodiment of the present invention, it specifically includes: acquiring road surface images of the predicted steering blind spot using an image sensor, and identifying abnormal road surface features using a deep learning model, wherein the abnormal features include color differences and / or local texture abrupt changes.

[0011] As an improvement to an embodiment of the present invention, it specifically includes: the second running data includes the movement trajectory and speed of the first target object; the change-of-direction probability is analyzed based on the second running data and the first ground features; if the probability is lower than a preset threshold, the first change-of-direction probability is determined to be low.

[0012] As an improvement to an embodiment of the present invention, it specifically includes: in response to detecting that the second ground feature is an obstacle for the second target object to enter the target area, and the speed change of the second target object meets a first preset condition, it is determined that the second change of direction probability is high, wherein the first preset condition is that the speed of the second target object decreases by more than a first threshold and / or the deceleration of the second target object exceeds a second threshold.

[0013] As an improvement to an embodiment of the present invention, the alarm signal specifically includes: triggering an audible and visual warning to prompt the driver and / or sending a graded braking command to the electronic braking device.

[0014] To achieve one of the aforementioned objectives, one embodiment of the present invention provides a blind spot braking system for commercial vehicles, comprising: a first detection module, configured to acquire first operating data of the commercial vehicle, determine a predicted steering blind spot and a direct rear blind spot of the commercial vehicle based at least on the first operating data, and detect a first ground feature of a first target area within the predicted steering blind spot; a first prediction module, configured to acquire second operating data of a first target object located within the predicted steering blind spot, and predict a first change-of-direction probability of the first target object moving away from the predicted steering blind spot based on the second operating data and the first ground feature; a second prediction module, configured to acquire third operating data of a second target object located in the direct rear blind spot, determine a second target area and its second ground feature based on the third operating data, and predict a second change-of-direction probability of the second target object entering the predicted steering blind spot; and a warning module, configured to output an alarm signal if the first change-of-direction probability is lower than a first preset value and / or the second change-of-direction probability is higher.

[0015] To achieve one of the above-mentioned objectives, one embodiment of the present invention provides a storage medium storing program instructions, which, when executed, implement the blind spot braking method for commercial vehicles as described in any of the preceding claims.

[0016] To achieve one of the above-mentioned objectives, one embodiment of the present invention provides an electronic terminal, including a processor and a memory, wherein the memory stores program instructions, and the processor executes the program instructions to implement the blind spot braking method for commercial vehicles as described in any of the preceding claims.

[0017] The method, system, storage medium, and electronic terminal for blind spot braking of commercial vehicles provided in this invention have the following advantages: This invention predicts the probability of a target object changing direction within the turning blind spot by combining ground features with analysis of its change-of-direction probability. An alarm is triggered only when the target object cannot be avoided, effectively reducing the false alarm rate. Furthermore, for target objects in the blind spot directly behind, the possibility of them entering the turning blind spot is predicted by detecting their path constraints, promptly identifying potential risks and significantly reducing the missed alarm rate. This avoids misjudgments caused by environmental interference or the target's active avoidance, while ensuring accurate capture of hidden risks, thereby achieving high-precision, low-false-and-missed-alarm active safety protection in complex traffic scenarios. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the blind spot braking method for commercial vehicles described in this invention.

[0019] Figure 2This is a schematic diagram of the location of the first target area of ​​the present invention.

[0020] Figure 3 This is a schematic diagram of the location of the second target area in this invention.

[0021] Figure 4 This is a schematic diagram of the blind spot braking system for commercial vehicles described in this invention.

[0022] Figure 5 This is a schematic diagram of the electronic terminal structure described in this invention. Detailed Implementation

[0023] The present invention will now be described in detail with reference to the specific embodiments shown in the accompanying drawings. However, these embodiments do not limit the present invention, and any structural, methodological, or functional modifications made by those skilled in the art based on these embodiments are included within the scope of protection of the present invention.

[0024] If the present invention involves orientation (e.g., up, down, left, right, front, back, outside, inside, etc.) when described, then the orientations involved need to be defined.

[0025] The scope of the embodiments described herein includes the entire scope of the claims and all available equivalents thereof. Throughout this document, the terms “first,” “second,” etc., are used only to distinguish one element from another without requiring or implying any actual relationship or order between the elements. Indeed, a first element can also be referred to as a second element, and vice versa. Furthermore, the terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a structure, apparatus, or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a structure, apparatus, or device. Without further limitations, an element defined by the phrase “comprising one…” does not exclude the presence of other identical elements in the structure, apparatus, or device that includes said element. The various embodiments described herein are presented in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably.

[0026] The terms "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer" used in this document to indicate orientation or positional relationships are based on the orientation or positional relationships shown in the accompanying drawings and are used only for the convenience of describing this document and simplifying the description. They do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as limiting the invention. In the description herein, unless otherwise specified and limited, the terms "installed," "connected," and "linked" should be interpreted broadly. For example, they can refer to mechanical or electrical connections, or internal connections between two elements, or direct connections or indirect connections through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms according to the specific circumstances.

[0027] This invention provides a method for blind spot braking of commercial vehicles, such as... Figure 1 As shown, the process includes the following steps: Step 101: Obtain the first operating data of the commercial vehicle, determine the predicted steering blind spot and the rear blind spot of the commercial vehicle based at least on the first operating data, and detect the first ground feature of the first target area in the predicted steering blind spot.

[0028] In practice, the first step is to acquire the initial operational data of the commercial vehicle. This commercial vehicle is equipped with an inertial measurement unit, an image sensor, and a millimeter-wave radar. The initial operational data includes the triaxial acceleration data collected by the inertial measurement unit. and triaxial angular velocity The system includes the driving trajectory of the commercial vehicle, environmental image information collected by the image sensor, and distance, speed, and azimuth information of the target object detected by the millimeter-wave radar. Simultaneously, the CenterFusion model is used to analyze the information collected by the image sensor and detected by the millimeter-wave radar to obtain the target object's trajectory and second speed. and the three-dimensional position information of the target object .

[0029] The inertial measurement unit's sampling frequency can be set to 100Hz, enabling it to acquire the commercial vehicle's three-axis acceleration in real time at a high frequency. and triaxial angular velocity Understandably, this ensures timely capture of vehicle dynamics. The three-axis acceleration measurement range is ±16g, meeting the acceleration measurement needs of commercial vehicles under various driving conditions, accurately measuring even minor acceleration changes during rapid acceleration, emergency braking, or normal driving. The three-axis angular velocity measurement range is ±2000° / s, accurately measuring vehicle rotation, such as angular velocity changes during cornering. Data output undergoes Kalman filtering for noise reduction. Kalman filtering is a highly efficient recursive filter that effectively removes noise interference from the inertial measurement unit's data based on a series of noisy measurement data, ensuring the stability of pose calculation. The image sensor resolution can be 1920×1080 pixels or higher; this embodiment is not limited, providing clear image information and ensuring accurate identification of the surrounding environment. The frame rate is 30fps or higher, ensuring continuous image acquisition without image stuttering or information loss.

[0030] For example, the camera is installed on the right front side of the vehicle and the rearview mirror, which can cover the right blind spot view. The wide-angle lens captures the blind spot environment image, expanding the range of image acquisition, so that the target object and environmental features in the right blind spot can be effectively captured. It can be understood that a camera is also set on the left front side of the vehicle, so that the target object and environmental features in the right blind spot can be effectively captured. This embodiment will not be described in detail.

[0031] Millimeter-wave radar, operating in the 77GHz band, offers higher resolution and accuracy. With a detection range of 200m, it can detect targets at greater distances, providing sufficient time for subsequent judgment and decision-making. Its speed detection accuracy is ±0.1m / s, accurately measuring the speed of moving targets. With a horizontal field of view of ±60° and a vertical field of view of ±15°, it can comprehensively detect the distance, speed, and azimuth of targets within a certain horizontal and vertical range. Based on this initial operational data, the predicted steering blind spot and rear blind spot of commercial vehicles are determined.

[0032] Specifically, the real-time pose of the commercial vehicle is calculated by integrating the triaxial acceleration α and triaxial angular velocity ω obtained by the inertial measurement unit, and the driver's steering intention is determined by solving the real-time pose of the commercial vehicle. The predicted steering blind spot is determined by combining the vehicle's steering trajectory. At the same time, the rear blind spot is determined based on the first operating data, the vehicle body structure parameters, and the sensor installation position.

[0033] Initialize the initial pose parameters of the commercial vehicle ;in The initial position, Set the initial attitude angle; set the learning rate. and convergence threshold The triaxial acceleration obtained based on the inertial measurement unit and triaxial angular velocity Perform integral calculation of pose change; let the elapsed time be... The change in position of the commercial vehicle The attitude change of the commercial vehicle , ;in, All are caused by the aforementioned triaxial acceleration Integrating, we obtain the pose estimation error function. The pose estimation error function This represents the difference between the estimated current pose of the commercial vehicle and the actual pose; the pose parameters are updated iteratively. ,in, Is the error function in The gradient at the point, up to the error Output the position and orientation of the commercial vehicle. According to the pose Yaw angle change rate and rate of change of velocity The driver's steering intention is determined by a pre-trained support vector machine classifier.

[0034] In practical applications, the gradient descent algorithm is used to solve for vehicle pose, and its specific implementation steps are as follows: The initial pose parameters are defined as follows: The initial attitude angles include the roll angle. Pitch angle Yaw angle The learning rate α was set to 0.05, and the convergence threshold was... Defined as error function The maximum number of iterations is 1000.

[0035] Triaxial acceleration based on inertial measurement unit and triaxial angular velocity The change in pose is calculated by integration: Among them, speed By acceleration The integral yields the result. The pose estimation error function is obtained using the Euclidean distance formula: In each iteration, the error gradient is calculated. , and according to Update the pose parameters until the convergence condition is met. Extract the yaw rate of change from the solved pose. and lateral acceleration The input is fed into a pre-trained support vector machine classifier, which outputs a driver intention label, which can be either going straight, turning left, or turning right. The classifier training dataset contains 5000 sets of driving scenario data, including urban roads and highways, with a classification accuracy of ≥95%.

[0036] In some embodiments, after determining the vehicle's steering intention, the theoretical steering blind spot range can be determined based on geometric relationships, such as the vehicle's structural parameters (e.g., body length, width, wheelbase, etc.) and the geometric relationships during steering. For example, when a vehicle turns, the crescent-shaped area formed by the difference in turning radii between the front inner wheel and the rear inner wheel is the inner wheel difference blind spot, which is part of the steering blind spot.

[0037] In some embodiments, the rearward blind spot refers to the area directly behind the driver that cannot be observed through the rearview mirror or side mirrors. The specific boundaries of the rearward blind spot can be determined based on actual operating data, actual vehicle structure parameters, and sensor installation locations. For example, the horizontal field of view angle of the rearview mirror is typically 50°-60°, symmetrically distributed around the vehicle's longitudinal centerline. This embodiment will not elaborate further.

[0038] Furthermore, during a turn, body roll causes the rearview mirror's field of vision to shift, potentially extending the central area directly behind the vehicle beyond the viewpoint. Steering intent is crucial for determining whether the vehicle has entered a "turning state," and changes in the rear blind spot are directly related to the vehicle's body posture deviation during a turn (body posture identification is explained earlier and will not be repeated here). When the vehicle is turning, the rear blind spot shifts in the opposite direction of the turn. For example, when turning left, the rear end swings to the right, expanding the right-side rear blind spot and shrinking the left-side. The smaller the turning radius (the smaller the R), the greater the rear end swing and the more pronounced the lateral expansion of the blind spot. This embodiment will not elaborate further.

[0039] After that, as Figure 2 As shown, ground features of the first target region in the predicted steering blind spot are detected. Road surface images of the predicted steering blind spot are acquired using the image sensor, and a deep learning model is used to identify abnormal road surface features, including color differences and / or local texture abrupt changes. Specifically, the operation can be based on the YOLOv5 object detection algorithm, using its CSPNet structure to extract image features. The CSPNet structure divides the input feature map into two parts; one part is processed by a convolutional layer, and the other part is downsampled. Finally, the two feature maps are concatenated and fused to obtain the processed feature map, which is then used to identify abnormal road surface features.

[0040] In practice, the input image is augmented with Mosaic data, and four images are randomly cropped and stitched together to simulate complex road surface scenes, enhancing the model's generalization ability. HSV color space conversion is used to extract road surface color features: hue (H), saturation (S), and brightness (V). Abnormal color judgment thresholds are set as ΔH > 15% or ΔV > 20%. The input image is divided into two parts using a CSPNet structure: one part undergoes convolutional layer extraction for texture features, and the other part is directly downsampled to the same size. After fusing the features of the two parts, a spatial pyramid pooling module is used to enhance multi-scale representation capabilities, identify local texture abrupt changes, and achieve a Sobel edge intensity > 50. It is understood that in this embodiment, the first target area does not have obvious signs or obstacles hindering the continued passage of the target object. Obvious signs or obstacles include, but are not limited to, fences, signs, or other barriers or warning signs.

[0041] Step 102: Obtain second running data of the first target object located within the predicted steering blind zone, and predict the first change of direction probability of the first target object moving away from the predicted steering blind zone based on the second running data and the first ground features; preferably, the present invention predicts the first change of direction probability of the first target object moving away from the predicted steering blind zone through a pre-trained first model, wherein the first model is a Bayesian network-based model.

[0042] Here, as Figure 2 As shown, second running data of a first target object located within the predicted steering blind zone 1 is obtained. The second running data includes the movement trajectory and speed of the first target object. The change-of-direction probability is analyzed based on the second running data and the first ground feature. If the probability is lower than a preset threshold, the first change-of-direction probability is determined to be low.

[0043] The second set of operational data is acquired collaboratively by millimeter-wave radar and an image sensor. The millimeter-wave radar, with its precise detection capabilities for target distance, velocity, and azimuth, provides basic motion information of the target object in space. The image sensor, through real-time imaging of the surrounding environment, provides visual evidence for identifying the target object's category and obtaining its detailed trajectory. The point cloud data from the millimeter-wave radar and the image data from the image sensor are spatiotemporally aligned and fused using the CenterFusion model to accurately output the second set of operational data, including the target object's trajectory and velocity.

[0044] Here, for example, the spatiotemporal alignment and fusion processing using the CenterFusion model can be achieved by aligning the timestamps of the image sensor and the millimeter-wave radar through a hardware synchronization signal, ensuring the synchronization of data acquisition. The millimeter-wave radar point cloud data is then calibrated using an external parameter calibration matrix. Mapping to the image coordinate system completes spatial alignment. Image data is input into the YOLOv5 backbone network to extract two-dimensional feature maps. Millimeter-wave radar point cloud data is enhanced by a point cloud dilation algorithm, converted into a depth map, and stitched with an RGB image to form a four-channel input. The fused feature map is input to the detection head prediction layer, outputting the target object's trajectory and second velocity. and the three-dimensional position information of the target object .

[0045] In practice, the Kalman filter algorithm is used to optimize the trajectory and speed data of the target object in order to reduce noise interference and ensure the accuracy and stability of the data.

[0046] Combining the first ground features, such as the changes in road material implied by differences in road surface color, for example, newly paved dark asphalt road surfaces, especially those containing human footprints, small animal footprints, or ruts, indicate that the road surface construction is not yet fully completed and is not suitable for normal traffic. In this case, compared to the surrounding normal road surfaces, it may cause the target to change their driving intention, thereby affecting their driving direction. Or, abnormal road conditions reflected by sudden changes in local texture, such as sudden potholes or bumps on the road surface, may cause the target to change direction in order to maintain a stable driving position. Analyze the possibility of their change of direction.

[0047] By constructing a Bayesian network-based model for assessing the probability of a change of direction, multiple factors such as the target object's trajectory, speed, and ground features are used as input variables. Through a trained Bayesian network model, probabilistic inference and calculation within the model output a numerical value indicating the probability of a change of direction. If this probability is lower than a preset threshold, the probability of a change of direction is considered low. The construction of the Bayesian network model specifically includes the following variable hierarchy: Root node: Ground feature attributes (containing 3 sub-networks); Material change indicators: Color difference / Traces density / Construction marking visibility; Texture mutation features: Pothole depth / Protrusion height / Crack density; Environmental interference factors: Footprint distribution rate / Rut coverage rate / Foreign object presence probability; Intermediate nodes: Dynamic behavior parameters; Trajectory deviation: Lateral displacement standard deviation / Heading angle fluctuation rate; Velocity variation coefficient: Instantaneous acceleration / Braking frequency / Velocity entropy; Leaf nodes: Change of direction decision probability.

[0048] Step 103: Obtain the third running data of the second target object located in the blind zone directly behind it, determine the second target area and its second ground features and the speed change of the second target object based on the third running data, and predict the second change of direction probability of the second target object entering the predicted turning blind zone.

[0049] Preferably, in response to detecting that the second ground feature is an obstacle for the second target object to enter the target area, and the speed change of the second target object meets a first preset condition, the probability of the second change of direction is determined to be high, wherein the first preset condition is that the speed of the second target object is less than a first threshold and / or the deceleration of the second target object exceeds a second threshold.

[0050] In practice, such as Figure 3 As shown, the third operational data of the second target object located in the blind zone 1 directly behind is also acquired by relying on millimeter-wave radar and image sensor, and is fused and processed by CenterFusion model; the specific acquisition process of the third operational data is the same as the acquisition process of the second operational data described above, and will not be repeated in this embodiment.

[0051] Based on this third operational data, the second target area and the second ground features are determined. For example... Figure 3 As shown, the image data is processed using a target detection algorithm to identify the ground features of the second target area in the first sub-region 3 and the second sub-region 4. The algorithm detects whether there are any obstacles in the first sub-region 3 and the second sub-region 4 that prevent the second target object from continuing to move towards the first sub-region 3 and the second sub-region 4. These obstacles include, but are not limited to, obstacles in the first sub-region 3 and the second sub-region 4, such as other vehicles, other objects, or roadblocks placed on the roadside (e.g., roadblocks placed due to road construction in the adjacent area), or other situations that prevent the second target object from entering, such as dead-end roads without traffic lights, green belts, etc., which prevent the second target object from continuing to move towards the first sub-region 3 and the second sub-region 4.

[0052] For the detection of the second ground feature, a similar deep learning model and image processing method as in the aforementioned steps are used, which will not be repeated in this embodiment.

[0053] If the first sub-region 3 and the second sub-region 4 are detected to obstruct the second target object from continuing to move towards the first sub-region 3 and the second sub-region 4, in order to further improve the accuracy of identifying whether the second target object has turned, the analysis of the second target's speed change will continue. If the second target's speed drops to zero or the speed difference of the deceleration is... ,in ,in The velocity before the change Given the changed speed, the target object will most likely need to turn. It's understandable that a target object typically decelerates when turning to avoid the first sub-region 3. Based on this, if... Figure 3As shown, the second target object may turn towards the second sub-region 4 or towards the commercial vehicle's turning blind spot 1. However, it is understandable that, based on the aforementioned analysis, since the second sub-region 4 also contains obstacles preventing the second target object from continuing to move forward, the probability of the second target object turning towards the commercial vehicle's turning blind spot 1 is higher in this case, i.e., the probability of the second target object entering the predicted turning blind spot is higher; if the target object's speed remains constant or the speed difference is... The speed difference here can be either a deceleration speed difference or an acceleration speed difference. Therefore, the second target object might directly pass through the first sub-region 3 and continue forward. and The value of can be set based on empirical or experimental values, and this embodiment does not impose any restrictions.

[0054] In some embodiments, the deceleration of the second target object can also be used. The rate at which the velocity decreases within a predetermined time t is used to determine whether the probability of a second change of direction is high. Specifically, the deceleration... If the probability of a second change of direction is high, it can be understood that when a target object makes a turn, it will usually reduce its speed to a safe turning speed within a short period of time. The value of can be set based on empirical or experimental values, and this embodiment does not impose any restrictions. Furthermore, it is understood that the probability of a second change of direction can also be determined by simultaneously satisfying the above conditions through the difference between the deceleration and the slowdown of the second target object; this will not be elaborated upon in this embodiment.

[0055] Step 104: If the first change-of-direction probability is lower than the first preset value and / or the second change-of-direction probability is higher, then an alarm signal is output.

[0056] Here, if the change-of-direction probability of the first target object is lower than a preset value and / or the change-of-direction probability of the second target object is higher, an alarm signal is output. For example, when the first change-of-direction probability is detected to be lower than the first preset value and / or the second change-of-direction probability is higher, an audible and visual warning is triggered, or a braking command is sent to the electronic braking device, or both simultaneously. The braking force can be increased step by step. The first preset value can be set according to experimental or empirical values, and this embodiment does not impose any restrictions.

[0057] like Figure 4As shown, Embodiment 2 of the present invention provides a blind spot braking system for commercial vehicles, comprising: a first detection module 201, configured to acquire first operating data of the commercial vehicle, determine at least based on the first operating data a predicted steering blind spot and a direct rear blind spot of the commercial vehicle, and detect a first ground feature of a first target area in the predicted steering blind spot; a first prediction module 202, configured to acquire second operating data of a first target object located within the predicted steering blind spot, and predict a first change-of-direction probability of the first target object moving away from the predicted steering blind spot based on the second operating data and the first ground feature; a second prediction module 203, configured to acquire third operating data of a second target object located in the direct rear blind spot, determine a second target area and its second ground feature based on the third operating data, and predict a second change-of-direction probability of the second target object entering the predicted steering blind spot; and a warning module 204, configured to output an alarm signal if the first change-of-direction probability is lower than a first preset value and / or the second change-of-direction probability is higher.

[0058] Embodiment 3 of the present invention provides a storage medium storing program instructions, which, when executed, implement the blind spot braking method for commercial vehicles as described in any of the preceding embodiments.

[0059] like Figure 5 As shown, Embodiment 4 of the present invention provides an electronic terminal, including a processor and a memory. The memory stores program instructions, and the processor executes the program instructions to implement the blind spot braking method for commercial vehicles as described in any of the preceding embodiments.

[0060] This invention can be an apparatus, method, and / or computer program product. A computer program product may include a readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of the invention.

[0061] Storage media can be tangible devices that hold and store instructions for use by instruction execution devices. Storage media can include, for example, electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination thereof.

[0062] It should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This way of describing the specification is only for clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

[0063] The detailed descriptions listed above are merely specific descriptions of feasible embodiments of the present invention, and are not intended to limit the scope of protection of the present invention. All equivalent embodiments or modifications made without departing from the spirit of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for blind spot warning in commercial vehicles, characterized in that, Includes the following steps: Acquire first operating data of the commercial vehicle, determine the predicted steering blind spot and the rear blind spot of the commercial vehicle based at least on the first operating data, and detect the first ground feature of the first target area in the predicted steering blind spot; Acquire second operational data of a first target object located within the predicted steering blind zone, and predict a first change-of-direction probability of the first target object moving away from the predicted steering blind zone based on the second operational data and the first ground feature; Acquire third operational data of a second target object located in the blind zone directly behind it, determine the second target area and its second ground features based on the third operational data, and predict the second change-of-direction probability of the second target object entering the predicted turning blind zone; wherein, the second change-of-direction probability of the second target object entering the predicted turning blind zone is determined based on the second target area and its second ground features and the speed change of the second target object. If the first change-of-direction probability is lower than the first preset value and / or the second change-of-direction probability is higher, an alarm signal is output. Specifically, in response to detecting that the second ground feature is an obstacle for the second target object to enter the second target area, and the speed change of the second target object meets the first preset condition, the probability of the second change of direction is determined to be high.

2. The method for blind spot alarm in commercial vehicles according to claim 1, characterized in that, include: Based on the second running data and the first ground features, the first change of direction probability of the first target object moving away from the predicted turning blind zone is predicted by the pre-trained first model.

3. The method for blind spot alarm in commercial vehicles according to claim 1, characterized in that, include: The road surface image of the predicted steering blind spot is acquired by an image sensor, and an abnormal road surface feature is identified by a deep learning model. The abnormal feature includes color difference and / or local texture abrupt change.

4. The method for blind spot alarm in commercial vehicles according to claim 1, characterized in that, include: The second operational data includes the movement trajectory and speed of the first target object. Based on the second operational data and the first ground features, the probability of its change of direction is analyzed. If the probability is lower than a preset threshold, the probability of the first change of direction is determined to be low.

5. The method for blind spot alarm in commercial vehicles according to claim 1, characterized in that, include: The first preset condition is that the speed of the second target object decreases by more than a first threshold and / or the deceleration of the second target object exceeds a second threshold.

6. The method for blind spot alarm in commercial vehicles according to claim 1, characterized in that, The alarm signals include triggering audible and visual warnings to prompt the driver and / or sending graded braking commands to the electronic braking device.

7. A blind spot warning system for commercial vehicles, characterized in that, include: The first detection module is used to acquire the first operating data of the commercial vehicle, determine the predicted steering blind spot and the rear blind spot of the commercial vehicle based at least on the first operating data, and detect the first ground features of the first target area in the predicted steering blind spot. The first prediction module is used to acquire second running data of a first target object located within the predicted steering blind zone, and predict a first change of direction probability of the first target object moving away from the predicted steering blind zone based on the second running data and the first ground features. The second prediction module is used to acquire third running data of the second target object located in the blind zone directly behind it, determine the second target area and its second ground features based on the third running data, and predict the second change of direction probability of the second target object entering the predicted turning blind zone; wherein, the second change of direction probability of the second target object entering the predicted turning blind zone is determined based on the second target area and its second ground features and the speed change of the second target object. The early warning module is used to output an alarm signal if the first change-of-direction probability is lower than a first preset value and / or the second change-of-direction probability is higher. Specifically, in response to detecting that the second ground feature is an obstacle for the second target object to enter the second target area, and the speed change of the second target object meets the first preset condition, the probability of the second change of direction is determined to be high.

8. A storage medium storing program instructions, characterized in that: When the program instructions are executed, the method for blind spot alarm of commercial vehicles as described in any one of claims 1 to 6 is implemented.

9. An electronic terminal, comprising a processor and a memory, wherein the memory stores program instructions, characterized in that: The processor executes program instructions to implement the blind spot alarm method for commercial vehicles as described in any one of claims 1 to 6.

Images