Unsafe parking detection

Abstract

Methods and systems for detecting unsafe parking scenarios to enhance road safety are provided. The method utilizes at least one processor to detect when a vehicle is stopped and receives visual data from an outward-facing camera on the vehicle. The visual data is then processed to determine the risk of other vehicles passing the stopped vehicle at an unsafe distance or speed. If the risk is determined to be above a certain threshold, an alert is generated for the driver of the vehicle, to effectively prevent accidents, alert drivers to potential hazards, and significantly improve overall road safety.

Description

UNSAFE PARKING DETECTIONCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of priority to U.S. Provisional Application No. 63 / 688,459, filed August 29, 2024, the entirety of which is incorporated by reference herein.TECHNICAL FIELD

[0002] The present disclosure relates generally to vehicle parking detection systems, and more specifically, to a computer-implemented method for detecting unsafe parking of a vehicle.BACKGROUND

[0003] In recent years, the increased number of vehicles on roads has led to a rise in parking- related incidents, posing significant threats to road safety. These incidents often involve vehicles parked unsafely on the side of the road, particularly on or near freeways, highways, or main roads. It is critical to detect such instances promptly to minimize the potential hazards they pose.

[0004] Current methods for detecting unsafe parking often rely on map-based approaches. These methods, however, have several limitations. They may not consider new roads and may require LTE connectivity for remotely stored maps. Furthermore, they lack precision in making distance estimates. In addition, map-based approaches may not adequately incorporate the environmental context of a parking event.

[0005] There is a need for a more robust and consistent system that can accurately detect unsafe parking situations that is attuned to environmental context, and that is not limited to detecting events that are on or near highways but that is also capable of detecting any unsafe parking event, and issue alerts promptly.SUMMARY

[0006] The present disclosure provides an improved method for detecting unsafe parking of a vehicle. The method includes detecting that a vehicle is stopped; receiving visual data captured by an outward-facing camera of the vehicle; processing the visual data to determine a risk that other vehicles will pass the stopped vehicle at an unsafe distance and speed; and responsive to detecting that the vehicle is parked unsafely, generating, by the at least one processer, an alert for a driver of the vehicle.BRIEF DESCRIPTION OF THE DRAWINGS

[0007] FIG. 1 illustrates an environment depicting one or more edge devices in communication with a cloud server, according to an embodiment of the disclosure.

[0008] FIG. 2 illustrates a block diagram of an edge device, according to an embodiment of the disclosure.

[0009] FIG. 3 A illustrates a first scenario related to determination of unsafe parking, according to an embodiment of the disclosure.

[0010] FIG. 3B illustrates a second scenario related to determination of unsafe parking, according to an embodiment of the disclosure.

[0011] FIG. 3C illustrates a third scenario related to determination of unsafe parking, according to an embodiment of the disclosure.

[0012] FIG. 3D illustrates a fourth scenario related to determination of unsafe parking, according to an embodiment of the disclosure.

[0013] FIG. 3E illustrates a fifth scenario related to determination of unsafe parking, according to an embodiment of the disclosure.

[0014] FIG. 3F illustrates a sixth scenario related to determination of unsafe parking, according to an embodiment of the disclosure.

[0015] FIG. 4A illustrates a view from a vehicle driving on a roadway.

[0016] FIG. 4B illustrates the view from the vehicle driving on the roadway in FIG.4A as the vehicle approaches an exit ramp.

[0017] FIG. 4C illustrates the view from the vehicle approaching the exit ramp in FIG.4B as the vehicle is driven along the exit ramp.

[0018] FIG. 4D illustrates the view from the vehicle driven along the exit ramp in FIG.4C as the vehicle continues along the exit ramp at a lateral distance from the roadway that exceeds a threshold, according to an embodiment of the disclosure.

[0019] FIG. 4E illustrates the view from the vehicle driven along the exit ramp in FIG.4D as the vehicle enters a rest area with parking available to the left.

[0020] FIG. 4F illustrates the view from the vehicle entering the rest area in FIG. 4E, after the vehicle has parked in a parking space that is more than a threshold distance closer to the roadway than the vehicle was at the time corresponding to FIG. 4E, according to an embodiment of the disclosure.

[0021] FIG. 5 illustrates a series of locations superimposed on an overhead view of a highway with a rest area in each direction.

[0022] FIG. 6A illustrates a view of a rear of a truck from a vehicle travelling on a road under construction.

[0023] FIG. 6B illustrates the view of the rear of the truck in FIG. 6A, a short time later, from which it may be inferred the visible truck is moving slowly, according to an embodiment of the disclosure.DETAILED DESCRIPTION

[0024] The detailed description set forth below, in connection with the appended drawings, is intended as a description of various configurations and is not intended to represent the only configurations in which the concepts described herein may be practiced. The detailed description includes specific details for the purpose of providing a thorough understanding of the various concepts. However, it will be apparent to those skilled in the art that these concepts may be practiced without these specific details. In some instances, well-known structures and components are shown in block diagram form to avoid obscuring such concepts.

[0025] Based on the teachings, one skilled in the art should appreciate that the scope of the disclosure is intended to cover any aspect of the disclosure, whether implemented independently of or combined with any other aspect of the disclosure. For example, an apparatus may be implemented, or a method may be practiced using any number of the aspects set forth. In addition, the scope of the disclosure is intended to cover such an apparatus or method practiced using other structure, functionality, or structure and functionality in addition to or other than the various aspects of the disclosure set forth. Any aspect of the disclosure disclosed may be embodied by one or more elements of a claim.

[0026] The word “exemplary” is used herein to mean “serving as an example, instance, or illustration.” Any aspect described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other aspects.

[0027] Although particular aspects are described herein, many variations and permutations of these aspects fall within the scope of the disclosure. Although some benefits and advantages of the preferred aspects are mentioned, the scope of the disclosure is not intended to be limited to particular benefits, uses or objectives. Rather, aspects of the disclosure are intended to be broadly applicable to different technologies, and system configurations, some of which are illustrated by way of example in the figures and in the following description of the preferred aspects. The detailed description and drawings are merely illustrative of the disclosure rather than limiting, the scope of the disclosure being defined by the appended claims and equivalents thereof.

[0028] Vehicles may not be safely parked if they are parked in locations where there is a substantial risk that other vehicles will pass the parked vehicle at an unsafe distance and speed. Unsafe parking may occur when a vehicle is parked proximate to roads on which vehicles tend to travel at high speeds. Drivers may park at certain locations, such as highway on ramps or off ramps, because drivers are unaware of risks involved in such unsafe parking of the vehicles, and / or because the drivers are unable to find safer or more appropriate places to park. In certain scenarios, drivers may stop the vehicle on or near a lane of the road (such as the lane nearest the shoulder or the side of the road), making it dangerous for the other vehicles travelling on that road in the lane. In certain other scenarios, drivers may stop the vehicle on or near a lane but in a way that is not unsafe, such as when the driver is waiting at a long-duration traffic light or when all traffic on the same road is temporarily halted. To accurately detect unsafe parking of the vehicle, and to distinguish other stopping or parking events which are similar in some respects but not considered unsafe, an edge device is installed in the cabin of the vehicle. The edge device may be configured to process visual data captured by a camera and determine whether the parking is safe or unsafe based on the inferences made by processing of the visual data along with inertial sensor data. Upon determining that the parking is unsafe, the edge device may generate an alert to be provided to the driver of the vehicle. The alert may alternatively or additionally be communicated to a remote server, which may be accessible by a safety manager associated with the driver, and / or other drivers in the vicinity of the driver who is parked in an unsafe location.

[0029] FIG. 1 depicts an illustration of an environment 100 in which one or more embodiments of the present disclosure may be implemented. The environment 100 includes a cloud server 102, vehicles 104, 106, and a network 108 for communication between the cloud server and edge devices 104a, 106a installed in the vehicles 104, 106.

[0030] The network 108 can be radio access networks such as GSM, LTE, 5G or the like. In one embodiment, the network can be Wi-Fi or other wireless broadband networks.

[0031] The cloud server 102 provides configuration settings and software related services to the edge devices. The cloud server may store the data received from the edge devices. Further, the cloud server may provide alerts to the fleet operator device and driver’s device upon detection of unsafe parking by an edge device.

[0032] In one embodiment, each edge device 104a, 106a may be an advanced driver assistance system (ADAS) that includes one or more cameras and sensors. Further, each edge device104a, 106a may include a processor and communication circuitry that includes radio interfaces and antennas. The radio interfaces may correspond to a plurality of radio access technologies including one or more of GSM, NB-IoT, LTE, 5G, WLAN, Bluetooth, BT-LE, NFC, radio frequency identifier (RFID), ultra-wideband (UWB), and the like. Each edge device 104a, 106a is installed in a vehicle 104, 106, respectively, to monitor driver behavior, which may include capturing a cabin of the vehicle with a driver-facing camera and includes capturing a driving (or parking) environment of the vehicle with other sensors and / or a forward-facing camera. Further, each edge device 104a, 106a may include an audio speaker or other output devices to provide notifications and alerts to the driver, or may be connected to devices, such as a driver’s smartphone, that may be used to convey generated alert notifications to the driver. Each edge device 104a, 106a may be a standalone device or a combination of devices. In one embodiment, the edge device installed in the vehicle may include a sensor device and a computing device, where the sensor device may include cameras and other inertial sensors. In such embodiments, the camera may or may not include a processor that is capable of processing the visual data to determine a risk indicative of unsafe parking. The computing device on the other hand may include one or more processors to process the information captured by the sensor device. The sensor device may relay the captured data to the computing device via a wired or wireless connection. The distribution of functionalities among devices may be implemented to reduce the size of devices that are installed near the windshield. The various components of an exemplary edge device are explained next with reference to FIG. 2.

[0033] FIG. 2 depicts a block diagram of the edge device 200 (similar to edge devices 104a, 106a) including various components. The edge device 200 includes a processor 202, a communication module 204, input / output (I / O) module 206, memory 208, a camera module 210 and a sensor module 212. In one embodiment, the edge device 200 may not include a camera module 210 and in those scenarios, the camera module 210 may be connected to edge device 200 through wired or wireless connection. The edge device 200 may include input sensors (which may include a forward-facing camera, a driver facing camera, connections to other cameras that are not physically mounted to the device, inertial sensors, car OBD-II port sensor data (which may be obtained through a Bluetooth connection), and the like) and compute capability. The compute capability may be a CPU or an integrated System-on-a-chip (SOC), which may include a CPU and other specialized compute cores, such as a graphics processor (GPU), gesture recognition processor, and the like. In some embodiments, the edge device 200 for determining, transmitting, and / or providing alerts to an operator of a vehicleand / or a device of a remote driver monitoring system may include wireless communication to cloud services, such as with Long Term Evolution (LTE) or Bluetooth communication to other devices nearby. For example, the cloud server 102 may provide real-time analytics assistance. In an embodiment involving cloud services, the cloud server 102 may facilitate aggregation and processing of data for offline analytics. The edge device 200 may also include a global positioning system (GPS) either as a separate module or integrated within a system-on-a-chip (e.g., included in sensor module 212).

[0034] The camera module 210 may optionally include an inward facing camera and an outward facing camera to capture visual data related to the cabin of the vehicle and the environment ahead of the vehicle, respectively. The outward facing camera is configured to capture visual data related to the environment around the vehicle and the visual data is relayed to the processor 202.

[0035] The processor 202 may process the captured visual data using machine learning (ML) models deployed on the edge device 200. The ML models may be included in an unsafe parking detection module 202a. The ML models may include image classifiers to classify objects captured in the visual data, full-frame classifiers that are configured to detect and distinguish various relevant scenes or environments, and / or video classifiers.

[0036] Unsafe parking may involve parking of the vehicle on or near the road such that the parking may lead to accidents or inconvenience to the other vehicles on the road. Parking on the side of the road may lead to accidents if vehicles are not parked outside the boundary of the roads. Parking the vehicle in no parking zones may cause inconvenience to the public and is also a traffic rule violation. These types of parking may be considered unsafe parking depending on environmental context, such as whether it may be likely that other vehicles will pass the parked vehicle at close distances and high speeds. The above scenarios are for exemplary purposes and unsafe parking may not be limited to these scenarios.

[0037] To determine whether the vehicle is parked safely or not, the edge device 200 may be configured to process the visual data to detect the traffic environment around the vehicle upon detecting stoppage of the vehicle. To that effect, the edge device 200 may include machine learning models to determine riskiness of parking based on inferences made from visual data. Broadly, relevant inferences based on processing of visual data may include actual detections of one or more vehicles passing the parked vehicle at a close distance and a high speed, and / orinferences that predict the same, such as whether the vehicle is near, partially on, or on an active lane of a road on which vehicles tend to travel at high speeds.

[0038] In one embodiment, the processor 202 may determine risk that other vehicles will pass the stopped vehicle (i.e., parked vehicle) at an unsafe distance and speed. To that effect, the processor 202 may detect other vehicles passing the parked vehicle in which the edge device 200 is installed. The vehicle identification may be done using object detection algorithms like R-CNN, Fast R-CNN, or YOLO. The processor 202 may estimate the lateral distance from the detected vehicle to the parked vehicle. In one example, lateral distance estimation may be performed using techniques like stereo vision, where the edge device 200 may have two or more cameras to estimate distances based on the parallax effect. Alternatively, the edge device 200 may use monocular depth estimation techniques, such as training a machine learning model to estimate distances from single images. The processor 202 may estimate the speed of the passing vehicles by tracking the position of each passing vehicle across multiple frames and calculating the change in position over time. An alert may be raised by the processor 202 indicating that the parking is unsafe if the distance between the passing vehicle and the distance is lower than a predefined threshold distance and the passing vehicle is moving faster than a predefined threshold speed.

[0039] The speed of the vehicles passing the parked vehicle may indicate the speed limits associated with the road near which the vehicle is parked. The processor 202 may determine a road type associated with the parked vehicle by comparing the speed of the vehicles passing the parked vehicle and the predefined threshold speed. Further, the processor 202 may keep track of the number of vehicles passing the parked vehicle within a predefined time window that meets the speed and distance thresholds. In one example, the processor 202 may use a counter to count the number of vehicles that pass the parked vehicle at an unsafe distance and speed (i.e., lateral distance less than distance threshold and speed greater than speed threshold) for a predefined time frame. For example, the predefined time frame may be 10 minutes or 30 minutes, the selection of which may depend on the density of traffic previously observed on the road. The frequency data may also be used to calculate a risk score or safety level, which can be part of the unsafe parking alert information that is communicated to a remote server.

[0040] In one embodiment, the magnitude of the determined risk may be modified based at least one determined frequency of the passing vehicles, where risk indicates that vehicles will pass the parked vehicle at the unsafe distance and speed. Further, the processor 202 may modifythe severity level of unsafe parking alerts based on frequency of the vehicles passing the parked vehicle at unsafe distances and speeds. Various parking alerts may be generated based on the determined safety level of the parking of the vehicle. The safety level may be determined by having multiple thresholds that indicate different safety levels. For example, severe alert may be generated upon detecting highly unsafe parking if a high number of vehicles are passing the parked vehicle at higher speeds, which may indicate that the vehicles is parked on or near a highway. In another example, a moderate alert may be raised if frequency of the vehicles passing the parked vehicle is greater than a moderate threshold but less than a severe threshold.

[0041] In one embodiment, the lateral distance between the parked vehicle and the passing vehicles may be determined by calculating the distance from center of the parked vehicle’s lane to the bottom center or bottom nearby comer of the bounding box for the passing vehicle. To that effect, the processor 202 may detect the lane lines in the image using an image classifier. The processor 202 may detect the passing vehicles and their bounding boxes using an object detection algorithm like R-CNN or YOLO. Thereafter, the distance from the center of the ego lane (i.e., lane on which the vehicle is parked, or a calibrated lane where the lane boundaries typically appear in the visual data when the vehicle is driving, or an estimated center of the vehicle in the direction of travel which may correspond to a vanishing point) to the bottom center or bottom nearby corner of the bounding box for the passing vehicle may be calculated in the image space, resulting in a raw pixel distance. The processor 202 may normalize the raw pixel distance based on the lane calibration width at that horizontal point in the image, thereby converting the pixel distance to a real-world distance.

[0042] In another embodiment, the processor 202 may determine the outer edges of the parked vehicle and the inner edge of the passing vehicle using edge detection techniques and / or using the bottom nearby corner of a bounding box. The processor 202 may calculate the distance between the detected edges in image space. The distance may be projected based on the lane / road calibration to normalize the distance along the path of travel. This may result in converting the pixel distance to a real-world distance. The processor 202 may improve the accuracy of the distance estimation by averaging or weighting the distances over several bounding box detections. The weights may be based on the size (width of vehicle), proximity (horizontal distance between the bottom nearby corner of the bounding box and the portion of the image corresponding to locations directly in front of the vehicle), or other features of the bounding boxes.

[0043] In one embodiment, the processor 202 may be configured to determine the type of road based on visual data captured by an outward facing camera of the edge device. Different road types correspond to different expectations of travel speeds of other vehicles, so determining the road type of the road that the vehicle is parked on or parked near can be a helpful component of a determination of a risk level of a parking location. The processor 202 may include an image classifier to determine the type of the road based on visual data. Prior to deployment on the edge device 200, the image classifier may be trained on a dataset of road images. This dataset should include different types of roads such as highways, freeways, residential streets, country roads, on-ramps, off-ramps, etc. Each image in the dataset is labeled with the type of road it represents and may be further classified based on whether it is a type of road for which roadside parking is considered safe or unsafe. In this way, a trained scene classifier may contribute to a determination whether parking is safe or unsafe. The scene classifier may holistically associate the visual features of each image with the type of road. A scene classifier may work in conjunction with an object classifier that is trained to detect individual objects on or near roads. These features (which may be learned in an internal representation of a scene classifier) or objects (which may be explicitly detected by an object classifier) could include the number of lanes, presence of road signs, types of road signs, information content of road signs, road markings, surroundings (like buildings, trees, or open fields), the presence or absence of sidewalks, etc. In some embodiments, a scene classifier and an object classifier may share a common trunk in a multi-headed neural network design.

[0044] In one embodiment, the processor 202 may also determine the position of the vehicle relative to the road based on the visual data. The processor 202 may determine the position of the vehicle relative to the road based on camera’s specific field of view, which typically includes part of the vehicle itself (like the hood or side mirrors), the road ahead, and sometimes the road behind or to the sides. In some embodiments, the position of the vehicle relative to the road may be inferred from auxiliary camera, such as one positioned on a side of a truck cabin that is positioned to monitor the area along the side of the truck and / or trailer. For example, the auxiliary camera may be positioned to support blind spot detection.

[0045] The object classifier may identify lane markings, curbs, sidewalks, and other road features by processing the visual data. The processor 202 may use edge detection methods or Convolutional Neural Networks (CNNs) to identify predefined objects in the visual data. Further, the classifier may identify parts of the vehicle within the image, such as the hood orside mirrors. The relative position of these parts to the detected road features may provide an estimate of the vehicle’s position on the road and / or proximity to an active lane of traffic on that road. To that effect, the classifier may analyze the spatial relationships between the detected road features and parts of the host vehicle to determine the position of the vehicle relative to the road, or more specifically to moving traffic on the road that the vehicle is parked on or near. The spatial relationships may include whether the vehicle is centered within the lane of the road, near the edge of the lane of the road, or on the shoulder of the road.

[0046] In one embodiment, a frame-based and / or video-based classifier may track changes over a series of frames captured by the camera to help determine the vehicle’s movement that include detecting lane departures or instances where the vehicle crosses onto the shoulder.

[0047] In some embodiments, the processor 202 may use data from other sensors such as GPS, inertial measurement units (IMUs), or vehicle telematics data in combination with the visual / image data to improve the accuracy of one or more processing steps that may underly a determination that the parking is risky / unsafe.

[0048] In some embodiments, the processor 202 may quantify the safety level of the parking based on additional parameters derived from the processing the visual data. The additional parameters may include lateral distance of the passing vehicles to the parked vehicle, the speed of the passing vehicles, the number of passing vehicles in a time interval, road signs identified in the visual data, and similar data.

[0049] In some scenarios, the vehicle may be stuck in traffic congestion or might have stopped near a road intersection for a longer time, which may lead to false detection of unsafe parking since these are traffic scenarios for which it is unlikely that other vehicles will pass the stopped vehicle at close distances and high speeds. To refine the detection of unsafe parking, the edge device 200 may be configured to identify scenarios that may lead to vehicle stoppage on the road but should not be classified under unsafe parking. Such scenarios may include, but are not limited to, temporary stoppage of all traffic on the road.Validation Framework Implementation

[0050] The edge device 200 implement a validation framework comprising one or more analytical stages that execute prior to alert generation. This framework may increase the precision of unsafe parking detection so that alerts that are generated for true unsafe parkingscenarios and false alarms are reduced. These analytical stages may include analyzing historical movement patterns and / or real-time environmental conditions. Two exemplary validation mechanisms are described below: movement pattern validation and environmental context validation, each of which may contribute to a final risk assessment determination of the parking of the vehicle.Movement Pattern Validation

[0051] In one embodiment, the edge device 200 may implement temporal analysis of vehicle movement patterns to distinguish between true unsafe parking events and temporary stops associated with normal traffic navigation. The processor 202 may process a first time series of speed data spanning a predetermined duration (e.g., 270 seconds) to establish the vehicle’s movement context. This first time series analysis determines whether the vehicle transitioned from high-speed travel to a complete stop, which may serve as an indicator of unsafe parking.

[0052] To refine the unsafe parking accuracy, the processor 202 may further analyze a second time series that represents a subset of the first time series, typically focusing on the most recent portion of the movement history (e.g., the last 50 seconds before stopping). Within this second time series, the processor 202 examines speed patterns for characteristic rise and fall profiles. A detected pattern showing speed increase followed by decrease within this window indicates the vehicle likely exited a main roadway, navigated to an adjacent area, and came to rest — a behavior consistent with rest stop usage rather than unsafe roadside parking within the context of a longer transition from a high speed to a complete stop. Upon detecting such patterns, the edge device may suppress alert generation.

[0053] The speed pattern analysis may be complemented by determining the final parking context. When the second time series exhibits the characteristic rise and fall pattern, the processor 202 may determine that the vehicle is parked safely in an area adjacent to, but separate from, the main roadway, such as a rest area parking lot or service plaza, or in an area adjacent to a parking lot where vehicles commonly stop after exiting the main roadway. This determination may prevent false positive alerts for vehicles that have properly exited the highway system before stopping.

[0054] In one embodiment, the processor 202 processes the series of distance measurements concurrently with the first time series of speed data, maintaining temporal synchronization between speed and distance data throughout the predetermined duration. Each distancemeasurement in the series may be temporally indexed to correspond with speed measurements in the first time series, enabling the processor 202 to correlate lateral movement patterns with speed variations during the same temporal window. This concurrent processing ensures that both speed-based exit patterns and distance-based lateral movements are evaluated within the same temporal context, enhancing the accuracy of identification of a safe parking scenario, such as rest area detection or parking lot detection.

[0055] Complementing the temporal speed analysis, the processor 202 may perform spatial tracking of the vehicle’s lateral position to further validate legitimate stopping scenarios.

[0056] In addition to or as an extension of visual position determination, the processor 202 may track the vehicle’s lateral distance from the roadway throughout the approach to the stopping event. The processor 202 processes a series of distance measurements, where each measurement represents the vehicle’s perpendicular distance from the road edge or centerline at a specific time point. This distance data series corresponds temporally with the speed data series, enabling correlated analysis of vehicle movement patterns.

[0057] The processor 202 evaluates this distance series to identify patterns indicative of safe parking behaviors. Identifying patterns includes comparing the final distance measurement (at the point of vehicle stoppage) with intermediate distance measurements throughout the series. When the final distance falls within a threshold range suggesting roadside parking (e.g., 1.5 to 15 meters from road edge), but intermediate measurements show the vehicle traveled beyond this threshold distance, the edge device 200 recognizes an appropriate parking event, which may correspond to a rest area entry pattern. This pattern where the vehicle moves away from and then back toward the roadway indicates the vehicle exited to an adjacent facility rather than simply pulling over on the shoulder.

[0058] Upon detecting that intermediate distance measurements exceeded the threshold while the final distance is within the threshold, the processor 202 suppresses alert generation. This suppression mechanism prevents false positives for vehicles that park at the periphery of rest areas or service facilities, even when such parking positions place them relatively close to the main roadway. The distance threshold may be dynamically adjusted based on road type, with highways requiring larger threshold distances than surface streets.

[0059] An example of movement pattern validation is illustrated in FIG 4A-4F and 5. FIGURES 4A - 4F illustrate a series of frames captured by a camera having a front facing view froma vehicle driving. In FIG 4A the vehicle is on a first roadway. In FIG. 4B the vehicle approaches an exit ramp. In FIG. 4C the vehicle is driven along the exit ramp. In FIG. 4D the vehicle is continuing along the exit ramp. At this point, the lateral distance from the roadway exceeds a first threshold, where the first threshold indicates that the truck has driven away from the first roadway a distance that exceeds parking on the shoulder of the first roadway. In FIG. 4E the vehicle enters a rest area with parking available to the left. In FIG. 4F the vehicle has parked in a parking space that was visible in FIG. 4E. The vehicle’s location in FIG. 4F is closer to the roadway than it was in FIG. 4D. The final distance from the roadway, corresponding to FIG. 4F may be close enough to the roadway that it could be considered within a range of distances that would otherwise be considered unsafe parking. While stationary, the camera in the vehicle continues to have a view of fast-moving traffic from the first roadway. However, this in not an unsafe parking location, and is instead a designated truck parking location of a rest area. As such, an unsafe parking alert should not be generated for this scenario and / or should not be communicated to a remote server.

[0060] FIG. 5 illustrates a series of locations superimposed on an overhead view of a highway with a rest area in each direction. Some of these superimposed locations correspond to the camera views depicted in FIGURES 4A- 4F. In FIG.5, the increase lateral distance from the first roadway, followed by a turn back in the direction of the roadway can be readily appreciated.Environmental Context Validation

[0061] In combination with movement pattern validation or independently, the processor 202 may be configured to analyze the immediate visual environment to identify real-time conditions that may explain the vehicle stoppage and negate unsafe parking risks.

[0062] In one embodiment, the processor 202 may further refine the risk determination by analyzing the motion state of vehicles detected in proximity to the stopped vehicle. Upon detecting a second vehicle at a close distance to the stopped vehicle, the processor 202 may determine whether the second vehicle is stationary or in motion. The detection of a stationary vehicle ahead or a slowly moving vehicle significantly reduces the level of assessed risk, as either of these conditions typically indicates queued traffic rather than an unsafe parking scenario. The processor 202 may utilize temporal analysis of sequential frames to determine the motion state of the detected second vehicle, comparing, for example, bounding box positions across multiple time intervals. Trucks that are queued at a vehicle weigh station, for example, may by stationary and behind other trucks that are stationary or slowly moving, andnearby traffic may pass these trucks at a high rate of speed, but these would not be considered an unsafe parking scenario.

[0063] The determination of whether the second vehicle is stationary may be enhanced through bounding box analysis. Specifically, the processor 202 may infer a bounding box around the back portion of the detected second vehicle using object detection algorithms trained to detect vehicle rear profiles. The consistency of this bounding box location across multiple frames, which may be combined with an inference of minimal pixel displacement, provides reliable indication of a stationary vehicle.

[0064] In certain implementations, the processor 202 may correlate the detection of stationary vehicles ahead with specific traffic scenarios. Upon determining that multiple vehicles are stationary in a queue formation, the processor 202 may infer the presence of a toll booth, vehicle inspection station, or similar checkpoint. Such determination may be reinforced by analyzing the spatial arrangement of detected vehicles, lane configurations, and infrastructure elements visible in the visual data.

[0065] The edge device 200 may implement multi-factor analysis combining visual detection, temporal patterns, and spatial relationships to provide accurate unsafe parking detection while minimizing false positives. This layered approach addresses various real-world scenarios where simple proximity-based detection could yield incorrect results. The integration of stationary vehicle detection, movement pattern analysis, and distance tracking creates a validation framework that helps distinguish between unsafe parking and legitimate stopping scenarios.

[0066] An example of environmental context validation is illustrated in FIG 6A-B. FIG. 6A illustrates a view of a rear of a truck from a vehicle travelling on a road under construction. FIG. 6B illustrates the same view of the rear of the truck a short time later, from which it may be inferred the visible truck is moving slowly. The bounding boxes around the visible truck (not shown), can be compared to determine the relative speed of the vehicle and the vehicle having the camera. In addition, the distance between the bounding box at other detected objects, such as the construction cone, along with the elapsed time between the two frames, can be analyzed to determine that the vehicle ahead is moving slowly. In general, when the vehicle directly ahead of the vehicle having the camera is moving slowly, the situation is not an unsafe parking scenario.

[0067] Through implementation of the validation framework described above, the edge device 200 effectively distinguishes between certain unsafe parking requiring driver alerts and legitimate stopping scenarios. However, the edge device further encompasses detection capabilities for emergency situations, such as when the vehicle was involved in a collision, that, while not classified as unsafe parking, nevertheless require immediate attention and specialized alert protocols.

[0068] Scenarios for which the stopped vehicle may be in danger, but which is not accurately described as unsafe parking may include breakdown of the vehicle in the middle of the road, collision of the vehicle, emergency stoppage by driver in the middle of the road. In these dangerous scenarios, a different type of alert may be generated so that the driver may be provided with appropriate and timely assistance.

[0069] In one embodiment, the processor 202 may determine that the vehicle is in middle of a lane of traffic on a road (not on a shoulder or in a lane of a city road for which parking may be permissible). The processor 202 may further determine whether there are any vehicles in front of the vehicle in which the edge device 200 is installed. The processor 202 may determine that the vehicle is stopped because of a traffic congestion on the road based on inertial sensor data along with visual data. Upon determining that there are obstructing vehicles in front of the vehicle, the processor 202 may determine that the vehicle may be involved in a collision based on the inertial sensor data along with the visual data. In these cases, the fleet manager may be notified if the vehicle is involved in a collision or if there is a vehicle breakdown or a case of emergency related to the driver. Whereas stopping in the middle of a lane due to temporary stoppages in traffic should be ignored from the perspective of generating unsafe parking alerts, the driver safety system as a whole should be attuned to properly handling situations for which the driver and vehicle were actually involved in the event that caused the temporary stoppage. The additional or separate processing that may detect that the vehicle was involved in a collision may result in alerts that are different from unsafe parking alerts. Such alerts may be presented to safety managers and support personnel differently and in a manner that may instead trigger emergency response protocols.

[0070] Upon determining that there are no vehicles in front of the vehicle, in contrast, the processor 202 may generate an unsafe parking alert (330), but it may additionally determine that there might be an issue with vehicle based on vehicle data received from vehicle diagnostics system installed in the vehicle other than the edge device 200. In these cases, thealert may still be considered an unsafe parking alert, but the additional information may be used to make the alert more informative. Not only does the fleet manager know that the vehicle is parked in an unsafe location, but the fleet manager may also know that the vehicle is experiencing a mechanical or similar issue which may prevent the driver from moving the vehicle to a safer location.

[0071] In one embodiment, the processor 202 may identify whether the vehicle is within a lane using an image classifier that recognizes lane markings. The position of the vehicle relative to these markings may then indicate whether the vehicle is parked within a lane of travel. Upon detecting that the vehicle is parked within a lane of the travel, the processor may determine whether there are any blocking objects in front of the vehicle using a scene and / or object classifier, either alone or in conjunction. An object classifier used for this purpose may be trained to recognize other vehicles, pedestrians, and common road obstacles. If no such objects are detected within a certain distance in front of the vehicle, the processor 202 may infer that the vehicle is not blocked. In this example, a scene classifier may be utilized to determine if the vehicle is waiting at an intersection. An object classifier could similarly be utilized to determine if there are objects in the environment, such as traffic lights, that indicate that the driver may be waiting at a long traffic light. The processor 202 may trigger an alert to a fleet safety manager or an operation center upon determination that the vehicle is within a lane, unblocked, not waiting at an intersection (320), and stopped. The alert may include the vehicle’s location (using GPS data) and other relevant information. The manager may then reach out to the driver to understand the situation and provide assistance if needed.

[0072] Returning to FIGURES 3A-F, FIG. 3A illustrates a first scenario for unsafe parking detection of a vehicle in which the edge device is installed, according to an embodiment of the present disclosure.

[0073] At step 302, the edge device 200 may detect that the vehicle is stopped. In one example, the edge device 200 may determine that the vehicle has stopped for more than a predefined time period. The determination of the vehicle stoppage is based on one or more of: speed from a GPS sensor, engine speed from vehicle data, inertial sensors, visual odometry, or visual consistency / motion vectors.

[0074] At step 310, the edge device 200 may process visual data captured by a camera module associated with the edge device 200. The edge device 200 may determine a risk that other vehicles will pass the stopped vehicle at an unsafe distance and speed.

[0075] At step 312, the edge device 200 may detect one or more vehicles passing the stopped vehicles at close distance and high speed. The edge device 200 may identify the vehicles using one or more object classifiers, image classifiers. Lateral distance between the stopped vehicle and the passing vehicle may be determined and compared with a predefined distance threshold. Further, the speed of the passing vehicle may be determined and compared with a predefined speed threshold.

[0076] At step 330, the edge device 200 may generate an alert to be provided to driver if lateral distance is less than the predefined distance threshold and the speed of the passing vehicle is greater than the predefined speed threshold. The alert may indicate that the vehicle stoppage at the current location is unsafe.

[0077] FIG. 3B illustrates a second scenario for unsafe parking detection of a vehicle in which the edge device is installed, according to an embodiment of the present disclosure.

[0078] At step 302, the edge device 200 may detect that the vehicle is stopped. In one example, the edge device 200 may determine that the vehicle has stopped for more than a predefined time period. The determination of the vehicle stoppage is based on one or more of: speed from a GPS sensor, engine speed from vehicle data, inertial sensors, visual odometry, or visual consistency / motion vectors.

[0079] At step 310, the edge device 200 may process visual data captured by a camera module associated with the edge device 200. The edge device 200 may determine a risk that other vehicles will pass the stopped vehicle at an unsafe distance and speed.

[0080] At step 314, the edge device 200 may determine a risk that other vehicles will pass the stopped vehicle at an unsafe distance and speed by determining position of stopped vehicle relative to a traffic lane. The relative position of the stopped vehicle may indicate whether the vehicle is parked on the road or on the side of the road.

[0081] At step 318, the edge device 200 may determine whether the vehicle is stopped in the traffic lane.

[0082] Upon determining that the vehicle is stopped in a traffic lane, at step 320, the edge device 200 may determine whether the vehicle is waiting at an intersection. The process may terminate upon determining that the vehicle is waiting at the intersection.

[0083] Upon determining that the vehicle is not waiting at the intersection, at step 324, the edge device 200 may determine whether the vehicle path is obstructed by processing the visual data. Upon determining that the vehicle path is not obstructed, the edge device 200 may retrieve additional sensor data that includes a check engine light indicator, a vehicle maintenance status indicator. The edge device 200 may generate an unsafe parking alert to be provided to the driver of the vehicle if there is no indication from the additional sensor data that the vehicle is stopped because of a breakdown.

[0084] Upon determining that the path of stopped vehicle is obstructed, at step 328, the edge device 200 may determine the causes of path obstruction. The path of the stopped vehicle may be obstructed if there is traffic congestion on lane in which the ego vehicle is moving, which may lead to the vehicle stoppage but cannot be considered as unsafe parking. In this case, alerts are not generated. In some cases, the vehicle may have been stopped because of its involvement in a collision. In this case, the vehicle’s path may be obstructed by an object (e.g., other vehicles, wall, traffic sign, divider) involved in collision. In addition to the visual data, the edge device 200 may process additional sensor data to determine that the vehicle may have been involved in a collision. In this case, the edge device 200 may generate a potential collision alert to be provided to safety fleet manager or emergency services.

[0085] FIG. 3C illustrates a third scenario for unsafe parking detection of a vehicle in which the edge device is installed, according to an embodiment of the present disclosure.

[0086] At step 302, the edge device 200 may detect that the vehicle is stopped. In one example, the edge device 200 may determine that the vehicle has stopped for more than a predefined time period. The determination of the vehicle stoppage is based on one or more of: speed from a GPS sensor, engine speed from vehicle data, inertial sensors, visual odometry, or visual consistency / motion vectors.

[0087] At step 310, the edge device 200 may process visual data captured by a camera module associated with the edge device 200. The edge device 200 may determine a risk that other vehicles will pass the stopped vehicle at an unsafe distance and speed.

[0088] At step 314, the edge device 200 may determine a risk that other vehicles will pass the stopped vehicle at an unsafe distance and speed by determining position of stopped vehicle relative to a traffic lane. The relative position of the stopped vehicle may indicate whether the vehicle is parked on the road or on the side of the road.

[0089] At step 316, the edge device 200 may determine road type based on scene and / or object classifier. The road type along with position of the stopped vehicle may be used to determine the risk that other vehicles will pass the stopped vehicle at an unsafe distance and speed. The edge device 200 may assess one or more of the following: a number of lanes of the road, whether travel is one way or two way, whether there is a visible intersection, whether there was a visible intersection within a threshold period of time, based on the visual data to determine the road type. Further, the edge device 200 may classify whether there is at least one of a sidewalk near the road or a shoulder on the road based at least on the visual data. The scene classifier may be configured to classify visual data as corresponding to one or more road types for which parking is unsafe and one or more road types for which parking is safe. The one or more road types are highway, on ramp, and off ramp for which the parking is unsafe. The obj ect classifier may be configured to detect one or more objects in visual data. The one or more objects include parking signs. The road type of a road may be determined based on visual data of the road on which the vehicle was travelling prior to the stop.

[0090] At step 318, the edge device 200 may determine whether the vehicle is stopped in the traffic lane.

[0091] Upon determining that the vehicle is not stopped in the traffic lane, at step 322, the edge device 200 may determine whether the parking is safe or unsafe given the determined road type from step 316.

[0092] Upon determining that the parking is unsafe given the road type, at step 330, the edge device 200 may generate an unsafe parking alert to be provided to the driver.

[0093] Upon determining that the parking is safe given the road type, at step 326, the edge device 200 may determine whether side-of-road parking is prohibited based on the road type.

[0094] Upon determining that side-of-road parking is prohibited for the road type, at step 330, the edge device 200 may generate an unsafe parking alert to be provided to the driver. Upondetermining that side-of-road parking is not prohibited for the road type, the unsafe parking detection process may end.

[0095] FIG. 3D illustrates a fourth scenario for unsafe parking detection of a vehicle in which the edge device is installed, according to an embodiment of the present disclosure.

[0096] At step 302, the edge device 200 may detect that the vehicle is stopped. In one example, the edge device 200 may determine that the vehicle has stopped for more than a predefined time period. The determination of the vehicle stoppage is based on one or more of: speed from a GPS sensor, engine speed from vehicle data, inertial sensors, visual odometry, or visual consistency / motion vectors.

[0097] At step 310, the edge device 200 may process visual data captured by a camera module associated with the edge device 200. The edge device 200 may determine a risk that other vehicles will pass the stopped vehicle at an unsafe distance and speed.

[0098] At step 316, the edge device 200 may determine road type based on scene and / or object classifier to determine the risk. The edge device 200 may assess one or more of the following: a number of lanes of the road, whether travel is one way or two way, whether there is a visible intersection, whether there was a visible intersection within a threshold period of time, based on the visual data to determine the road type. Further, the edge device 200 may classify whether there is at least one of a sidewalk near the road or a shoulder on the road based at least on the visual data. The scene classifier may be configured to classify visual data as corresponding to one or more road types for which parking is unsafe and one or more road types for which parking is safe. The one or more road types are highway, on ramp, and off ramp for which the parking is unsafe. The object classifier may be configured to detect one or more objects in visual data. The one or more objects include parking signs. The road type of a road may be determined based on visual data of the road on which the vehicle was travelling prior to the stop.

[0099] At step 322, the edge device 200 may determine whether the parking is safe or unsafe given the determined road type from step 316.

[0100] Upon determining that the parking is unsafe given the road type, at step 330, the edge device 200 may generate an unsafe parking alert to be provided to the driver.

[0101] Upon determining that the parking is safe given the road type, at step 326, the edge device 200 may determine whether side-of-road parking is prohibited based on the road type.

[0102] Upon determining that side-of-road parking is prohibited for the road type, at step 330, the edge device 200 may generate an unsafe parking alert to be provided to the driver. Upon determining that side-of-road parking is not prohibited for the road type, the unsafe parking detection process may end.

[0103] FIG. 3E illustrates a fifth scenario for unsafe parking detection of a vehicle in which the edge device is installed, according to an embodiment of the present disclosure.

[0104] At step 302, the edge device 200 may detect that the vehicle is stopped. In one example, the edge device 200 may determine that the vehicle has stopped for more than a predefined time period. The determination of the vehicle stoppage is based on one or more of: speed from a GPS sensor, engine speed from vehicle data, inertial sensors, visual odometry, or visual consistency / motion vectors.

[0105] At step 310, the edge device 200 may process visual data captured by a camera module associated with the edge device 200. The edge device 200 may determine a risk that other vehicles will pass the stopped vehicle at an unsafe distance and speed.

[0106] At step 316, the edge device 200 may determine road type based on scene and / or object classifier to determine the risk. The edge device 200 may assess one or more of the following: a number of lanes of the road, whether travel is one way or two way, whether there is a visible intersection, whether there was a visible intersection within a threshold period of time, based on the visual data to determine the road type. Further, the edge device 200 may classify whether there is at least one of a sidewalk near the road or a shoulder on the road based at least on the visual data. The scene classifier may be configured to classify visual data as corresponding to one or more road types for which parking is unsafe and one or more road types for which parking is safe. The one or more road types are highway, on ramp, and off ramp for which the parking is unsafe. The object classifier may be configured to detect one or more objects in visual data. The one or more objects include parking signs. The road type of a road may be determined based on visual data of the road on which the vehicle was travelling prior to the stop.

[0107] At step 322, the edge device 200 may determine whether the parking is safe or unsafe given the determined road type from step 316.

[0108] Upon determining that the parking is unsafe given the road type, at step 324, the edge device 200 may determine whether the vehicle path is obstructed.

[0109] Upon determining that the vehicle path is obstructed, at step 328, the edge device 200 may determine the causes of path obstruction. The path of the stopped vehicle may be obstructed if there is traffic congestion on lane in which the ego vehicle is moving, which may lead to the vehicle stoppage but cannot be considered as unsafe parking. In this case, alerts are not generated. In some cases, the vehicle may have been stopped because of its involvement in a collision. In this case, the vehicle’s path may be obstructed by an object (e.g., other vehicles, wall, traffic sign, divider) involved in collision. In addition to the visual data, the edge device 200 may process additional sensor data to determine that the vehicle may have been involved in a collision. In this case, the edge device 200 may generate a potential collision alert to be provided to safety fleet manager or emergency services.

[0110] Upon determining that vehicle path is not obstructed, at step 330, the edge device 200 may generate an unsafe parking alert to be provided to the driver.

[0111] Upon determining that the parking is safe given the road type, at step 326, the edge device 200 may determine whether side-of-road parking is prohibited based on the road type.

[0112] Upon determining that side-of-road parking is prohibited for the road type, at step 330, the edge device 200 may generate an unsafe parking alert to be provided to the driver. Upon determining that side-of-road parking is not prohibited for the road type, the unsafe parking detection process may end.

[0113] FIG. 3F illustrates a sixth scenario for unsafe parking detection of a vehicle in which the edge device is installed, according to an embodiment of the present disclosure. FIG. 3F is a combination of all the scenarios of FIGS. 3A-3E, which includes the steps 302-330.

[0114] As used herein, the term “determining” encompasses a wide variety of actions. For example, “determining” may include calculating, computing, processing, deriving, investigating, looking up (e.g., looking up in a table, a database or another data structure), ascertaining and the like. Additionally, “determining” may include receiving (e.g., receivinginformation), accessing (e.g., accessing data in a memory) and the like. Furthermore, “determining” may include resolving, selecting, choosing, establishing and the like.

[0115] As used herein, a phrase referring to “at least one of’ a list of items refers to any combination of those items, including single members. As an example, “at least one of: a, b, or c” is intended to cover: a, b, c, a-b, a-c, b-c, and a-b-c.

[0116] The methods disclosed herein comprise one or more steps or actions for achieving the described method. The method steps and / or actions may be interchanged with one another without departing from the scope of the claims. In other words, unless a specific order of steps or actions is specified, the order and / or use of specific steps and / or actions may be modified without departing from the scope of the claims.

[0117] The processing system may be configured as a general-purpose processing system with one or more microprocessors providing the processor functionality and external memory providing at least a portion of the machine-readable media, all linked together with other supporting circuitry through an external bus architecture. Alternatively, the processing system may comprise one or more specialized processors for implementing the neural networks, for example, as well as for other processing systems described herein.

[0118] Thus, certain aspects may comprise a computer program product for performing the operations presented herein. For example, such a computer program product may comprise a computer-readable medium having instructions stored (and / or encoded) thereon, the instructions being executable by one or more processors to perform the operations described herein. For certain aspects, the computer program product may include packaging material.

[0119] Further, it should be appreciated that modules and / or other appropriate means for performing the methods and techniques described herein can be downloaded and / or otherwise obtained by a user terminal and / or base station as applicable. For example, such a device can be coupled to a server to facilitate the transfer of means for performing the methods described herein. Alternatively, various methods described herein can be provided via storage means (e.g., RAM, ROM, a physical storage medium such as a compact disc (CD) or floppy disk, etc.), such that a user terminal and / or base station can obtain the various methods upon coupling or providing the storage means to the device. Moreover, any other suitable technique for providing the methods and techniques described herein to a device can be utilized.

[0120] It is to be understood that the claims are not limited to the precise configuration and components illustrated above. Various modifications, changes and variations may be made in the arrangement, operation and details of the methods and apparatus described above without departing from the scope of the claims.

Claims

CLAIMS1. A computer-implemented method for detecting unsafe parking, the method comprising: detecting (302), by at least one processor, that a vehicle is stopped; receiving, by the at least one processor, visual data captured by an outward-facing camera of the vehicle; processing (310), by the at least one processor, the visual data to determine a risk that other vehicles will pass the stopped vehicle at an unsafe distance and speed; and responsive to detecting, by the at least one processor, that the risk is above a threshold, generating (330), by the at least one processer, an alert for a driver of the vehicle.

2. The method of claim 1, wherein a first processor of the at least one processor is housed in an edge device installed in the vehicle; and wherein processing the visual data to determine the risk that other vehicles will pass the stopped vehicle at an unsafe distance and speed is based on processing of the first processor.

3. The method of claim 1, wherein processing (310) the visual data to determine the risk that other vehicles will pass the stopped vehicle at an unsafe distance and speed comprises: detecting (312) one or more vehicles passing the vehicle at a close distance and a high speed.

4. The method of claim 3, wherein detecting (312) the one or more vehicles at a close distance and a high speed comprises: detecting a passing vehicle; determining a lateral distance between the vehicle and the passing vehicle; and determining a speed of the passing vehicle.

5. The method of claim 4, wherein detecting (312) the one or more vehicles at a close distance and a high speed further comprises: comparing the lateral distance with a predefined distance threshold; and comparing the speed of the passing vehicle with a predefined speed threshold; and wherein the alert is generated (330) if the lateral distance is below the predefined distance threshold and the speed of the passing vehicle is above the predefined speed threshold.

6. The method of claim 3, wherein processing the visual data to determine the risk that other vehicles will pass the stopped vehicle at an unsafe distance and speed comprises determining a frequency of vehicles passing the stopped vehicle.

7. The method of claim 6, further comprising: modifying a magnitude of the risk based on the determined frequency, such that a higher frequency corresponds to a greater likelihood that the alert will be generated.

8. The method of claim 6, further comprising: modifying a severity of the alert based on the determined frequency.

9. The method of claim 3, wherein detecting (312) the one or more vehicles at a close distance and a high speed comprises: calculating a distance in the visual data from a center of a calibrated ego lane location to a bottom center of a bounding box or a nearby bottom corner of the bounding box for the one or more passing vehicles.

10. The method of claim 1, wherein processing the visual data to determine the risk that other vehicles will pass the stopped vehicle at an unsafe distance and speed comprises: determining (314) a position of the stopped vehicle relative to a traffic lane of a road.

11. The method of claim 10, wherein determining (314) the position of the stopped vehicle relative to the traffic lane of the road comprises: determining (318) if the vehicle is stopped in the traffic lane of the road.

12. The method of claim 11, further comprising: in response to determining (318) that the vehicle is stopped in the traffic lane of the road, determining (320) if the vehicle is waiting at an intersection or if a path of the vehicle is obstructed (324); and in response to determining that the vehicle is not waiting in an intersection and the path of the vehicle is not obstructed, generating the alert.

13. The method of claim 11, further comprising: in response to determining that the vehicle is not stopped in the traffic lane of the road, determining a road type of the road; and in response to determining that roadside parking is unsafe given the road type, generating the alert.

14. The method of claim 10, wherein determining (314) the position of the stopped vehicle relative to the traffic lane of the road comprises determining the vehicle’s distance off the road.

15. The method of claim 1, wherein processing the visual data to determine the risk that other vehicles will pass the stopped vehicle at an unsafe distance and speed comprises: determining (316) a type of road using a scene classifier, an object classifier, or some combination thereof.

16. The method of claim 15, wherein determining (316) the type of road comprises assessing, based on the visual data, one or more of: a number of lanes of the road, whether travel is one way or two way, whether there is a visible intersection, whether there was a visible intersection within a threshold period of time.

17. The method of claim 15, wherein determining (316) the type of road comprises classifying whether there is at least one of a sidewalk near the road or a shoulder on the road based at least on the visual data.

18. The method of claim 15, wherein the scene classifier is configured to classify visual data as corresponding to one or more road types for which parking is unsafe and one or more road types for which parking is safe; wherein the one or more road types for which parking is unsafe comprises: highway, on ramp, and off ramp.

19. The method of claim 15, wherein the object classifier is configured to detect one or more objects in visual data, wherein the one or more objects comprises parking signs.

20. The method of claim 15, wherein determining a type of road comprises:determining a road type of a road on which the vehicle was travelling prior to the stop.

21. The method of claim 1, wherein detecting that the vehicle has stopped is based on one or more of: speed from a GPS sensor, engine speed from vehicle data, inertial sensors, visual odometry, or visual consistency / motion vectors.

22. The method of claim 1, further comprising: receiving additional sensor data from the vehicle; wherein the additional sensor data comprises at least one of a check engine light indicator, a vehicle maintenance status indicator; processing the additional sensor data for a potential collision event; and generating a potential collision alert in response to a determination that the vehicle may have been involved in a collision.

23. The method of claim 1, wherein detecting that the vehicle is stopped further comprises detecting that the vehicle has been stopped for at least a threshold amount of time.

24. The method of claim 1, wherein processing (310) the visual data to determine the risk that other vehicles will pass the stopped vehicle at an unsafe distance and speed comprises: detecting a second vehicle at a close distance to the vehicle; determining that the risk is below a threshold based on a determination that the second vehicle is not moving or is moving slowly.

25. The method of claim 24, wherein the determination that the second vehicle is not moving or is moving slowly is based on an inferred bounding box around the back of the second vehicle.

26. The method of claim 24, further comprising: determining that the vehicle is in a line for a toll booth or a vehicle inspection station.

27. The method of claim 1, wherein detecting that the vehicle is stopped comprises: processing, by the at least one processor, a first time series of speed data from the vehicle to determine that the vehicle was travelling at a high rate of speed at the start of the first time series and that the vehicle is stopped at the end of the first time series; the method further comprising:processing, by the at least one processor, a second time series of speed data from the vehicle, wherein the second time series is a subset of the first time series, to determine whether the vehicle exhibited a rise and fall of speed within the second time series; and suppressing the alert in response to a determination that the vehicle exhibited a rise and fall of speed within the second time series.

28. The method of claim 27, wherein processing the visual data to determine a risk that other vehicles will pass the stopped vehicle at an unsafe distance and speed comprises determining that the vehicle is parked safely in an area adjacent to a parking lot.

29. The method of claim 1, further comprising: processing, by the at least one processor, a series of distance data during the first time series, wherein each distance data of the series corresponds to a distance from a road; determining, by the at least one processor, that a final distance from the road of the series of distance data is within a threshold distance; and suppressing the alert is response to a determination that an intermediate distance from the road of the series of distance data is beyond the threshold distance.

30. A system for detecting unsafe parking, the system comprising one or more processors coupled to a non-transitory memory, the one or more processors configured to perform the steps of the method of any of claims 1 to 29.

Images