Animal collision-based risk warning method, vehicle device, and electronic device

Abstract

The application provides a risk early warning method based on animal collision, a vehicle device and an electronic device. The method predicts the expected behavior of an animal target according to the information of the animal target and environmental information, obtains a risk increasing factor in the information of the animal target and / or the expected behavior, and is used for evaluating the collision risk level. In one aspect, the different risk increasing factors in the information of the animal target and / or the expected behavior can be evaluated differently. In another aspect, the collision risk evaluation is not only based on the current state, but also combines the expected behavior of the animal target, so that the collision risk can be predicted in advance, and the accuracy of the collision risk evaluation is improved. Different early warning methods are used for early warning according to different collision risk levels, so that the driver can intuitively perceive the collision risks of different degrees, and the early warning efficiency is improved.

Description

Technical Field

[0001] This application relates to the field of driver assistance technology, and in particular to a risk warning method, vehicle equipment, and electronic equipment based on animal collisions. Background Technology

[0002] With the continuous expansion of road networks, the risk of collisions between vehicles and wild animals is increasing. Such accidents not only cause direct economic losses and casualties, but also pose a sustained threat to biodiversity conservation. Therefore, timely and accurate early warning of potential animal collision risks has become a crucial element in assisting drivers' decision-making and improving road safety.

[0003] Currently, mainstream collision warning technologies primarily rely on visual sensors to identify animal targets and trigger alarms by calculating the time-to-collision (TTC) between the animal and the vehicle. However, this method offers a relatively singular assessment dimension. In real-world road scenarios, different species exhibit significant differences in size, behavior, and movement patterns, meaning the level of danger they pose and the necessary response strategies can vary drastically. For example, the appearance of an individual gregarious animal often foreshadows the herd following; similarly, flashing high beams at easily startled animals like wild boars may actually trigger dangerous behavior such as them ramming into vehicles. Simply relying on TTC parameters makes it difficult to conduct a refined and differentiated risk assessment of these complex factors, potentially leading to untimely warnings or unnecessary, disruptive alerts. Summary of the Invention

[0004] To address the existing technical problems, this application provides a risk warning method, vehicle equipment, and electronic equipment based on animal collisions that enables differentiated assessment and effective early warning.

[0005] Firstly, a risk warning method based on animal collisions is provided, the method comprising: Based on the animal recognition results of the image sequence, information about the animal target in the image is obtained; Acquire environmental information and, based on the information of the animal target and the environmental information, predict the expected behavior of the animal target; Calculate the collision time of the animal target, and determine the collision risk level based on the collision time, information about the animal target, and / or risk-increasing factors in the expected behavior; The warning information is output using a warning method that matches the collision risk level.

[0006] In a second aspect, a vehicle device is provided, including a sensor module, a processor, a display, a speaker, and a memory connected to the processor. The memory stores a computer program that can be executed by the processor. When the computer program is executed by the processor, it implements the steps of the risk warning method based on animal collision as described in the above embodiments.

[0007] Thirdly, an electronic device is provided, including a sensor module, a processor, and a memory connected to the processor. The memory stores a computer program that can be executed by the processor. When the computer program is executed by the processor, it implements the steps of the risk warning method based on animal collision as described in the above embodiments.

[0008] This animal collision-based risk warning method predicts the expected behavior of the animal target based on information about the animal and the environment. It then uses the information about the animal target and / or the risk-increasing factors in its expected behavior to assess the collision risk level. On one hand, it allows for differentiated assessments based on different risk-increasing factors in the animal target's information and / or expected behavior. On the other hand, the collision risk assessment is not only based on the current situation but also incorporates the expected behavior of the animal target, enabling early prediction of collision risks and improving the accuracy of the assessment. Differentiated warning methods are used for different collision risk levels, allowing drivers to intuitively perceive different degrees of collision risk and improving warning efficiency.

[0009] The vehicle equipment and electronic equipment provided in the above embodiments belong to the same concept as the corresponding animal collision-based risk warning method embodiments, and thus have the same technical effects as the corresponding animal collision-based risk warning method embodiments, which will not be repeated here. Attached Figure Description

[0010] Figure 1 This is a structural block diagram of the vehicle equipment in one embodiment.

[0011] Figure 2 This is a flowchart of a risk warning method based on animal collisions in one embodiment.

[0012] Figure 3 This is a flowchart of the steps involved in constructing an animal behavior knowledge base in one embodiment.

[0013] Figure 4 This is a schematic diagram of the structured information of behavior preset rules in a behavior knowledge base in one embodiment.

[0014] Figure 5 This is a structural block diagram of the vehicle equipment in one embodiment.

[0015] Figure 6This is a flowchart of the early warning data in one embodiment.

[0016] Figure 7 This is a flowchart illustrating the specific judgment and early warning processes of the risk assessment and decision-making module in one embodiment.

[0017] Figure 8 This is a schematic diagram of a visual warning for the fourth collision risk level in one embodiment. Detailed Implementation

[0018] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] In the following description, the phrase "some embodiments" refers to a subset of all possible embodiments. It should be noted that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict.

[0021] In the following description, the terms "first, second, and third" are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first, second, and third" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0022] A first aspect of this application provides a vehicle device, see reference Figure 1 The vehicle equipment 10 includes a display terminal 11, a sensor module 12, a controller 13, and a speaker 14. The sensor module 12 collects data during the vehicle equipment 10's operation. The controller 13, based on the image data collected by the sensor module 12 and the motion data of the vehicle equipment 10, implements a risk warning method based on animal collisions. Specifically, the controller 13 includes a processor and a memory connected to the processor. The memory stores a computer program executable by the processor. When the computer program is executed by the processor, it implements a risk warning method based on animal collisions.

[0023] Specifically, the risk warning method based on animal collisions includes: obtaining information about the animal target in the image based on the animal recognition results of the image sequence; acquiring environmental information and predicting the expected behavior of the animal target based on the information of the animal target and the environmental information; calculating the collision time of the animal target and determining the collision risk level based on the collision time, as well as risk-increasing factors in the information of the animal target and / or the expected behavior; and outputting warning information using a warning method matching the collision risk level. The warning information can be displayed on the display terminal 11 and / or played through the speaker 14.

[0024] This vehicle-mounted device predicts the expected behavior of an animal target based on information about the animal and its environment. It then uses this information and / or risk-increasing factors within the expected behavior to assess the collision risk level. On one hand, it can perform differentiated assessments based on different risk-increasing factors in the animal target's information and / or expected behavior. On the other hand, the collision risk assessment is not only based on the current situation but also incorporates the expected behavior of the animal target, enabling early prediction of collision risks and improving the accuracy of the assessment. Differentiated warning methods are used to address different collision risk levels, allowing drivers to intuitively perceive varying degrees of collision risk and improving warning efficiency.

[0025] The sensor module 12 can be a module integrating one or more sensors, and there can be multiple sensor modules 12, which can be installed in different locations on the vehicle equipment 10. The sensor module 12 includes, but is not limited to, multispectral vision sensors, environmental perception sensors, and motion attitude sensors. The multispectral sensor includes, but is not limited to, a combination of one or more of the following sensors: thermal imaging sensors, visible light image sensors, millimeter-wave sensors, lidar sensors, infrared thermal imaging sensors, and depth sensors. The environmental perception sensor includes, but is not limited to, a combination of one or more of the following environmental sensors: brightness sensors, temperature sensors, fog sensors, etc. The motion attitude sensor includes, but is not limited to, a combination of one or more of the following: inertial measurement unit (IMU), velocity sensors, acceleration sensors, gyroscope sensors, geomagnetic sensors, rotation vector sensors, steering wheel angle sensors, level sensors, tilt sensors, vibration sensors, displacement sensors, and gravity sensors, etc.

[0026] The screen is the surface on which the display terminal 11 displays images. The screen is positioned in front of the driver so that the driver can view the screen. The display terminal 11 includes, but is not limited to, a display such as an LCD screen, a projector with a projector screen, etc. The display terminal 11 can be a central display on the vehicle equipment 10, or a projector located at the rear of the vehicle equipment 10 to project images onto the projector screen.

[0027] The controller 13 can be one or more. When there are multiple controllers 13, they can be integrated on a single chip or independently located on each chip. The vehicle device 10 can be a device installed on any type of mobile body, such as a vehicle, electric vehicle, hybrid electric vehicle, motorcycle, bicycle, personal mobile device, airplane, drone, boat, or robot, etc. In some embodiments, the controller 13 and the display terminal 11 can be integrated.

[0028] The image acquisition device can be a combination of one or more sensor modules 12. The image acquisition device can be a monocular vision sensor or a multi-view vision sensor. For example, it can be a combination of one or more sensors such as a thermal imaging sensor, a visible light image sensor, a millimeter-wave sensor, a lidar sensor, an infrared thermal imaging sensor, and a depth sensor.

[0029] Common driver assistance sensors include visible light, infrared, millimeter-wave radar, and lidar. Visible light is easily affected by ambient light. LiDAR is unaffected by light, but it still misses many small targets and may fail in rain or snow. Millimeter-wave radar has low spatial resolution and insufficient target classification capabilities. Infrared thermal imaging, on the other hand, detects the thermal radiation of objects, is unaffected by light, has a longer detection range at night, can penetrate fog, and has high resolution. Therefore, as an alternative implementation, infrared thermal imaging sensors are used to solve the animal collision warning system in special scenarios such as nighttime, low light, and fog.

[0030] Another aspect of this application provides an electronic device, which includes a sensor module, a processor, and a memory connected to the processor. The memory stores a computer program that can be executed by the processor. When the computer program is executed by the processor, it implements the steps of a risk warning method based on animal collision.

[0031] Specifically, the risk warning method based on animal collisions includes: obtaining information about the animal target in the image based on the animal recognition results of the image sequence; acquiring environmental information and predicting the expected behavior of the animal target based on the information of the animal target and the environmental information; calculating the collision time of the animal target and determining the collision risk level based on the collision time, as well as risk-increasing factors in the information of the animal target and / or the expected behavior; and outputting warning information using a warning method matching the collision risk level. The electronic device can be a personal mobile device, drone, transportation equipment, or outdoor robot, etc.

[0032] In one embodiment, a risk warning method based on animal collisions is provided, which is applied to, for example... Figure 1 The controller of the vehicle equipment shown. (As...) Figure 2 As shown, this risk warning method based on animal collisions includes: Step 202: Based on the animal recognition results of the image sequence, obtain information about the animal target in the image.

[0033] In this embodiment, the image acquisition device is positioned in front of the vehicle equipment 10 to acquire image data of the area in front of the vehicle during operation. The image acquisition device can be at least one of a visible light camera and an infrared camera. The image sequence here refers to the visible light image sequence, the infrared image sequence, or a fused image sequence of both acquired by the image acquisition device of the vehicle equipment.

[0034] By identifying animal targets in an image, information about the animal targets can be determined. In this embodiment, an object detection algorithm can be used to identify animal targets from image data and obtain detection data of the animal targets. The identified animal target can be one or more, and the detection data includes, but is not limited to: the coordinates of the center point of the detection box, the size of the detection box, the length, width, and height of the target, etc. The object detection model algorithm includes, but is not limited to, MonoCon-based object detection model algorithms, SMOKE-based monocular object detection model algorithms, neural network-based object detection algorithms, etc.

[0035] In one optional implementation, a neural network-based object detection algorithm is used as an example to describe the object detection algorithm. Performing object detection on the image data of the current frame using this algorithm includes: pre-collecting an animal sample dataset, which includes animal samples of various categories, non-animal samples, and corresponding category labels for each sample. The animal samples of various categories include image samples of animals of the same category in various poses. For example, the animal samples of various categories include, but are not limited to, image data of wild boars, elk, coyotes, foxes, sheep, bison, horses, and other animals in various poses. The non-animal samples include, but are not limited to, image data of cars, buses, other vehicles, pedestrians, two-wheeled vehicles, three-wheeled vehicles, and other vehicles in various forms. The category label corresponding to each sample is the label for that sample; for example, the label for a fox is "animal," and the label for a pedestrian is "non-animal."

[0036] The target detection model is trained using animal sample datasets as input data until it reaches the training termination condition. During training, the model uses the labels from the training dataset as its training target, learning the features of animal and non-animal samples in the dataset. This allows it to identify animal and non-animal targets after training.

[0037] Understandably, single-type image acquisition devices have limitations. For example, the performance of visible light cameras drops sharply at night, dusk, dawn, or in inclement weather. While a single infrared camera can detect heat, it lacks texture and color information about the target, making it difficult to accurately distinguish animal species (such as wild boar, elk, sheep, coyotes, and cattle) and their postures. As a result, the system can usually only classify such targets as obstacles and cannot provide targeted warnings.

[0038] To address this issue, in a preferred embodiment, a visible light and long-wave infrared camera with hardware time synchronization is used. The infrared and visible light images are fused using a pixel-level fusion algorithm to obtain a fused image. Then, a target detection algorithm is used to detect and classify the target in the fused image, identifying information such as the animal target's category (e.g., wild boar, elk, sheep, coyote, cattle), age (adult or juvenile), initial position, posture (e.g., standing, running, crouching), and environment.

[0039] Based on this, animal recognition results from image sequences are used to track animal targets using target tracking algorithms, and the tracking data is recorded to obtain the animal's movement trajectory and posture changes. The tracking data includes detection data of the animal target in each frame of historical frames. Target tracking algorithms include, but are not limited to, target tracking based on mean-shift algorithm, target tracking based on Kalman filter, target tracking based on particle filter, etc.

[0040] Step 204: Obtain environmental information and predict the expected behavior of the animal target based on the information of the animal target and the environmental information.

[0041] Environmental information includes geographic environmental information and temporal environmental information. Temporal environmental information includes seasons, day / night cycles, etc. Geographic environmental information includes geographical location, topography, etc.

[0042] In one embodiment, geographic and temporal environmental information can be obtained based on other information systems of the vehicle equipment. For example, geographic environmental information can be obtained based on the vehicle's positioning system, and temporal environmental information can be obtained based on a time system. Furthermore, image recognition technology can be combined to assist in determining geographic environmental information.

[0043] The inventors of this application discovered that different animals exhibit significant differences in body size, habits, and movement patterns, resulting in varying levels of danger and corresponding response strategies. For example, the presence of an individual gregarious animal often foreshadows a following group; conversely, flashing high beams as a warning to easily startled animals like wild boars may actually trigger dangerous behavior such as them ramming into vehicles. Therefore, when assessing animal collision risk, in addition to considering factors such as the time of collision, it is crucial to conduct advance assessments based on the behavioral characteristics of different animals to improve the accuracy of collision judgments. Specifically, if a gregarious animal is identified and a following group is predicted, the collision risk increases accordingly; similarly, if an easily startled animal such as a wild boar is identified, using inappropriate high beam warnings may also increase the collision risk.

[0044] Based on the above findings, this application predicts the expected behavior of identified animal targets in front of a vehicle, using information about the animal target and the environment. Expected behavior refers to the predicted action the animal target might take next, based on current observations and its behavioral habits. For example, if the identified animal is a elk, the time is dusk or dawn, and the environment is a road near a forest, its expected behavior might be to cross the road. If the animal is a coyote, and its posture is detected as crouching, its expected behavior might be to prepare to charge.

[0045] In this embodiment, the expected behavior of the animal target is predicted based on the information of the animal target and the environmental information, and the expected behavior is used as one of the factors in the collision risk assessment. This makes the collision risk assessment not only based on the current state, but also combined with the predicted behavior of the animal target. That is, the impact of the moving target's next behavior on the collision risk is assessed, thereby improving the accuracy of the collision risk assessment.

[0046] Step 206: Calculate the collision time of the animal target and determine the collision risk level based on the collision time, as well as the information of the animal target and / or risk-increasing factors in the expected behavior.

[0047] Risk-increasing factors refer to those factors in the information about the animal target and the expected behavior of the animal target that may increase the risk of collision. For example, the type of animal target is easily startled, and the expected behavior of the animal target is to cross over, etc., are factors that may increase the risk of collision.

[0048] In one embodiment, risk-increasing factors in the information and / or expected behavior of the animal target can be preset, and the identified information and / or expected behavior of the animal target can be matched with the preset risk-increasing factors to determine whether there are preset risk-increasing factors in the information and / or expected behavior of the animal target.

[0049] In one embodiment, risk-increasing factors associated with the expected behavior include: the animal target exhibiting path conflict behavior; and / or, risk-increasing factors associated with information about the animal target include at least one of: the animal target being a stressful animal, the animal target being in motion, the animal target being a social animal, and the animal target being an offspring.

[0050] Specifically, based on the relative motion between the animal target and the vehicle, the collision time is calculated, and this collision time corresponds to a baseline risk value. Then, based on factors that increase the risk associated with the animal target, the baseline risk value is adjusted to determine the final risk level.

[0051] It should be understood that as an animal's movement causes changes in its own state, and as the vehicle's movement causes changes in the relative distance and position to the animal target, the collision time and factors increasing risk may change accordingly. Therefore, the collision risk level of an animal target is also dynamic. For example, when an animal target is far from the vehicle and is stationary, its collision risk level is low. However, when the expected behavior of the same animal target changes to path intrusion, its collision risk increases. The collision risk level is highest when the collision time is less than the risk threshold.

[0052] Step 208: Output warning information using a warning method that matches the collision risk level.

[0053] In this embodiment, different warning methods can be used based on the collision risk assessment results. Specifically, in assisted driving scenarios, warning information is output using a warning method that matches the collision risk level. Assuming there are four risk levels: low, medium, high, and emergency, in assisted driving scenarios, low-risk levels can only be indicated by an icon, while medium-risk levels can issue mild visual and audible warnings, such as displaying a yellow animal icon on the screen with a mild alert sound. High-risk levels can issue strong visual and audible warnings, such as a red warning on the screen and an urgent alert sound. Emergency risk levels can issue the highest priority audible and visual alarms.

[0054] The aforementioned animal collision-based risk warning method predicts the expected behavior of the animal target based on information about the animal and the environment. It then uses the information about the animal target and / or the risk-increasing factors in its expected behavior to assess the collision risk level. On one hand, it allows for differentiated assessments based on different risk-increasing factors in the animal target's information and / or expected behavior. On the other hand, the collision risk assessment is not only based on the current situation but also incorporates the expected behavior of the animal target, enabling early prediction of collision risks and improving the accuracy of the assessment. Furthermore, by matching different collision risk levels with differentiated warning methods, drivers can intuitively perceive different degrees of collision risk, improving warning efficiency.

[0055] Existing animal collision-based warning systems also have significant shortcomings in terms of human-computer interaction. Their warning information is mostly a uniform "obstacle" icon or a generic sound and light alarm. This interaction method is too simplistic and abstract, failing to effectively convey crucial cognitive information to the driver such as "what animal is ahead" and "what it might do," making it difficult for the driver to make the most appropriate decision instantly, thus limiting the improvement of human-computer collaboration efficiency.

[0056] To address this issue, in this embodiment, the warning information includes visual warning information and / or voice warning information.

[0057] Based on information about risk-increasing factors and animal targets, visual and / or audible warning messages are determined; the visual and audible warning messages respectively include at least one of the following: prompts for animal targets, warnings about risk-increasing factors, and alert strategies for risk-increasing factors.

[0058] In other words, after determining the collision risk level of the animal target, this embodiment further generates matching visual and / or voice warning information based on the risk-increasing factors and information about the animal target, thereby providing the driver with multi-dimensional risk alerts from both visual and auditory dimensions.

[0059] Specifically, the content of visual and / or voice warning information is directly derived from information about risk-increasing factors and animal targets, ensuring that the warning output is highly matched with the actual dangerous situation. For example, visual and voice warning information may include at least one of the following: category cues for animal targets, warnings about risk-increasing factors, and at least one of the following alert strategies for risk-increasing factors.

[0060] Specifically, category cues for animal targets are used to inform drivers of the specific species or category of the animal target ahead. For example, visual warning information can display specific animal icons (such as deer, wild boar, cow, etc.) that match the recognition results, thus concretizing the abstract "obstacle" into a clear animal category.

[0061] Specifically, warnings about increased risk factors can alert users to the factors that increase the risk of an animal ahead, and what risky behaviors these factors might lead the animal to exhibit. For example, a warning like "Social animal, be careful when following" could be given for social animals. This allows drivers to be aware in advance of the potential risky behaviors of the animal.

[0062] Specifically, warning strategies targeting risk-increasing factors refer to specific and actionable safety recommendations provided to drivers based on identified risk-increasing factors. Their purpose is to guide drivers to proactively reduce or avoid corresponding risks in an appropriate manner. For example, for identified easily startled animals (such as wild boars), the safety warning strategy could explicitly advise, "Please slow down smoothly and avoid honking or flashing lights," thereby preventing stress-induced collision behavior from animals due to improper interaction.

[0063] In this way, drivers can intuitively obtain the following key information: what type of animal is ahead, what factors might increase the risk of collision, and what response strategy is recommended for the current situation. This effectively conveys key cognitive information to the driver, such as "what animal is ahead" and "what it might do." This not only enriches the driver's cognitive dimensions regarding animal targets ahead but also provides clear guidance for making accurate and reasonable driving decisions in a short time, thereby significantly improving the efficiency of human-machine collaboration and driving safety.

[0064] In one embodiment, a warning method matching the collision risk level is used to output warning information, including: Visual warning information is output using visual parameters that match the collision risk level, and / or, voice warning information is output using sound parameters that match the collision risk level; The intensity of both visual and auditory parameters increases with the level of collision risk. This allows for differentiated, tiered warnings for different collision risk levels, enabling drivers to intuitively perceive the level of risk and make quick decisions.

[0065] In one embodiment, the visual parameters include at least one of the following: hue, size, flashing state, and flashing frequency of the animal target category identifier. That is, the intensity of at least one of the following: hue, size, flashing state, and flashing frequency of the animal target category identifier increases with the increase of the collision risk level. The animal target category identifier may be an icon representing an animal category.

[0066] For example, risk levels can be categorized from low to high using progressively different hues such as yellow, orange, and red. Drivers can understand the degree of collision risk by observing the color changes of the animal category icon. The icon color can be either the background color or the color of the outer frame of the animal category icon.

[0067] For example, the higher the risk level, the relative size of the label on the screen can be appropriately enlarged, or additional highlighting effects such as illumination or bright outlines can be added.

[0068] For example, at low risk, the sign can remain static; at medium risk, slow flashing can be used; and at high risk to emergency risk, fast or even rapid flashing can be used.

[0069] In one embodiment, the sound parameters include at least one of volume, speed, tone, and frequency. That is, the intensity of at least one of the volume, speed, tone, and frequency of the voice warning information increases as the collision risk level increases.

[0070] For example, as the risk level increases, the volume of the prompt can be appropriately increased, and the tone of the synthesized speech can change from "stable" to "urgent," "serious," or even "urgent."

[0071] For example, when the risk level is low, the voice broadcast is delivered at a slow pace; when the risk level is high, the speech is delivered at a faster pace to convey a sense of urgency.

[0072] For example, different timbres or sound effects can be used to distinguish risk levels, such as transitioning from a soft broadcast voice to a sharp buzzing sound.

[0073] In this embodiment, by mapping abstract collision risk levels to sensory parameters that are intuitively perceptible to the driver and whose intensity changes continuously, the driver does not need to interpret complex data. They can quickly understand the level of urgency of the danger simply by observing icon color, flashing speed, or the urgency of the alarm sound. This shortens the driver's decision-making time.

[0074] In one embodiment, risk-increasing factors associated with the expected behavior include path conflict behavior of the animal's target.

[0075] Path conflict behavior of animal targets refers to behaviors exhibited by an animal target whose expected trajectory conflicts with the vehicle's current or planned path, thereby significantly increasing the risk of collision. Typical path conflict behaviors include crossing the road and approaching from the front.

[0076] Animals' path-conflict behaviors can lead to collisions. This embodiment incorporates predicted path-conflict behaviors into the risk level classification, enabling the prediction of the likelihood of conflict based on the animal's behavioral posture (such as turning towards the road or accelerating) before the animal actually enters the conflict path. This allows for an earlier adjustment of the risk level, gaining valuable early warning and preparation time.

[0077] In one embodiment, risk-increasing factors associated with information about the animal target include at least one of the following: the animal target is a stressful animal, the animal target is in motion, the animal target is a social animal, and the animal target is an animal cub.

[0078] Specifically, stress-prone animals include those that are prone to stress and those that have already exhibited a stress response. Stress-prone animals are those that are startled, tense, or defensive in response to external stimuli (such as approaching vehicles, lights, or noise). Their physiological manifestations may include ruffled fur, vocalizations, stiffness, aggressive postures, or postural changes (such as changing from a lowered head to an raised head). The system pre-classifies animals into stress-prone categories based on their behavioral characteristics. In practical applications, the system determines whether an animal target belongs to the stress-prone category in real time based on the animal's category identified through image recognition. It can also determine whether an animal target has exhibited a stress response based on the physiological manifestations or postural changes of the identified animal target during the application process.

[0079] Animal stress significantly increases the mutability and unpredictability of their behavior. Stressed animals may instantly shift from a state of rest to charging or fleeing, and their direction of movement is difficult to predict, leading to unstable time-to-collision (TTC) estimates and a sharp increase in the probability of collision. In this embodiment, by incorporating the animal's target being a stressful animal into the risk level classification as a risk amplification factor, the risk level can be adjusted upwards in advance, gaining valuable early warning and preparation time.

[0080] An animal target being in motion means that the animal is actively moving, rather than remaining stationary. Its motion state can be further subdivided based on speed and direction (e.g., walking slowly, running, turning). Specifically, the motion state of an animal target can be identified based on the results of trajectory tracking.

[0081] Movement itself implies a change in the animal's trajectory, potentially indicating it is actively cutting into the vehicle's path. Compared to stationary targets, movement directly leads to changes in relative speed and a shortened time-to-travel (TTC), a direct prerequisite for triggering path conflict behavior. In this embodiment, by incorporating the animal's movement as a risk-increasing factor into the risk level classification, the risk level can be adjusted upwards in advance, gaining valuable early warning and preparation time.

[0082] Social animals are those whose species are social and typically live in groups (such as deer, sheep, and cattle). The system pre-classifies animals as social animals based on their social attributes. In practical applications, the system determines the animal category in real-time based on image recognition results, thus determining whether an animal is a social animal.

[0083] In road scenarios, the presence of an individual often indicates the presence of a group. The behavior of social animals is highly contagious and prone to imitation. A single individual, especially a leader, engaging in path-conflicting behavior (such as crossing a path) can trigger the entire group to follow. In this embodiment, by incorporating the animal's target being a social animal into the risk level classification as an amplifying risk factor, the risk level can be raised in advance, gaining valuable early warning and preparation time.

[0084] Animal cubs refer to juvenile individuals, typically small in size and exhibiting immature behavioral patterns. Whether an animal target is an animal cub can be determined based on its size as identified through image recognition.

[0085] Young animals are prone to unpredictable behavior and lack awareness of danger, and may suddenly dart out, further amplifying the risk. In this embodiment, by considering the animal target as a young animal as a risk-increasing factor and incorporating it into the risk level classification, the risk level can be raised in advance, gaining valuable early warning and preparation time.

[0086] In this embodiment, by incorporating factors that increase risk, such as the animal target being a stress-prone animal, the animal target being in motion, the animal target being a social animal, and the animal target being a young animal, into the risk collision level classification, the collision assessment essentially considers not only the collision time but also animal behavioral characteristics, enabling early risk prediction. For example, after identifying the "social animal" attribute, the system can predict the possibility of the group following as soon as the leader shows the intention to cross, thereby increasing the risk level in advance and buying valuable reaction time for the driver or autonomous driving system. Furthermore, different risk-increasing factors correspond to different optimal safety strategies. For example, stimulation should be avoided for "stress-prone animals" (e.g., not honking or flashing lights), while more time and space should be allowed for "social animals" to avoid them. Integrating these risk-increasing factors into the risk level classification facilitates the subsequent generation of warning content targeting these factors.

[0087] In one embodiment, if the animal target is detected to be a social animal, an alert is output for the social animal.

[0088] When a social animal is detected ahead of the vehicle, an alert is generated and output regardless of its current individual behavior or the collision risk level at the time of collision. This alert informs the driver of the potential for further risks due to the group following the animal, guiding the driver to take precautionary measures. For example, the alert could read, "Social animal ahead, be wary of following." Upon receiving this alert, the driver might adopt a driving strategy of maintaining a low speed, continuously observing, and preparing to yield again. This allows sufficient time and space for the group to pass, preventing the driver from prematurely relaxing their vigilance and increasing speed after the first animal has passed, thus avoiding a collision.

[0089] In one embodiment, the collision risk level is determined based on the collision time and information about the animal target and / or risk-increasing factors in the expected behavior, including: if the animal target is a stress-prone animal, then the animal target is determined to have a first collision risk level; The warning information corresponding to the first collision risk level includes at least one of the following: indicating the category of the animal target, warning that the animal target is a stressful animal, and vigilance strategies for stressful animals; the vigilance strategies for stressful animals are used to prompt the adoption of behavioral patterns that reduce animal stress response.

[0090] The system pre-classifies animals into stressor categories based on their behavioral characteristics. During practical application, it uses real-time image recognition to determine whether an animal target is a stressor. It can also assess whether the animal is exhibiting a stress response based on its physiological behavior or posture changes observed during the application.

[0091] It is worth noting that the time of collision (TTC) is not required as a factor in determining the first type of collision risk. That is, once an animal target is detected and identified as a stressed animal, the first collision risk alarm is triggered. This design takes into account that animal stress significantly increases the mutability and unpredictability of its behavior. Stressed animals may instantly shift from a stationary state to charging or fleeing, and their direction of movement is difficult to predict, leading to unstable TTC estimations and a sharp increase in the probability of collision.

[0092] For animals experiencing stress, the output warning information can include indicating the category of the animal target. For example, displaying a specific icon (such as a wild boar icon) corresponding to the stressful animal on the visual interface, or indicating the animal category in the voice broadcast.

[0093] For animals prone to stress, the warning information output can include: an alert indicating that the animal is a stress-prone species. In this way, it can be clearly indicated that the animal is a species that is prone to stress response, such as visually displaying a "frightened animal" label, or verbally announcing "Caution, frightened animal ahead".

[0094] For animals experiencing stress, the output warning information can include: alert strategies for stress-prone animals. These alert strategies are used to prompt the adoption of behavioral patterns that reduce the animal's stress response. Specifically, they are behavioral guidelines to reduce the animal's stress response and avoid escalating risks. For example, visual displays or voice announcements such as: "Please slow down smoothly and avoid honking, flashing high beams, and sudden turns."

[0095] In one specific embodiment, an image of the area in front of the vehicle is acquired, and when a stressed animal is identified in the image, the animal target is determined to belong to the first collision risk level. Visual warning information and voice warning information are then output. The visual warning information may display a specific icon corresponding to the stressed animal on the visual interface. The voice warning information may state, "Please switch to high beams and avoid horn-like alarms."

[0096] In this way, drivers can be guided to adopt driving methods that reduce animal stress response and avoid escalating risks, and avoid inducing or aggravating dangerous animal behavior (such as ramming into vehicles) due to improper interaction.

[0097] The determination of the first collision risk level can provide early warning of the possibility of behavioral changes in stressed animals due to stimulation, and provide warning strategies, effectively reducing the probability of collisions caused by unpredictable animal behavior and enhancing driving safety.

[0098] In one embodiment, the collision risk level is determined based on the collision time, and information about the animal target and / or risk-increasing factors in the expected behavior, including: If the animal target is a young animal that is not moving, then the animal target is determined to have the first collision risk level. The warning information corresponding to the first collision risk level includes at least one of the following: prompts regarding the animal target category, warnings that the animal target is a cub, and alert strategies for cubs.

[0099] Young animals are unpredictable and lack awareness of danger, potentially darting out and amplifying the risk. By capturing images of the vehicle ahead, the system identifies the type, posture, and size of animal targets in the images, tracks their movement, and determines whether the animal is in motion and whether it is a young animal. If a non-moving young animal is identified in the image, it is classified as a Level 1 collision risk.

[0100] It is worth noting that the time of collision is not required as a factor in determining the first collision risk. That is, once an animal target is detected, as long as it is identified as a non-moving fawn, the first collision risk level is determined. This setting takes into account the mutability and unpredictability of fawn behavior. Even if the current collision time is relatively long (appearing lower risk due to the animal being stationary), there is still potential behavioral mutability in the fawn, which could increase the collision risk. For example, under the same TTC (Time to Collision) conditions, a stationary fawn will have a higher collision risk level than a stationary adult deer.

[0101] For animal cubs that are not moving, the output warning information can include indicating the category of the animal target. For example, displaying a specific icon corresponding to the species of the cub (such as a deer icon) on the visual interface, or clearly stating the animal category in the voice broadcast.

[0102] For animal cubs that are not moving, the output warning information can include: an alert indicating that the animal's target is an animal cub. For example, a visual "cub" label can be displayed or a specific color (such as light yellow) can be used to mark it, and a voice announcement can be made saying "Attention, there is an animal cub ahead".

[0103] For inactive animal pups, the warning information output can include: alert strategies for animal pups. These alert strategies can provide drivers with specific behavioral guidance in response to sudden changes in the pups' behavior. For example, visual displays or voice announcements such as, "Caution: Pups, please slow down and remain alert," or "Pups may suddenly dart away, please prepare to avoid them."

[0104] In one specific embodiment, when an image of the area in front of the vehicle is acquired and a non-moving animal cub is detected in the image, the animal target is determined to belong to the first collision risk level. Visual warning information and voice warning information are then output. The visual warning information may display a specific icon of the animal cub on the visual interface. The voice warning information may state, "Caution: There is a cub ahead. Avoid it."

[0105] In this way, drivers can be guided to adopt defensive driving strategies, leaving a safety margin for possible sudden starts, escapes, or other behaviors by the cubs.

[0106] The determination of the first collision risk level can provide early warning of unstable behavior in animal cubs, such as sudden starts or escape attempts, thus alerting the driver and enhancing driving safety.

[0107] It is evident that the risk of the first collision is relatively low. It focuses on using image recognition information, combined with expected behavior, to provide early warnings for animal cubs that are not in motion in the image, as well as the behavioral uncertainty of stressed animals.

[0108] When the animal target identified by image recognition is a non-moving animal cub or a stressed animal, the animal target is usually far away from the vehicle and the collision risk level is low. Therefore, a warning can be given in a normal way, such as a voice warning at a normal volume and speed, or a static icon of the animal category can be displayed on the screen.

[0109] In one embodiment, the collision risk level is determined based on the collision time, and information about the animal target and / or risk-increasing factors in the expected behavior, including: If the collision time of a moving animal target is greater than the first-class threshold, or if the non-moving animal target is a cub and the collision time is greater than the first-class threshold, then the animal target is determined to have a second collision risk level; wherein, the first-class threshold of a moving animal target is greater than the first-class threshold of a non-moving animal target. The warning information corresponding to the second collision risk level includes at least one of the following: indication of the type of animal target, indication of the presence of an animal target, warning to animal cubs, and warning to moving animals.

[0110] The first threshold is a time boundary value used to assess whether a target has a low collision risk level, and this value is usually relatively large. If a moving animal target has a collision time greater than the first threshold, or if a non-moving animal target is a cub and its collision time is greater than the first threshold, it usually indicates a low collision risk, but attention should be paid to this.

[0111] Generally speaking, when the collision time of an animal target exceeds the first-class threshold, it is in a relatively safe state. In this embodiment, the risk level of moving animal targets or non-moving animal cub targets that have a collision time exceeding the first-class threshold has been increased. This is because the behavioral uncertainty of moving animal targets and animal cubs is taken into account, so as to provide early warning for these animal targets and give the driver sufficient reaction time.

[0112] The first type of threshold has a base threshold, such as 5 seconds. When an animal target is determined to be in motion, the first type of threshold for that animal target is increased by a preset value, such as 1 second. Thus, if the collision time of a moving target is greater than 6 seconds, it is determined to have a second collision risk level. Similarly, if the collision time of a non-moving moving cub is greater than 5 seconds, it is determined to have a second collision risk level. This is because motion implies variability and uncertainty in animal behavior, requiring earlier warnings to allow sufficient reaction time. For non-moving animal targets, whose behavior is relatively stable, a relatively smaller first type of threshold can be used to avoid unnecessary warning interference. The base threshold value for the first type of threshold can be obtained through experimental calibration or learning based on historical data.

[0113] Among them, the warning information corresponding to the second collision risk level can include the category of the animal target, such as clarifying the category of the animal through visual icons or voice broadcasts, such as displaying an icon of "deer" or broadcasting "deer ahead".

[0114] The warning information corresponding to the second collision risk level can include a notification of the presence of an animal target. For example, a general warning symbol or a short voice message such as "Animal ahead" can be used to inform the driver of the presence of an animal target.

[0115] The warning for the animal cub can be to indicate that the animal ahead is a cub, and the warning for the moving animal can be to indicate that the animal ahead is a moving animal.

[0116] Understandably, the risk level of a second collision is higher than that of a first collision; therefore, the warning intensity can be higher in the warning system. For example, the volume and / or tone of the voice warning can be increased.

[0117] In one specific embodiment, an image of the area in front of the vehicle is acquired, and an animal target in the image is identified and tracked. When it is determined that the animal target is a stressed animal and is in motion, preset values ​​are added to both the first and second category thresholds for that animal target, based on the base threshold. When the animal target is a moving target and its collision time exceeds the first category threshold, the animal target is determined to have a second collision risk level; or, when the animal target is a non-moving animal cub and its collision time exceeds the first category threshold, the animal target is determined to have a second collision risk level, and visual and voice warning information is output. The visual warning information may display an icon of the animal on the visual interface. The voice warning information may state "An animal is ahead," and be broadcast in a very calm manner.

[0118] In this embodiment, the sensitivity of risk assessment can be adaptively adjusted according to the movement state of the animal target. At the same time, the risk level is determined by combining the collision time, and when a lower second collision risk level is determined, a matching warning information is provided. This allows for timely and appropriate warnings of potential risks without excessively interfering with the driver.

[0119] In one embodiment, if the animal target is in a non-moving state and the collision time is between a first-class threshold and a second-class threshold, then the animal target is determined to have a third collision risk level; wherein, the first-class threshold for an animal target in a moving state is greater than the first-class threshold for an animal target in a non-moving state; and the second-class threshold for an animal target in a moving state is greater than the second-class threshold for an animal target in a non-moving state. The warning information corresponding to the third collision risk level includes at least one of the following: animal target category and animal target presence indication.

[0120] In this embodiment, considering the higher uncertainty of the behavior of moving targets, a longer warning lead time is required to ensure safety. Therefore, both the first and second type thresholds for moving targets are increased from their base thresholds to enable early warnings to be triggered earlier.

[0121] The Type I threshold is typically higher than the Type II threshold. The Type I threshold is a time boundary value used to assess whether a collision risk level is low, and this value is usually large. If the collision time of an animal target, whether moving or stationary, exceeds the Type I threshold, it generally indicates a low collision risk, but should still be noted. The Type II threshold is a boundary value used to assess whether a collision risk level is high, and this value is usually small. When the collision time of an animal target is between the Type I and Type II thresholds, there is a moderate collision risk. The base threshold for the Type II threshold can be obtained through experimental calibration or learned based on historical data; for example, it could be 3 seconds.

[0122] In one embodiment, the collision risk level is determined based on the collision time, and information about the animal target and / or risk-increasing factors in the expected behavior, including: If the animal target is in motion but does not exhibit path intrusion behavior, and the collision time is not greater than the first-class threshold, then the animal target is determined to have a third-class collision risk level. The warning information corresponding to the third collision risk level includes at least one of the following: animal target category and animal target presence indication.

[0123] In this embodiment, considering the higher uncertainty of the behavior of moving targets, a longer warning lead time is required to ensure safety. Therefore, both the first and second type thresholds for moving targets are increased from their base thresholds to enable early warnings to be triggered earlier.

[0124] The Type I threshold is typically higher than the Type II threshold. The Type I threshold is a time boundary value used to assess whether a collision risk level is low, and this value is usually larger. If the collision time of a moving or stationary animal target exceeds the Type I threshold, it generally indicates a low collision risk, but attention should be paid. The Type II threshold is a time boundary value used to assess whether a collision risk level is high, and this value is usually smaller.

[0125] If an animal target is in motion but does not exhibit path intrusion behavior, and the collision time is no greater than the first-class threshold, then the animal target is determined to have a third-class collision risk level. This means that the target is moving, but its current trajectory does not conflict with the vehicle (e.g., running parallel or moving away), and the collision time is no greater than the first-class threshold; however, it is necessary to be wary of the risk of behavioral mutation caused by its motion state.

[0126] Among them, the warning information corresponding to the third collision risk level can include the category of the animal target, such as clearly identifying the animal category through visual icons or voice broadcasts, such as displaying an icon for "deer" or broadcasting "deer ahead".

[0127] The warning information corresponding to the second collision risk level can include a notification of the presence of an animal target. For example, a general warning symbol or a short voice message such as "Animal ahead" can be used to inform the driver of the presence of an animal target.

[0128] Understandably, the risk level of a third collision is higher than that of a second collision; therefore, the warning intensity can be higher in the warning system. For example, the volume and / or tone of the voice warning can be increased, and the animal icon can change color to yellow and flash.

[0129] In one specific embodiment, an image of the area in front of the vehicle is acquired, and an animal target in the image is identified and tracked. When the animal target is determined to be a stressed animal and is in motion, a preset value is added to both the first and second category thresholds for that animal target, based on the base threshold. When the animal target is in a non-moving state and the collision time is between the first and second category thresholds, the animal target is determined to have a third collision risk level. Alternatively, when the animal target is in motion but does not exhibit path intrusion behavior, and the collision time is not greater than the first category threshold, the animal target is determined to have a third collision risk level. Visual and voice warning information is output. The visual warning information may display an icon of the animal on the visual interface, and the icon may be set to yellow to serve as a warning. The voice warning information may state "Animal ahead," and be broadcast in a relatively urgent manner, such as by increasing the volume or tone.

[0130] When an animal target is determined to have a third-level collision risk, the corresponding warning message can be further differentiated based on whether the animal is a social animal. For example, if the animal target is in a non-moving state and the collision time is between the first and second category thresholds, then the animal target is determined to have a third-level collision risk. If the animal target is a non-social animal, the warning message includes conventional visual and audio warnings (such as displaying an icon of the animal on the visual interface and setting the icon to yellow for a warning effect; the audio warning message could be "An animal ahead," broadcast in a relatively urgent manner, such as increasing the volume or tone). If the animal target is a social animal, in addition to conventional visual and audio warnings, the warning message also includes prompts about social animals, such as indicating that a group may be following.

[0131] Understandably, the urgency of the third collision risk level is higher than that of the second collision risk level. Therefore, the intensity of the voice and visual warnings for animal targets at the third collision risk level should be higher than that for the second collision risk level. For example, for animal targets at the third collision risk level (animals that are not moving but are social animals), the warning method would be: visually display an animal icon that flashes and turns yellow, along with a voice prompt: "Please be aware of the animal ahead; it may move at any time." For animal targets at the second collision risk level (such as those that are not moving and whose collision time exceeds the first-class threshold), the warning method would be: visually display an animal icon, along with a calm voice prompt: "An animal ahead."

[0132] In this embodiment, risk is assessed based on the animal's path intrusion behavior, social attributes, and collision time, rather than a one-size-fits-all approach based solely on collision time, thus improving the precision of risk assessment. Matching different warning content to risks of different natures ensures that the amount of information received by the driver precisely corresponds to the nature of the risk. This differentiated warning information helps build driver trust and reduces warning fatigue caused by ineffective alerts.

[0133] In one embodiment, the collision risk level is determined based on the collision time, and information about the animal target and / or risk-increasing factors in the expected behavior, including: If the animal target is in motion, exhibits path intrusion behavior, and the collision time is between the first and second type thresholds, then the animal is determined to have a fourth collision risk level. The warning information corresponding to the fourth collision risk level includes at least one of the following: animal target category, warning of path intrusion behavior, warning of path intrusion trajectory, and warning strategy for path intrusion behavior. Among them, the first-class threshold for animal targets in motion is greater than the first-class threshold for animal targets in non-motion; the second-class threshold for animal targets in motion is greater than the second-class threshold for animal targets in non-motion.

[0134] For moving animal targets whose collision time falls between the first and second category thresholds, if the moving animal target exhibits path intrusion behavior, its collision risk level is higher than that of a non-moving animal target whose collision time falls between the first and second category thresholds. This is because the animal target's behavior is identified as path intrusion behavior, meaning that the animal target is crossing the road or approaching a vehicle, thus increasing its risk level.

[0135] The warning information corresponding to the fourth collision risk level includes animal target categories. Specifically, the categories of animal targets are quickly identified through specific icons (such as deer or cow icons) or voice prompts.

[0136] The warning information corresponding to the fourth collision risk level includes alerts for path intrusion behavior. In this way, warnings for path intrusion behavior can be clearly displayed, such as a visual display "An animal is crossing the road ahead!", or a voice announcement "An animal is approaching the lane ahead!", or a voice announcement "An animal is approaching the lane, please slow down". The warning information corresponding to the fourth collision risk level includes a path intrusion trajectory indication. In this way, path conflicts can be indicated in a visual way, such as by overlaying arrows on the display showing the predicted direction of an animal's movement, allowing the driver to anticipate the point of conflict.

[0137] The warning information corresponding to the fourth collision risk level includes alert strategies for path intrusion behavior. For example, different alert strategies can be used for different path intrusion behaviors. For instance, for lateral intrusion, the strategy might be: "Please slow down or stop immediately to give way!" While for a frontal collision, the strategy might be: "Prepare for emergency braking!" It should be understood that the urgency level of the fourth collision risk level is higher than that of the third collision risk level. Therefore, the intensity of voice and visual warnings should be higher than that of the third collision risk level. For example, visual warnings could include animal icons and predicted path arrows, and voice warnings could include phrases like "An animal is approaching the lane, please slow down." In one specific embodiment, an image of the area in front of the vehicle is acquired, and an animal target in the image is identified and tracked. When it is determined that the animal target is a stress-prone animal and is in motion, preset values ​​are added to both the first and second category thresholds for that animal target, based on the base threshold. If the animal target is in motion, exhibits path intrusion behavior, and the collision time is between the first and second category thresholds, then the animal is determined to have a fourth collision risk level. Visual and voice warning information is output. The visual warning information may display a dynamic animal icon (e.g., flashing) and a predicted path arrow on the visual interface to serve as a warning. The voice warning information may state "An animal is approaching the lane, please slow down," and be broadcast in a relatively urgent manner, such as by increasing the volume or tone.

[0138] In this embodiment, animal targets that are in motion and exhibit path intrusion behavior are set to a higher risk level, which can provide an early warning when the animal just shows the intention to intrude into the path and begins to act (even if the TTC value may still be high at this time), so that the driver can have more time to make decisions and respond before the collision risk is fully manifested.

[0139] In one embodiment, the collision risk level is determined based on the collision time, and information about the animal target and / or risk-increasing factors in the expected behavior, including: If the collision time of the animal target is less than the second-class threshold, the animal target is determined to have a fifth collision risk level; the second-class threshold of the animal target in motion is greater than the second-class threshold of the animal target in non-motion. The warning information corresponding to the fifth collision risk level includes collision risk alerts and the type of animal.

[0140] The second threshold is a time boundary value used to assess whether a high-risk collision level exists; this value is typically small. This threshold indicates that the danger has entered an emergency phase, and a collision is imminent and unavoidable unless maximum-intensity intervention is implemented immediately. When the collision time of the animal target is less than the second threshold, the animal target is determined to have a fifth-level collision risk.

[0141] The warning information corresponding to the fifth collision risk level includes a collision risk alert and the animal category. Specifically, the collision risk alert uses the highest priority alarm method, such as displaying a red icon of the animal on the display screen, while simultaneously triggering a continuous, high-frequency, and loudest beeping or warning sound.

[0142] In one specific embodiment, an image of the area in front of the vehicle is acquired, and an animal target in the image is identified and tracked. When it is determined that the animal target is a stressed animal and is in motion, preset values ​​are added to both the first and second category thresholds for that animal target, based on the base threshold. If the collision time of the animal target is less than the second category threshold, the animal target is determined to have a fifth collision risk level. Visual and audible warning information is then output. For example, a red icon of the animal is displayed on the screen, and simultaneously, a continuous, high-frequency, and loudest beeping or warning sound is triggered.

[0143] This embodiment triggers the highest level of emergency response by determining that the collision time of the animal target is less than the second type of threshold, so as to slow down the collision speed or avoid the collision to the greatest extent.

[0144] In one embodiment, predicting the expected behavior of an animal target based on information about the animal target and environmental information includes: querying an animal behavior knowledge base to obtain the expected behavior of the animal target based on information about the animal target and environmental information; the information about the animal target includes at least one of the following: the category of the animal target, the animal's behavior, age, social context, behavior, and posture.

[0145] Animal behavior knowledge base is a pre-built structured database or rule base. The data can come from ecological research and field observation, and it can map animal categories and environmental information to typical expected behaviors.

[0146] In one embodiment, the animal behavior knowledge base can store: the inherent habits of different animal categories (such as deer, wild boar, and cattle), such as activity rhythms (diurnal habits), diet, habitat preferences, and seasonal migration patterns; defined behavioral rules that animals may adopt under specific environments and conditions; for social animals, defined social roles (leader, follower, vigilant) and group interaction rules; and typical stress behaviors (such as fleeing in fright, stiffening, and charging) of easily startled animals when faced with specific stimuli (such as bright light, whistling, or rapid approach). For example, one record might be: Deer, social animals, commonly found on forest roads at dusk; after the leader crosses, subsequent individuals are highly likely to follow. Another example is: Wild boar, easily startled by bright light, may charge in a straight line when startled.

[0147] The prediction process involves a multi-condition query. The system uses real-time perceived information as query conditions and inputs them into an animal behavior knowledge base. The knowledge base outputs one or more expected behaviors and their probabilities through rule-based reasoning or probability matching. For example, if the perceived information includes: category = "elk", time environment = "dusk", geographical environment = "forest road", behavioral posture = stationary posture, and social context = solitary, the output expected behavior can be in a structured form: crossing the road, with a confidence level of 65%.

[0148] In this embodiment, the expected behavior is predicted by introducing and querying an animal behavior knowledge base. This makes risk warning no longer solely based on the collision time, but also by predicting the animal target's behavioral intentions, i.e., assessing the impact of the moving target's next actions on the collision risk, thereby improving the accuracy of collision risk assessment.

[0149] In one embodiment, the method further includes the step of constructing a behavioral knowledge base, such as... Figure 3 As shown, the steps for constructing a behavioral knowledge base include: Step 302: Preprocess the animal behavior data; each piece of animal behavior data includes animal information, environmental information, and animal behavior.

[0150] Specifically, a wide range of data were collected from publicly available animal ecology research literature, wildlife observation databases, nature reserve monitoring reports, historical data from vehicle recorders, and field observation data from specialized organizations.

[0151] The data includes not only the animal's basic attributes (such as species, age, and sex), but more importantly, environmental information (such as geographical location, habitat type, season, weather, and time of day) and behavioral sequence information (such as stillness, walking, running, foraging, alertness, crossing roads, and group interaction).

[0152] The collected multi-source, heterogeneous data is cleaned, denoised, and standardized in a unified format (such as JSON or XML) to lay the foundation for building a knowledge base.

[0153] Step 304: Based on the preprocessed animal behavior data, structured behavior prediction rules are generated by injecting expert rules based on ecological prior knowledge and mining potential patterns in the data through machine learning algorithms.

[0154] Based on the prior knowledge of ecologists, some strong association rules were manually constructed. For example, the following were directly entered: "If the animal type is 'deer', the time is 'dusk or dawn', and the environment is 'a road near the forest', then the probability of its intention to 'cross the road' is significantly increased"; "If the animal type is 'coyote', and its behavior posture is detected as 'crouching', then the probability of its intention to 'prepare to charge' increases."

[0155] Using preprocessed data, machine learning models (such as decision trees, random forests, Bayesian networks, or association rule mining algorithms) can be trained to discover complex patterns hidden within the data that experts may not have summarized. For example, the model might discover that "when the population density is greater than 5 individuals per 100 square meters and the leader's movement speed exceeds 2 meters per second, the probability that the following individuals will initiate following within 1.5 seconds exceeds 85%".

[0156] This approach integrates expert rules with rules mined through machine learning. When rules conflict, they are resolved using pre-defined strategies (such as confidence level, data support, and expert weights), resulting in a unified, consistent, and consistent set of structured behavior prediction rules. An example of such structured behavior prediction rules is as follows: Figure 4 As shown.

[0157] Step 306: The behavior prediction rules are stored in a structured manner to construct the animal behavior knowledge base.

[0158] The fused behavior prediction rules are stored in a structured manner in a relational database or graph database, or typically in a knowledge graph structure of "entity-relationship-attribute". Each rule describes a behavioral pattern that an animal may exhibit under specific conditions. During real-time prediction, matching rules can be efficiently retrieved based on the input animal target information and environmental information.

[0159] It should be understood that the establishment of an animal behavior knowledge base is a continuous learning and optimization process. By combining ecological knowledge with machine learning, the limitations of relying solely on expert experience (limited coverage, slow updates) or purely data-driven approaches (requiring massive amounts of data, poor interpretability) are overcome. Expert rules ensure the accuracy of core knowledge, while machine learning efficiently mines new knowledge from massive amounts of data. The two complement each other, accelerating the construction of a high-quality knowledge base. Structured storage facilitates the addition, deletion, modification, and retrieval of knowledge. As new data is continuously collected, machine learning processes can be initiated periodically to discover new rules or adjust the confidence levels of existing rules; simultaneously, experts can inject new knowledge in real time based on new research findings. This makes the animal behavior knowledge base a continuously learning, dynamically evolving intelligent agent, capable of adapting to different geographical and seasonal changes, as well as potential long-term shifts in animal behavior and habits.

[0160] One embodiment of the risk warning method based on animal collision is applied to, for example Figure 5 The vehicle equipment shown includes: a visible light camera 51, an infrared camera 52, and an in-vehicle terminal 53, wherein the in-vehicle terminal 53 includes a controller 531, a display screen 532, and a speaker 533.

[0161] The controller 531 is equipped with an animal behavior knowledge base and a risk warning device based on animal collisions. The device includes a dual-light fusion sensing module, an animal identification and tracking module, and a risk assessment and decision-making module.

[0162] Specifically, such as Figure 6 The diagram showing the flow of early warning data is as follows: The dual-light fusion sensing module is used to acquire visible light images captured by a visible light camera and infrared images captured by an infrared camera, and to fuse the visible light images and infrared images to obtain a fused image.

[0163] The animal recognition and tracking module utilizes target detection algorithms to perform target detection and classification on fused images, identifying the animal target's category (e.g., wild boar, elk, sheep, coyote, cattle), age (adult or juvenile), initial position, posture (e.g., standing, running, crouching), and environment. Based on the animal recognition results from the image sequence, a target tracking algorithm is used to track the animal target and record the tracking data, obtaining the animal's movement trajectory and posture changes.

[0164] The risk assessment and decision-making module is used to obtain environmental information. Based on the information of the animal target and the environmental information, it queries the animal behavior knowledge base to obtain the expected behavior of the animal target. The information of the animal target includes at least one of the following: the category of the animal target, the animal's behavior, age, social context, behavior, and posture.

[0165] The risk assessment and decision-making module is also used to calculate the collision time of the animal target based on the vehicle's speed and position, and to determine the collision risk level based on the collision time, the information of the animal target, and / or risk-increasing factors in the expected behavior. Based on the risk-increasing factors and the information of the animal target, it determines visual warning information and / or voice warning information. The visual warning information and voice information respectively include at least one of the following: prompts for the animal target, warnings for risk-increasing factors, and alert strategies for risk-increasing factors. The visual warning information is output using visual parameters matched with the collision risk level, and / or, the voice warning information is output using sound parameters matched with the collision risk level. The intensity of the visual parameters and sound parameters increases with the increase of the collision risk level.

[0166] The display screen 532 and the speaker 533 execute the warning information. Specifically, they display corresponding icons, text and graphics on the vehicle's central control screen or head-up display, and play corresponding voice prompts through the vehicle's speakers, thereby completing a complete warning process.

[0167] Among them, risk-increasing factors related to expected behavior include: path conflict behavior of the animal target; and / or, risk-increasing factors related to information about the animal target include at least one of: the animal target being a stressful animal, the animal target being in motion, the animal target being a social animal, and the animal target being an offspring.

[0168] In one embodiment, the initial first and second type thresholds for the animal target are both base thresholds. The first type threshold is typically greater than the second type threshold. The first type threshold is a time boundary value used to assess whether the animal has a low collision risk level, while the second type threshold is a boundary value used to assess whether it has a high collision risk level. In practical applications, an upper limit for the collision time can also be set, such as 15 seconds. If the animal target does not have factors that increase risk and the collision time exceeds the upper limit, a collision warning will not be triggered.

[0169] like Figure 7 As shown, the specific judgment and early warning processes of the risk assessment and decision-making module include the following steps: Step 1: Upon detecting an animal target in the foreground image, acquire information about the animal target, its expected behavior, and the collision time. Information about the animal target includes at least one of the following: animal category, animal behavior, age, social context, and posture.

[0170] Step 2: Determine if the animal is a stress-prone animal. Specifically, determine whether it is a pre-defined stress-prone animal based on the animal target's category, or identify the animal target's stress behaviors based on its image. If yes, proceed to Step 3; otherwise, proceed to Step 4.

[0171] Step 3: Determine if the animal target has a first collision risk level. Visual warning: Display the category icon of the animal target. Voice warning: Announce "Remind to switch to high beams; avoid horn-style alarms". Set the risk adjustment value TTC_OFFSET = 2 for the animal target, which is to add 2 to the first and second category thresholds of the animal target based on their respective base thresholds.

[0172] Step 4: Set the risk adjustment value for the animal target to TTC_OFFSET = 0.

[0173] Step 5: Determine if the animal target is in motion. If not, proceed to step 6; if yes, proceed to step 15. Whether the animal target is in motion can be determined based on the tracking results of its movement trajectory.

[0174] Step 6: Determine if the target animal is a cub. If yes, proceed to step 7; otherwise, proceed to step 8.

[0175] Step 7: Determine if the animal target has a first collision risk level. Visual warning: Display the category icon of the animal target. Voice warning: Announce "Caution: cubs, unpredictable behavior". Then proceed to step 8.

[0176] Step 8: Determine if the collision time is greater than the first-class threshold of the animal target. If the base threshold of the first-class threshold is 5, then determine if TTC > 5 + TTC_OFFSET. If yes, proceed to step 9; otherwise, proceed to step 10.

[0177] Step 9: Determine if the animal target has a second collision risk level. Visual warning: Display the category icon of the animal target. Voice warning: Announce "An animal is ahead" in a very calm tone.

[0178] Step 10: Determine whether the collision time is between the first type threshold and the second type threshold. For example, if the base threshold of the second type threshold is 3, then determine whether TTC>3+TTC_OFFSET&&TTC<5+TTC_OFFSET. If yes, then proceed to step 11. If no, that is, TTC<3+TTC_OFFSET, then proceed to step 13.

[0179] Step 11: Determine if the target animal is a social animal. If yes, output a warning for social animals, such as a visual warning or early warning: beware of being followed, and then proceed to step 12; if no, proceed directly to step 12.

[0180] Step 12: Determine that the animal target has a third-level collision risk. Visual warning: Display the animal target's icon and flash it; the icon color changes to yellow. Voice warning: "Please be aware of the animal ahead; it may move at any time." Step 13: Determine if the target animal is a social animal. If yes, output a warning to social animals, such as a visual or verbal warning: Beware of being followed, and then proceed to step 14; otherwise, proceed directly to step 14.

[0181] Step 14: Determine that the animal target has a collision risk level of 5. Visual warning: Display the animal's icon, change its color to red, and flash it across the entire screen. Voice warning: Emits a short, loud beeping sound.

[0182] Step 15: Set the risk adjustment value TTC_OFFSET = 1 for the animal target, which means adding 1 to the first and second category thresholds of the animal target based on their respective base thresholds.

[0183] Step 16: Determine if the collision time is greater than the first type threshold, for example, whether TTC > 6 + TTC_OFFSET. If yes, proceed to step 17; otherwise, proceed to step 18.

[0184] Step 17: Determine if the animal target has a second collision risk level. Visual warning: Display the category icon of the animal target; Voice warning: Announce "Animal ahead" in a very calm tone.

[0185] Step 18: Determine whether the animal target exhibits path intrusion behavior, for example, based on the movement trajectory of the tracked animal target, determine whether there is a high probability of it crossing or approaching. If not, proceed to step 19; if yes, proceed to step 20.

[0186] Step 19: Determine if there is a third collision risk level. Visual warning: Display an icon of the animal target and flash it, the color of which changes to yellow. Voice warning: Announce "Please be aware of animals ahead".

[0187] Step 20: Determine if the target animal is a social animal. Is it social? If so, output a warning for social animals, such as a visual warning or a voice warning: Beware of being followed, and then proceed to step 21; otherwise, proceed directly to step 21.

[0188] Step 21: Determine whether the collision time is between the first type threshold and the second type threshold. For example, determine whether TTC > 4 + TTC_OFFSET && TTC < 6 + TTC_OFFSET. If not, proceed to step 22; if yes, proceed to step 23.

[0189] Step 22: Determine that the animal target has a collision risk level of 5. Visual warning: Display the category icon of the animal target, the icon turns red, and flashes across the entire screen. Voice warning: Issue a short, loud beeping sound.

[0190] Step 23: Determine if the animal target has a fourth collision risk level. Visual warning: Display a dynamic animal icon and a predicted path arrow, such as... Figure 8 As shown. Voice warning: "An animal is approaching the lane, please slow down." The risk warning method based on animal collisions proposed in this application has the following technical effects: 1. Animal Fine Classification and Posture Estimation Based on Dual-Light Fusion: Different early warning strategies are implemented for different animal categories and animal size recognition (adult or cub). It not only utilizes infrared and visible light fusion to detect animals, but also leverages the rich features resulting from the fusion to perform species-specific classification (e.g., wild boar, elk, sheep, coyote, cattle) and real-time posture estimation (e.g., standing, running, head down) using an AI model. Furthermore, it uses movement trajectories to determine whether the animal is stationary or wandering, providing crucial information input for subsequent behavior prediction.

[0191] 2. A pre-set animal behavior knowledge base was established, which binds environmental characteristics, time characteristics, and the identified animal type and size to specific behavioral characteristic models. This elevates the prediction of simple physical movement trajectories to the prediction of intentions combined with animal behavior, making risk assessment more realistic and forward-looking.

[0192] 3. On the display screen, differentiated icons (deer-shaped, cow-shaped), behavioral status text prompts ("May cross!, please slow down"), and graphical displays of predicted paths (virtual arrows), combined with speaker voice prompts, directly and intuitively convey the system's "cognitive results" to the driver, improving human-machine collaboration efficiency. This application enables intelligent animal early warning systems that "recognize its shape, know its nature, predict its actions, and provide detailed information."

[0193] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0194] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0195] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A risk warning method based on animal collision, characterized in that, The method includes: Based on the animal recognition results of the image sequence, information about the animal target in the image is obtained; Acquire environmental information and, based on the information of the animal target and the environmental information, predict the expected behavior of the animal target; Calculate the collision time of the animal target, and determine the collision risk level based on the collision time, information about the animal target, and / or risk-increasing factors in the expected behavior; The warning information is output using a warning method that matches the collision risk level.

2. The risk pre-warning method based on animal collision according to claim 1, characterized in that, The warning information includes visual warning information and / or voice warning information; the method further includes: Based on the information of the risk-increasing factors and the animal target, the visual warning information and / or the voice warning information are determined; the visual warning information and the voice information respectively include at least one of: prompting the animal target, warning about the risk-increasing factors, and vigilance strategies for the risk-increasing factors.

3. The animal collision-based risk pre-warning method according to claim 2, wherein, The method of outputting warning information using a warning method that matches the collision risk level includes: The visual warning information is output using visual parameters that match the collision risk level, and / or the voice warning information is output using sound parameters that match the collision risk level; The intensity of the visual parameters and the sound parameters increases as the collision risk level increases.

4. The animal collision-based risk pre-warning method according to claim 1, wherein, The risk-increasing factors associated with the expected behavior include: path conflict behavior of the animal target; and / or, the risk-increasing factors associated with information about the animal target include at least one of: the animal target being a stress-prone animal, the animal target being in motion, the animal target being a social animal, and the animal target being an offspring.

5. The animal collision-based risk pre-warning method according to claim 1, wherein, The method further includes: If the animal target is detected to be a social animal, a warning for social animals will be issued.

6. The risk warning method based on animal collision according to claim 4 or 5, characterized in that, The determination of the collision risk level based on the collision time, information about the animal target, and / or risk-increasing factors in the expected behavior includes: If the animal target is a stress-prone animal, then the animal target is determined to have a first collision risk level; correspondingly, the warning information corresponding to the first collision risk level includes at least one of: a prompt indicating the category of the animal target, a warning that the animal target is a stress-prone animal, and a warning strategy for the stress-prone animal; the warning strategy for the stress-prone animal is used to prompt the adoption of behavioral patterns that reduce the animal's stress response; And / or, If the animal target is a non-moving animal cub, then the animal target is determined to have a first collision risk level; accordingly, the warning information corresponding to the first collision risk level includes at least one of the following: a prompt indicating the type of animal target, a warning that the animal target is an animal cub, and a warning strategy for the animal cub.

7. The risk pre-warning method based on animal collision according to claim 4 or 5, characterized in that, The determination of the collision risk level based on the collision time, information about the animal target, and / or risk-increasing factors in the expected behavior includes: If the collision time of the animal target in motion is greater than the first type threshold, or if the animal target in non-motion is an animal cub and the collision time is greater than the first type threshold, then the animal target is determined to have a second collision risk level; wherein the first type threshold of the animal target in motion is greater than the first type threshold of the animal target in non-motion. The warning information corresponding to the second collision risk level includes at least one of the following: a prompt indicating the type of the animal target, a prompt indicating the presence of the animal target, a warning to the animal cubs, and a warning to moving animals.

8. The risk pre-warning method based on animal collision according to claim 4 or 5, characterized in that, The determination of the collision risk level based on the collision time, information about the animal target, and / or risk-increasing factors in the expected behavior includes: If the animal target is in a non-moving state and the collision time is between a first-class threshold and a second-class threshold, then the animal target is determined to have a third collision risk level; wherein, the first-class threshold of the animal target in a moving state is greater than the first-class threshold of the animal target in a non-moving state; and the second-class threshold of the animal target in a moving state is greater than the second-class threshold of the animal target in a non-moving state. The warning information corresponding to the third collision risk level includes at least one of the animal target category and the presence indication of the animal target.

9. The risk pre-warning method based on animal collision according to claim 4 or 5, characterized in that, Determining the collision risk level based on the collision time, information about the animal target, and / or risk-increasing factors in the expected behavior includes at least one of the following: The first type: If the animal target is in motion but does not exhibit path intrusion behavior, and the collision time is not greater than the first-class threshold, then the animal target is determined to have a third-class collision risk level. The warning information corresponding to the third collision risk level includes at least one of the animal target category and the presence indication of the animal target; The second type: If the animal target is in motion and exhibits path intrusion behavior, and the collision time is between the first and second thresholds, then the animal is determined to have a fourth collision risk level. The warning information corresponding to the fourth collision risk level includes at least one of the following: the animal target category, warnings for path intrusion behavior, warnings for path intrusion trajectories, and warning strategies for path intrusion behavior. Wherein, the first type threshold of the animal target in motion is greater than the first type threshold of the animal target in non-motion; the second type threshold of the animal target in motion is greater than the second type threshold of the animal target in non-motion.

10. The risk pre-warning method based on animal collision according to claim 4 or 5, characterized in that, The determination of the collision risk level based on the collision time, information about the animal target, and / or risk-increasing factors in the expected behavior includes: If the collision time of the animal target is less than the second-class threshold, then the animal target is determined to have a fifth collision risk level; the second-class threshold of the animal target in motion is greater than the second-class threshold of the animal target in non-motion. The warning information corresponding to the fifth collision risk level includes: collision risk warning and the category of the animal.

11. The risk warning method based on animal collision according to claim 6, further comprising: If the animal target is a stress-prone animal, then a preset value is added to the base threshold of the first and second type thresholds for the animal target; the first and second type thresholds are the time boundary values ​​for determining the collision risk level.

12. The risk warning method based on animal collision according to any one of claims 1 to 5, characterized in that, The step of predicting the expected behavior of the animal target based on the information of the animal target and the environmental information includes: Based on the information of the animal target and the environmental information, the expected behavior of the animal target is obtained by querying the animal behavior knowledge base. The information about the animal target includes at least one of the following: the category of the animal target, the animal's behavior, age, social context, behavior, and posture.

13. The animal collision-based risk pre-warning method according to claim 12, wherein, The method further includes: Collect and preprocess animal behavior data; each piece of animal behavior data includes animal information, environmental information, and animal behavior. Based on the preprocessed animal behavior data, behavior prediction rules are generated by injecting expert rules based on ecological prior knowledge and by mining potential patterns in the data through machine learning algorithms. The behavior prediction rules are structured and stored to construct the animal behavior knowledge base.

14. A vehicle apparatus characterized by comprising: The device includes a sensor module, a processor, a display, a speaker, and a memory connected to the processor. The memory stores a computer program that can be executed by the processor. When the computer program is executed by the processor, it implements the steps of the risk warning method based on animal collision as described in any one of claims 1 to 13.

15. An electronic device, comprising: The method includes a sensor module, a processor, and a memory connected to the processor. The memory stores a computer program that can be executed by the processor. When the computer program is executed by the processor, it implements the steps of the risk warning method based on animal collision as described in any one of claims 1 to 13.

Images