An unmanned vehicle passability prediction and path selection system based on visual detection

Abstract

The application relates to the technical field of path navigation, in particular to an unmanned vehicle passability prediction and path selection system based on visual detection. The application constructs a comprehensive evaluation model by integrating environmental perception data, a vehicle dynamics model and cargo state information; when obstacles are encountered, not only the passing posture of the vehicle can be simulated, but also the instantaneous impact load that will be borne by each part of the chassis can be predicted, and the cumulative vibration damage risk of the cargo due to the impact can be evaluated, so that the safety of the cargo is improved.

Description

Technical Field

[0001] This application relates to the field of path navigation technology, and in particular to a vision-based unmanned vehicle passability prediction and path selection system. Background Technology

[0002] With the rapid development of autonomous driving technology and intelligent logistics systems, autonomous vehicle decision-making in complex road conditions has become a key challenge. Especially on unstructured roads or temporary construction zones, autonomous vehicles need to identify and assess in real time the potential impact of obstacles (such as potholes, gravel, and protrusions) on the vehicle itself and its cargo in order to make safe and economical driving decisions. Current technologies often rely on preset physical parameter thresholds or simple sensor feedback to determine vehicle passability, lacking a comprehensive quantitative analysis of the dynamic interactions between the vehicle, cargo, and obstacles. This results in decisions that are often conservative or mismatched with actual risks.

[0003] Currently, common solutions mainly focus on the following two levels: First, obstacle recognition and classification based on environmental perception. This involves acquiring road information through sensors such as vision and lidar, and using image processing or deep learning algorithms to identify the type, size, and location of obstacles. However, these methods often remain at the stage of extracting the static geometric features of obstacles, failing to further combine vehicle status and cargo attributes to assess the dynamic deformation and vibration effects that obstacles may cause when vehicles run over or pass over them. Second, passability prediction based on vehicle dynamics models. Some studies simulate the vibration response of vehicles crossing obstacles by establishing mechanical models of the vehicle's suspension system and chassis structure. However, these methods typically ignore the impact resistance characteristics of the cargo itself and its cumulative damage effects during transportation, and also fail to incorporate the contact mechanical characteristics (such as hardness and sharpness) between the vehicle chassis and the obstacle into the analysis, leading to discrepancies between the predicted results and the actual cargo damage risk.

[0004] In summary, existing technologies have the following shortcomings: they do not fully quantify the deformation characteristics of obstacles under vehicle loads and their impact on vehicle vibration, nor do they establish a correlation model between historical vibration accumulation and current obstacle passage risk. Therefore, there is an urgent need for an intelligent decision-making method that can integrate visual recognition, deformation mechanics analysis, vibration transmission prediction, and cargo damage accumulation assessment, so that unmanned vehicles can make passage judgments and path selections that ensure both safety and economy in complex road conditions. Summary of the Invention

[0005] To overcome the shortcomings and deficiencies of existing technologies, this application constructs a comprehensive evaluation model by integrating environmental perception data, vehicle dynamics models, and cargo status information. When encountering obstacles, it can not only simulate the vehicle's passing posture, but also predict the instantaneous impact loads that the chassis components will bear, and assess the risk of cumulative vibration damage to the cargo, thereby improving cargo safety.

[0006] To achieve the above objectives, this application adopts the following technical solution: Firstly, this application provides a method for predicting the passability and selecting a path for an autonomous vehicle based on visual detection, including the following specific steps: Step 1: Use visual recognition components to acquire and identify road obstacles, vehicle movement, and cargo information. Step 2: Perform vehicle travel deformation analysis based on vehicle driving conditions, cargo conditions, and identified obstacles; Step 3: Analyze the obstacle deformation analysis results and vehicle driving conditions to predict vehicle overpass vibrations; Step 4: Predict the cargo damage caused by vehicle vibration based on the vehicle vibration prediction results, the vehicle's historical driving data, and the cargo condition. Step 5: Analyze the damage to the vehicle and cargo to determine whether it is necessary to cross the corresponding road obstacle.

[0007] In one implementation of this application, acquiring and identifying road obstacle conditions includes the following specific steps: The first step involves acquiring images and sizes of obstacles on the road ahead using a visual recognition component. These images and sizes are then compared to obstacle types in a database to obtain image and size similarity scores. Image and size similarity scores can be obtained using any existing image similarity calculation algorithm. For example, image similarity is calculated by comparing a sequence of pixel values ​​at each point with a sequence of pixel values ​​for each obstacle type using a cosine similarity calculation formula. Similarly, size similarity is calculated by comparing a sequence of dimensions at each location with a sequence of dimensions for each location of the obstacle type using a cosine similarity calculation formula. The second step is to perform a weighted sum of the image similarity and size similarity with each obstacle type in the obstacle database to obtain the obstacle similarity with each obstacle type in the obstacle database. The third step is to obtain the obstacle type corresponding to the highest obstacle similarity and set it as the recognition result of the corresponding obstacle. The elastic deformation and hardness characteristics of the corresponding obstacle are obtained; at the same time, the thickness and width characteristics are obtained; and the obtained obstacle features are stored in the corresponding storage module. The fourth step is to obtain information on vehicle weight, vehicle vibration while driving on the road, and cargo impact resistance. Vehicle weight includes the total weight of the vehicle and cargo, and cargo impact resistance includes the safe vibration amplitude of the cargo.

[0008] The overall benefits of this step are as follows: This step uses visual recognition components to collect and analyze road obstacles, vehicle status, and cargo information in real time, providing a comprehensive data foundation for subsequent decision-making. Its core advantage lies in achieving the synchronous acquisition and fusion of multi-source information. It can accurately identify the physical characteristics of obstacles (such as size and type) and combine vehicle driving parameters (such as weight and vibration) and cargo attributes (such as impact resistance) to form a holistic perception of the driving environment, thereby avoiding the limitations of a single data source.

[0009] In one implementation of this application, the driving process via deformation analysis includes the following specific steps: The first step is to obtain information about the vehicle chassis, the height of the obstacle, and the sharpness and hardness of the side facing the chassis. The sharpness is represented by the cosine of the angle. Obstacle hardness is obtained by dividing the obstacle hardness by the chassis hardness. Height is obtained by dividing the obstacle height by the chassis height. Sharpness is obtained by ensuring the obstacle's sharpness is within a safe level. Damage to the chassis caused by the obstacle is obtained by weighted summation of hardness, height, and sharpness anomalies. The second step is to obtain the vehicle weight, the elastic deformation and hardness characteristics of the corresponding obstacle; by dividing the vehicle weight into gravity and the elastic deformation coefficient of the obstacle, the deformation of the obstacle is obtained; by subtracting the influence coefficient of the deformation on the thickness from the overall thickness and multiplying it by the value of the deformation, the vehicle loading thickness is obtained; by dividing the vehicle loading thickness by the safe height, the abnormal vehicle vibration height is obtained; and by dividing the hardness of the corresponding obstacle by the safe hardness, the abnormal hardness is obtained. The third step is to compare the abnormal damage to the chassis caused by the obstacle with the set abnormal damage threshold. If the abnormal damage to the chassis is greater than or equal to the set abnormal damage threshold, it is determined to detour directly; if the abnormal damage to the chassis is less than the set abnormal damage threshold, proceed to step three.

[0010] The overall benefits of this step are as follows: by comprehensively analyzing the characteristics of the vehicle chassis, the attributes of the obstacle (such as height, hardness, and sharpness), and the vehicle load, the potential damage risk of the obstacle to the vehicle chassis is quantitatively assessed. The qualitative description of the obstacle is transformed into a calculable damage anomaly value, and rapid risk classification is achieved by comparing it with a preset threshold. This not only avoids the error of subjective judgment, but also identifies high-risk obstacles in advance and triggers detour decisions in a timely manner, thereby ensuring the structural safety of the vehicle while reducing unnecessary subsequent calculation burden. In one implementation of this application, the vehicle's vibration prediction includes the following specific details: The first step involves obtaining the abnormal vehicle vibration height and hardness for the corresponding obstacle, and then performing a weighted summation to obtain the obstacle obstruction anomaly. Next, the obstruction anomaly of the corresponding vehicle when crossing historical obstacles during this trip, along with the vibration amplitude and frequency at the time of crossing, are obtained. The vibration anomaly is calculated using the average vibration amplitude and average vibration frequency of historical vehicle crossings. The vibration anomaly is calculated as follows: the amplitude anomaly is obtained by dividing the average vibration amplitude by the safe amplitude; the frequency anomaly is obtained by dividing the average vibration frequency by the safe frequency; and the vibration anomaly of the corresponding obstacle is obtained by multiplying the amplitude anomaly by the frequency anomaly. This process combines the vibration anomaly and obstacle obstruction anomaly obtained during the current trip. The second step is to obtain the obstacle influence coefficient by dividing the vibration anomaly of the historical obstacle by the obstacle obstruction anomaly. The average value of the obstacle influence coefficient is then multiplied by the obstacle obstruction anomaly to be analyzed in this case to obtain the vibration anomaly caused by the obstacle to be analyzed in this case.

[0011] The overall benefit of this step is that, based on historical obstacle-crossing data (vibration amplitude and frequency) and the current obstacle's anomaly value, it estimates the vibration anomalies that the vehicle may generate when crossing the current obstacle. Its core value lies in dynamically correcting the prediction results using historical experience data. By quantifying vibration predictions as anomalies, the system can more accurately assess the impact of obstacle-crossing behavior on vehicle stability, providing crucial input for subsequent cargo damage analysis.

[0012] In one implementation of this application, the damage to goods caused by vehicle vibration includes the following specific details: The cumulative damage impact coefficient of the cargo is obtained by multiplying the sum of the vibration anomalies of the vehicle crossing historical obstacles during this trip by the historical vibration impact coefficient. The normal coefficient of the cargo is obtained by subtracting the cumulative damage impact coefficient of the cargo from the value of 1 and dividing it by the vibration anomalies caused by the obstacle to be analyzed in this trip.

[0013] The benefits of this step are as follows: This step combines the current vibration prediction value with the historical cumulative damage impact to calculate the normal coefficient of the cargo, and quantitatively assess the comprehensive impact of obstacle crossing on cargo safety. Its advantage lies in the introduction of the concept of cumulative damage, which not only considers the instantaneous impact of a single obstacle crossing, but also takes into account the superimposed effect of multiple vibrations on the cargo, thus more realistically reflecting the actual state of the cargo in long-distance transportation. This dynamic cumulative assessment method helps to prevent hidden damage caused by continuous small-amplitude vibrations and improves the comprehensiveness of cargo safety early warning.

[0014] In one implementation of this application, the analysis of whether it is necessary to cross the corresponding road obstacle includes the following specific steps: The calculated normal coefficient of the cargo is compared with the set normal coefficient threshold. If the normal coefficient of the cargo is greater than or equal to the set normal coefficient threshold, it means that the cargo is in a normal state when crossing, and it is determined that the vehicle can cross the corresponding road obstacle. If the normal coefficient of the cargo is less than the set normal coefficient threshold, it means that the cargo will be damaged when crossing, and it is determined that the vehicle needs to go around the corresponding road obstacle.

[0015] Secondly, this application also provides a vision-based autonomous vehicle passability prediction and path selection system, including the following specific modules: The module includes a data acquisition module, an obstacle recognition module, a deformation analysis module, a vibration prediction module, a cargo damage analysis module, and a path guidance module. The data acquisition module is used to acquire information about the vehicle's movement and the cargo's status. The obstacle recognition module is used to acquire and identify road obstacle conditions through a visual recognition component; The deformation analysis module is used to perform vehicle travel deformation analysis based on vehicle driving conditions, cargo conditions, and identified obstacle conditions. The vibration prediction module is used to analyze the vibration prediction of the vehicle crossing by analyzing the obstacle deformation analysis results and the vehicle driving conditions. The cargo damage analysis module predicts cargo damage caused by vehicle vibration based on vehicle vibration prediction results, vehicle driving conditions, and cargo conditions. The path guidance module is used to analyze whether it is necessary to cross the corresponding road obstacle based on the damage to the vehicle and cargo.

[0016] Thirdly, this application provides an electronic device comprising: a processor and a memory, wherein the memory stores a computer program that can be called by the processor, and the processor executes a vision-based unmanned vehicle passability prediction and path selection method by calling the computer program stored in the memory.

[0017] Fourthly, this application provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform a vision-based unmanned vehicle passability prediction and path selection method.

[0018] Compared with the prior art, this application has the following advantages: By integrating environmental perception data, vehicle dynamics models, and cargo status information, a comprehensive evaluation model is constructed. When encountering obstacles, it can not only simulate the vehicle's passing posture, but also predict the instantaneous impact load that the chassis components will bear, and assess the risk of cumulative vibration damage to the cargo, thereby improving cargo safety. Attached Figure Description

[0019] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a schematic diagram of the overall process structure of the method in this application; Figure 2 This is a schematic diagram illustrating the process of acquiring and identifying road obstacle situations using the method described in this application; Figure 3 This is a schematic diagram of the system structure of this application. Detailed Implementation

[0020] The technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments and specific features in the embodiments are detailed descriptions of the technical solution of this application, rather than limitations thereof. In the absence of conflict, the embodiments and technical features in the embodiments can be combined with each other.

[0021] Please see Figures 1 to 2 , Figure 1 This is a schematic diagram of the overall process of a vision-based unmanned vehicle passability prediction and path selection method provided in an embodiment of this application, which specifically includes the following steps: Step 1: Use visual recognition components to acquire and identify road obstacles, vehicle movement, and cargo information. In this embodiment, acquiring and identifying road obstacle conditions includes the following specific steps: The first step involves acquiring images and sizes of obstacles on the road ahead using a visual recognition component. These images and sizes are then compared to obstacle types in a database to determine image and size similarity. Image and size similarity can be obtained using any existing image similarity calculation algorithm. For example, image similarity is calculated by comparing a sequence of pixel values ​​at each point with a sequence of pixel values ​​for each obstacle type, using a cosine similarity calculation formula. Similarly, size similarity is calculated by comparing a sequence of dimensions at each location with a sequence of dimensions for each location within the obstacle type, using a cosine similarity calculation formula. This step elevates visual perception from simply seeing an obstacle to quantifying its similarity to known types, laying the foundation for subsequent accurate identification and attribute acquisition. It improves the accuracy of the recognition system by measuring two independent and complementary feature dimensions (image appearance and physical size). For example, even if changes in lighting cause image features to become blurred, its stable size features can still provide a reliable matching basis. By calculating the cosine similarity with each template in the predefined obstacle library, the system can place the currently observed obstacle in a multi-dimensional feature space and quantify its distance from various prototypes. Cosine similarity is particularly suitable for comparing shape, texture pattern (image pixel sequence) and proportional relationship (size sequence) because it is insensitive to vector amplitude, thus effectively capturing structural similarity rather than absolute numerical differences. The second step involves weighted summation of the obtained image similarity and size similarity with each obstacle type in the obstacle database to obtain the obstacle similarity score. This step, through weighted summation, merges the two independent similarity indicators from the first step into a unified and more discriminative comprehensive score. It allows the system to adjust the weights of different features according to the actual scene (e.g., higher image weight on well-lit urban roads; higher size weight at night). Image weight is directly proportional to illumination intensity, ranging from 0.4 to 0.8; while size weight is inversely proportional to illumination intensity, ranging from 0.2 to 0.6. Experiments show that under normal lighting conditions, the values ​​are: image weight 0.6; size weight 0.4. This results in a more reliable overall matching judgment, avoiding recognition errors caused by misjudgment of a single feature. In pattern recognition, multi-feature fusion usually achieves better performance than single-feature fusion. Weighted summation is an intuitive and effective linear fusion method, based on the fact that different features contribute differently to the final recognition task. By assigning different weights to image and size similarity, the system is able to simulate the human emphasis on different cues when recognizing objects. The final comprehensive obstacle similarity provides a clear and comparable scalar value for selecting the best match. The third step involves obtaining the obstacle type corresponding to the highest obstacle similarity and setting it as the identification result. This includes acquiring the elastic deformation and hardness characteristics of the corresponding obstacle, as well as its thickness and width. The acquired obstacle features are stored in the corresponding storage module. The system directly retrieves and outputs the key physical properties (elasticity, hardness, thickness, and width) of the obstacle category from the knowledge base. These properties are essential engineering parameters for subsequent risk assessment (such as calculating impact energy and deformation degree), directly converting the visual recognition results into calculable physical quantities. The system pre-constructs a mapping database (obstacle library) containing the visual features of various obstacles and their inherent physical properties. Once the most likely category is determined through similarity matching, the system no longer needs to perform complex online property analysis but directly looks up the recognized typical physical properties of that category. This method is efficient and deterministic, and its effectiveness depends on a sufficiently complete obstacle library and accurate attribute labeling. The fourth step involves acquiring information on vehicle weight, vehicle vibration during road travel, and cargo impact resistance. Vehicle weight includes the total weight of the vehicle and cargo, while cargo impact resistance includes the safe vibration amplitude of the cargo. The system shifts from a general assessment of obstacles to a specific risk assessment for this particular journey. It quantifies the current state of the risk-bearing parties (vehicle and cargo), allowing for dynamic adjustments to subsequent safety decisions (such as whether to proceed and at what speed). For example, for the same obstacle, vibration must be avoided as much as possible when fully loaded with fragile goods, while it can be tolerated to a certain extent when empty with solid cargo. The vehicle's total mass directly determines its inertia, affecting kinetic energy changes and suspension system response when passing obstacles. Vibration during travel reflects the real-time operating state of the suspension system and road surface background excitation, serving as a baseline for assessing additional impact effects. The safe vibration amplitude of the cargo defines the constraints of this transportation mission. Risk assessment must be bidirectional: it must assess both the strength of the threat source (obstacle) and the vulnerability of the load-bearing components (vehicle-cargo system) to make intelligent decisions that protect the vehicle structure while meeting the requirements of the transportation mission. Step 2: Perform vehicle travel deformation analysis based on vehicle driving conditions, cargo conditions, and identified obstacles; In this embodiment, the deformation analysis of the driving process includes the following specific steps: The first step involves obtaining information about the vehicle's chassis, obstacle height, and the sharpness and hardness of the obstacle's edge facing the chassis. Sharpness is represented by the cosine of the angle. Obstacle hardness anomalies are obtained by dividing obstacle hardness by chassis hardness, height anomalies by dividing obstacle height by chassis height, and sharpness anomalies by ensuring the obstacle's sharpness remains within a safe range. The weighted sum of hardness, height, and sharpness anomalies yields the damage anomalies to the chassis. By quantifying these three key geometric and material property anomalies (height, sharpness, and hardness), the potential damage to the chassis from the obstacle is transformed from a qualitative concern into a quantitative damage anomaly. The weighted summation method allows for adjusting risk weights based on different chassis design characteristics (e.g., off-road vehicles prioritize height, while precision instrument transport vehicles prioritize sharpness), making the assessment more targeted. This step quickly identifies extreme obstacles that will inevitably cause chassis scratches, punctures, or impacts. The second step involves obtaining information on vehicle weight, the elastic deformation of the corresponding obstacle, and its hardness characteristics. The deformation of the obstacle under impact is obtained by dividing the vehicle weight (converted to gravity) by the obstacle's elastic deformation coefficient. The vehicle's loaded thickness is obtained by subtracting the influence coefficient of the deformation on the thickness from the overall thickness and multiplying by the deformation value. The abnormal vehicle vibration height is obtained by dividing the vehicle's loaded thickness by the safe height, and the abnormal hardness is obtained by dividing the hardness of the corresponding obstacle by the safe hardness. The focus is on analyzing the dynamic behavior of the vehicle-obstacle coupling system. By introducing vehicle weight (excitation source) and obstacle elasticity (buffer medium), the estimated crushing deformation and the resulting vehicle loaded thickness are calculated, thus deriving the abnormal height that may cause severe vibrations. This step accurately simulates a typical scenario of a vehicle bumping through an elastic obstacle (such as a speed bump). Its evaluation objective is to protect the integrity of the cargo and prevent damage or component fatigue due to excessive vibration. The third step compares the abnormal damage to the chassis caused by the obstacle with the set abnormal damage threshold. If the abnormal damage to the chassis is greater than or equal to the set abnormal damage threshold, a detour is directly determined; if the abnormal damage to the chassis is less than the set abnormal damage threshold, proceed to step three. Using the abnormal damage threshold as a rigid red line reflects the principle of safety first and prevention foremost. As long as the direct damage risk exceeds the safety tolerance, regardless of the subsequent vibration situation, a detour is directly determined. This avoids unnecessary complex calculations or risky attempts when there is a definite damage risk. Step 3: Analyze the obstacle deformation analysis results and vehicle driving conditions to predict vehicle overpass vibrations; In this embodiment, the vehicle overcoming vibration prediction includes the following specific aspects: The first step involves obtaining the abnormal vehicle vibration height and hardness at the corresponding obstacle, and then weighting and summing them to obtain the obstacle obstruction anomaly. It also involves obtaining the obstruction anomaly of the corresponding vehicle when crossing historical obstacles during this trip, as well as the vibration amplitude and frequency at the time of crossing. The vibration anomaly is calculated using the average vibration amplitude and average vibration frequency of historical vehicle crossings. The calculation method for vibration anomaly is as follows: dividing the average vibration amplitude by the safe amplitude yields the amplitude anomaly; dividing the average vibration frequency by the safe frequency yields the frequency anomaly; and multiplying the amplitude anomaly by the frequency anomaly yields the vibration anomaly of the corresponding obstacle. This process combines the vibration anomaly and obstacle obstruction anomaly obtained during the current trip. Instead of simply recording that the vehicle has crossed a particular obstacle, it quantifies the dynamic response (vibration amplitude and frequency) and static attributes (obstruction anomaly) at the time of crossing. By comparing the average vibration data with a safety threshold, the vibration anomaly is calculated, directly reflecting the actual dynamic impact level of the obstacle on the vehicle's suspension system and cargo. At the same time, by associating it with its inherent obstruction anomalies (based on geometry, hardness, etc.), a key database is established for subsequent analysis: that is, how much vehicle vibration (vibration anomaly) a specific type of obstacle (obstruction anomaly) usually causes. The second step involves dividing the vibration anomaly of historical obstacles by the obstacle's obstruction anomaly to obtain the obstacle influence coefficient. The average of this coefficient is then multiplied by the obstruction anomaly of the obstacle being analyzed to obtain the vibration anomaly caused by the obstacle in question. By calculating the obstacle influence coefficient (vibration anomaly / obstruction anomaly), the core principle implicit in historical data—the average vibration level caused by a unit obstruction anomaly—is cleverly extracted. This coefficient can be understood as the system's obstacle sensitivity learned from experience. Applied to the current new obstacle (whose obstruction anomaly is known), the average obstacle influence coefficient is calculated. It assumes a roughly linear proportional relationship between vibration anomaly and obstruction anomaly: vibration anomaly ≈ k * obstruction anomaly. The coefficient k is estimated from historical data by averaging the vibration anomaly / obstruction anomaly calculations. While simple, this model captures the core correlation of the problem. The underlying physical assumption is that for vehicles with similar chassis and loads, obstacles causing similar obstruction anomalies (such as protrusions of similar height and hardness) also cause roughly proportional vibration intensity (vibration anomaly). Step 4: Predict the cargo damage caused by vehicle vibration based on the vehicle vibration prediction results, the vehicle's historical driving data, and the cargo condition. In this embodiment, the damage to goods caused by vehicle vibration includes the following specific details: The cumulative damage impact coefficient of the cargo is obtained by multiplying the sum of vibration anomalies caused by obstacles crossed during the current trip by a historical vibration impact coefficient. The normal coefficient of the cargo is obtained by subtracting the cumulative damage impact coefficient of the cargo from a value of 1 and dividing the result by the vibration anomalies caused by the obstacle to be analyzed in this trip. Cargo damage (such as packaging wear, internal structural fatigue, and precision instrument malfunction) has a cumulative effect. By summing the vibration anomalies caused by all obstacles crossed in this trip and multiplying by a historical vibration impact coefficient learned from global historical data, the system obtains a specific cumulative damage impact coefficient for the current vehicle, the cargo in this trip, and the current road conditions. This coefficient starts at 0 and increases as the journey progresses, intuitively reflecting the total vibration pressure the cargo has already endured, providing crucial contextual information for subsequent risk decisions. Vibration anomalies can be viewed as the intensity of a stress cycle. The historical vibration impact coefficient, calibrated from long-term data, converts the intensity of a single vibration into a weighted coefficient contributing to the damage to this type of cargo. The sum of their products provides a simplified yet effective estimate of the cumulative damage already incurred during this journey. The ingenuity of the formula (1 - cumulative damage impact coefficient) / current vibration anomaly lies in its combination of two key factors: 1) the cargo's remaining health (1 - cumulative damage); 2) the intensity of the upcoming single impact (current vibration anomaly). The calculated normal coefficient is a ratio. If the coefficient is high (e.g., much greater than 1), it indicates sufficient remaining health of the cargo to easily absorb the impact, and the system can determine a low risk of passage. If the coefficient is low (close to or less than 1), it means either the cargo is already severely damaged or the impact is too severe, and the system should issue a high-level warning or recommend avoidance. Step 5: Analyze the damage to the vehicle and cargo to determine whether it is necessary to cross the corresponding road obstacle; In this embodiment, analyzing whether it is necessary to cross the corresponding road obstacle includes the following specific steps: The system compares the calculated normality coefficient of the cargo with a set normality coefficient threshold. If the normality coefficient is greater than or equal to the threshold, the cargo is in a normal condition when crossing the obstacle, and the vehicle can proceed. If the normality coefficient is less than the threshold, the cargo will be damaged, and the vehicle must detour around the obstacle. This achieves an optimal dynamic balance between ensuring cargo safety and improving transportation efficiency. It avoids both a one-size-fits-all conservative strategy (such as detouring at every bump, leading to inefficiency) and a reckless, aggressive strategy (such as forcing passage regardless of cargo condition). By comparing the dynamically calculated normality coefficient (reflecting the current remaining safety margin) with a preset threshold representing the minimum acceptable safety level, the system can provide personalized passes or prohibitions for each obstacle crossing. This significantly reduces cargo damage and hidden costs caused by accumulated vibration (such as precision instrument calibration and fragile item breakage), and maximizes path efficiency when cargo is in good condition, reducing unnecessary detours and delays.

[0022] It should be noted that the method for setting parameters not specifically described in this step is as follows: obtain at least 500 sets of historical vehicle driving conditions, cargo conditions, and road obstacle conditions data, and obtain the judgment results on whether the cargo is intact after driving through; import the historical data into each step of this application for calculation and analysis on whether detour is required; import the judgment results and calculation results into MATLAB fitting software for continuous data fitting, and output the set of parameters that meets the highest judgment accuracy.

[0023] The overall benefits of this solution are as follows: by integrating environmental perception data, vehicle dynamics models, and cargo status information, a comprehensive evaluation model is constructed; when encountering obstacles, it can not only simulate the vehicle's passing posture, but also predict the instantaneous impact loads that the chassis components will bear, and assess the risk of cumulative vibration damage to the cargo, thereby improving cargo safety.

[0024] Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of a vision-based autonomous vehicle passability prediction and path selection system provided in an embodiment of this application, including: Includes the following specific modules: The module includes a data acquisition module, an obstacle recognition module, a deformation analysis module, a vibration prediction module, a cargo damage analysis module, and a path guidance module. The data acquisition module is used to acquire information about the vehicle's movement and the cargo's status. An obstacle recognition module is used to acquire and identify road obstacles through a visual recognition component. The deformation analysis module is used to perform vehicle travel deformation analysis based on vehicle driving conditions, cargo conditions, and identified obstacle conditions. The vibration prediction module is used to analyze the vibration prediction of a vehicle crossing by combining the results of obstacle deformation analysis and the vehicle's driving conditions. The cargo damage analysis module predicts cargo damage caused by vehicle vibration based on vehicle vibration prediction results, vehicle driving conditions, and cargo conditions. The path guidance module is used to analyze the damage to vehicles and cargo to determine whether it is necessary to cross the corresponding road obstacles.

[0025] The parameters and steps for implementing the corresponding functions of each unit module in the vision-based unmanned vehicle passability prediction and path selection system of this application can be referred to the parameters and steps in the embodiments of the vision-based unmanned vehicle passability prediction and path selection method described above, and will not be repeated here.

[0026] Embodiments of this application also provide an electronic device, including a memory, a processor, and a communication bus; the memory and the processor are connected via the communication bus. The memory stores a vision-based unmanned vehicle passability prediction and path selection method that can be loaded and executed by the processor, as provided in the above embodiments.

[0027] The memory can be used to store instructions, programs, code, code sets, or instruction sets. The memory may include a program storage area and a data storage area. The program storage area may store instructions for implementing an operating system, instructions for at least one function, and instructions for implementing the vision-based unmanned vehicle passability prediction and path selection method provided in the above embodiments. The data storage area may store data involved in the vision-based unmanned vehicle passability prediction and path selection method provided in the above embodiments.

[0028] A processor may include one or more processing cores. The processor executes instructions, programs, code sets, or instruction sets stored in memory, and calls data stored in memory to perform various functions and process data as described in this application. The processor may be at least one of a specific application-specific integrated circuit, a digital signal processor, a digital signal processing device, a programmable logic device, a field-programmable gate array, a central processing unit, a controller, a microcontroller, and a microprocessor. It is understood that, for different devices, the electronic devices used to implement the above-described processor functions may also be other types, and the embodiments of this application do not specifically limit this.

[0029] A communication bus may include a pathway for transmitting information between the aforementioned components. The communication bus can be a PCI bus or an EISA bus, etc. Communication buses can be categorized into address buses, data buses, control buses, etc.

[0030] This application provides a computer-readable storage medium storing a computer program that can be loaded by a processor and executed as described in the above embodiments, a vision-based unmanned vehicle passability prediction and path selection method.

[0031] In this embodiment, a computer-readable storage medium can be a tangible device that holds and stores instructions used by an instruction execution device. The computer-readable storage medium can be, but is not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any combination thereof. Specifically, the computer-readable storage medium can be a portable computer disk, a hard disk, a USB flash drive, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), spoofing random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory stick, floppy disk, optical disk, magnetic disk, mechanical encoding device, or any combination thereof.

[0032] The term includes, or any other variation thereof, is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0033] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the foregoing application concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions claimed in this application.

Claims

1. A vision-based unmanned vehicle passability prediction and path selection system, characterized in that, Includes the following specific modules: The module includes a data acquisition module, an obstacle recognition module, a deformation analysis module, a vibration prediction module, a cargo damage analysis module, and a path guidance module. The data acquisition module is used to acquire information about the vehicle's movement and the cargo's status. The obstacle recognition module is used to acquire and identify road obstacle conditions through a visual recognition component; The deformation analysis module is used to perform vehicle travel deformation analysis based on vehicle driving conditions, cargo conditions, and identified obstacle conditions. The deformation analysis of the driving process includes the following specific steps: The first step is to obtain information about the vehicle chassis, the height of the obstacle, and the sharpness and hardness of the side facing the chassis. Obstacle hardness is obtained by dividing the obstacle hardness by the chassis hardness, height is obtained by dividing the obstacle height by the chassis height, and sharpness is obtained by checking if the obstacle's sharpness is within a safe level. The damage abnormality of the obstacle to the chassis is obtained by weighted summation of the obstacle hardness abnormality, height abnormality, and sharpness abnormality. The second step is to obtain the vehicle weight, the elastic deformation of the corresponding obstacle, and the hardness characteristics. The deformation of the obstacle is obtained by dividing the vehicle weight into gravity and the elastic deformation coefficient of the obstacle. The vehicle loading thickness is obtained by subtracting the influence coefficient of the deformation on the thickness from the overall thickness and multiplying it by the value of the deformation. The abnormal vehicle vibration height is obtained by dividing the vehicle loading thickness by the safe height. The abnormal vehicle hardness is obtained by dividing the hardness of the corresponding obstacle by the safe hardness. The third step is to compare the abnormal damage to the chassis caused by the obstacle with the set abnormal damage threshold. If the abnormal damage to the chassis is greater than or equal to the set abnormal damage threshold, it is determined to detour directly. If the abnormal damage to the chassis is less than the set abnormal damage threshold, the vibration prediction of the vehicle crossing is analyzed based on the obstacle deformation analysis results and the vehicle driving conditions. The vibration prediction module is used to analyze the vibration prediction of the vehicle crossing by analyzing the obstacle deformation analysis results and the vehicle driving conditions. The vehicle overpass vibration prediction includes the following specific details: The first step is to obtain the vehicle vibration height anomaly and vehicle hardness anomaly corresponding to the obstacle, and then perform a weighted summation to obtain the obstacle obstruction anomaly. Next, obtain the obstruction anomaly of the corresponding vehicle when crossing historical obstacles during this trip, as well as the vibration amplitude and frequency when the vehicle crossed them. The vibration anomaly is calculated using the average vibration amplitude and average vibration frequency of historical vehicles crossing obstacles. The vibration anomaly is calculated as follows: the amplitude anomaly is obtained by dividing the average vibration amplitude by the safe amplitude; the frequency anomaly is obtained by dividing the average vibration frequency by the safe frequency; and the vibration anomaly of the corresponding obstacle is obtained by multiplying the amplitude anomaly by the frequency anomaly. This is achieved by obtaining the vibration anomaly and obstacle obstruction anomaly of the vehicle when crossing historical obstacles during this trip. The second step is to obtain the obstacle influence coefficient by dividing the vibration anomaly of the historical obstacle by the obstacle obstruction anomaly. Then, multiply the average value of the obstacle influence coefficient by the obstacle obstruction anomaly of the obstacle to be analyzed in this case to obtain the vibration anomaly caused by the obstacle to be analyzed in this case. The cargo damage analysis module predicts cargo damage caused by vehicle vibration based on vehicle vibration prediction results, vehicle driving conditions, and cargo conditions. The damage to the cargo caused by vehicle vibration includes the following specific details: The cumulative damage impact coefficient of the cargo is obtained by multiplying the sum of the vibration anomalies caused by the vehicle crossing historical obstacles during this trip by the historical vibration impact coefficient. The normal coefficient of the cargo is obtained by subtracting the cumulative damage impact coefficient of the cargo from the value of 1 and dividing it by the vibration anomalies caused by the obstacle to be analyzed in this trip. The path guidance module is used to analyze whether it is necessary to cross the corresponding road obstacle based on the damage to the vehicle and cargo.

2. The vision-based unmanned vehicle passability prediction and path selection system according to claim 1, characterized in that, The process of acquiring and identifying road obstacles includes the following specific steps: The first step is to acquire images and sizes of obstacles on the road ahead using a visual recognition component, and then compare the similarity between the images and sizes with the obstacle types in the obstacle database to obtain the image similarity and size similarity with each obstacle type in the obstacle database. The second step is to perform a weighted sum of the image similarity and size similarity with each obstacle type in the obstacle database to obtain the obstacle similarity with each obstacle type in the obstacle database. The third step is to obtain the obstacle type corresponding to the highest obstacle similarity and set it as the recognition result of the corresponding obstacle. The elastic deformation and hardness features of the corresponding obstacle are obtained; at the same time, the thickness and width features are obtained; and the obtained obstacle features are stored in the corresponding storage module. The fourth step is to obtain information on vehicle weight, vehicle vibration while driving on the road, and cargo impact resistance. Vehicle weight includes the total weight of the vehicle and cargo, and cargo impact resistance includes the safe vibration amplitude of the cargo.

3. The unmanned vehicle passability prediction and path selection system based on vision detection according to claim 1, characterized in that, The analysis of whether it is necessary to cross the corresponding road obstacle includes the following specific steps: The calculated normal coefficient of the cargo is compared with the set normal coefficient threshold. If the normal coefficient of the cargo is greater than or equal to the set normal coefficient threshold, it means that the cargo is in a normal state when crossing, and it is determined that the vehicle can cross the corresponding road obstacle. If the normal coefficient of the cargo is less than the set normal coefficient threshold, it means that the cargo will be damaged when crossing, and it is determined that the vehicle needs to go around the corresponding road obstacle.

4. The unmanned vehicle passability prediction and path selection system based on vision detection according to claim 2, characterized in that, The image similarity is calculated by importing a sequence of pixel values ​​at each point and a sequence of pixel values ​​of obstacle types into the cosine similarity calculation formula; the size similarity is calculated by importing a sequence of dimensions at each position and a sequence of dimensions of obstacle types into the cosine similarity calculation formula.

5. A method for predicting the passability and selecting the path of an unmanned vehicle based on visual detection, implemented based on the unmanned vehicle passability prediction and path selection system based on visual detection as described in any one of claims 1-4, characterized in that, The specific steps include the following: Step 1: Use visual recognition components to acquire and identify road obstacles, vehicle movement, and cargo information. Step 2: Perform vehicle travel deformation analysis based on vehicle driving conditions, cargo conditions, and identified obstacles; Step 3: Analyze the obstacle deformation analysis results and vehicle driving conditions to predict vehicle overpass vibrations; Step 4: Predict the cargo damage caused by vehicle vibration based on the vehicle vibration prediction results, the vehicle's historical driving data, and the cargo condition. Step 5: Analyze the damage to the vehicle and cargo to determine whether it is necessary to cross the corresponding road obstacle.

6. An electronic device, comprising: A processor and a memory, wherein the memory stores a computer program that can be called by the processor; characterized in that the processor executes, as described in claim 5, a vision-based unmanned vehicle passability prediction and path selection method by calling the computer program stored in the memory.

Images