Intelligent path analysis method for new energy vehicle on fire

Abstract

The invention relates to the field of on-fire vehicle self-recognition, in particular to an intelligent path analysis method for on-fire new energy vehicles. The method comprises the following steps: S1, collecting and pre-processing fire new energy vehicle environment data, the vehicle environment data comprising temperature sensing data and license plate color data, performing data fusion processing on the pre-processed vehicle environment data to obtain fused vehicle environment data, and performing adaptive denoising processing on the fused vehicle environment data to obtain a new energy vehicle environment data set; obtaining automobile environment data after noise filtering; and S2, generating an initial driving path of the new energy vehicle based on the basic information data after intelligent processing, and optimizing the initial driving path based on the initial driving path and the environment information data after intelligent processing to obtain a predictive driving path. The technical problem of inaccurate control caused by inaccurate recognition of a new energy vehicle on fire and inaccurate planning of a fire avoidance path is solved.

Description

Technical Field

[0001] The present invention relates to the field of self-identification of vehicles on fire, and in particular to an intelligent path analysis method for new energy vehicles on fire. Background Art

[0002] With the popularization of new energy vehicles, there are more and more accidents involving new energy vehicles catching fire when they are involved in traffic accidents. For example, the collisions with guardrails and trees reported in the news before did not cause great impacts, but the new energy vehicles quickly caught fire and caused casualties of varying degrees. With the increasing number of motor vehicles and the frequent occurrence of traffic accidents, the requirements for the safety and reliability of new energy vehicles are also getting higher and higher. Therefore, the problem of how to plan an avoidance path after a collision of new energy vehicles needs to be solved urgently.

[0003] In particular, there are still many technical challenges in data acquisition, environmental perception, path planning and vehicle control.

[0004] Therefore, an intelligent path analysis method for new energy vehicles on fire is proposed to improve the accuracy of identifying new energy vehicles on fire and improve the accuracy of avoidance paths for new energy vehicles on fire. Summary of the invention

[0005] The present invention provides an intelligent path analysis method for a new energy vehicle on fire, so as to solve the technical problem of inaccurate control caused by inaccurate identification of the new energy vehicle on fire and inaccurate fire avoidance path planning.

[0006] The present invention provides an intelligent path analysis method for a new energy vehicle on fire, which specifically includes the following technical solutions:

[0007] An intelligent path analysis method for a new energy vehicle on fire comprises the following steps:

[0008] An intelligent path analysis method for a new energy vehicle on fire, characterized in that it comprises the following steps:

[0009] S1. Collect and pre-process the environmental data of the new energy vehicle on fire, the vehicle environmental data includes temperature sensing data and license plate color data, perform data fusion processing on the pre-processed vehicle environmental data to obtain fused vehicle environmental data, and perform adaptive denoising processing on the fused vehicle environmental data to obtain noise-filtered vehicle environmental data;

[0010] S2. Generate a preliminary driving path for the new energy vehicle based on the basic information data after intelligent processing, optimize the preliminary driving path based on the preliminary driving path and the environmental information data after intelligent processing to obtain a predictive driving path, formulate a driving strategy based on the predictive driving path, and generate operating instructions; convert the operating instructions into vehicle control instructions to control the actual actions of the vehicle.

[0011] Preferably, S1 specifically includes:

[0012] The preprocessing algorithm for preliminary denoising and filtering is used to preprocess the original environmental data through a combination of short-time Fourier transform and wavelet transform to achieve multi-level time-frequency domain analysis.

[0013] Preferably, S1 specifically includes:

[0014] In the process of implementing the preprocessing algorithm based on multi-layer time-frequency domain transformation and preliminary denoising filtering, sub-band signals at different levels are obtained by performing multi-layer time-frequency domain transformation on the original environmental data; an improved non-local mean algorithm is introduced to perform preliminary denoising filtering on the sub-band signals.

[0015] Preferably, S1 specifically includes:

[0016] The pre-processed environmental data is fused using a fusion algorithm based on multi-dimensional weight optimization. The fusion algorithm based on multi-dimensional weight optimization introduces a multi-dimensional weight optimization mechanism to dynamically adjust the weights of different sensor data, monitor the temperature of new energy vehicles, and also monitor the license plate color of new energy vehicles.

[0017] Preferably, S1 specifically includes:

[0018] The environmental data after noise filtering is enhanced by a data enhancement algorithm based on a multi-level recursive convolutional network; the environmental data after noise filtering is predicted and corrected by combining multi-level recursive convolution with a temporal neural network.

[0019] Preferably, S2 specifically includes:

[0020] The driving trajectory of the new energy vehicle that caught fire is intelligently estimated during driving, and multi-level fitness calculations are performed during the path optimization process.

[0021] Preferably, S2 specifically includes:

[0022] The driving trajectory data is processed nonlinearly to generate comprehensive automobile road information data.

[0023] Preferably, S2 specifically includes:

[0024] According to the vehicle road information data, the new energy vehicle that caught fire automatically went to a road section with fewer vehicles and people, and reported the location information of the new energy vehicle that caught fire to the police.

[0025] The beneficial effects of the technical solution of the present invention are:

[0026] 1. Car license plates are binarized to form binary black-and-white images, reducing the impact of other colors on the recognition of green. When the brightness and saturation of the license plate are ensured, the green range occupies more than 60% of the overall range of the license plate, and the vehicle can be identified as a new energy vehicle, reducing the impact of license plate obstructions on license plate recognition and improving recognition accuracy.

[0027] 2. The pre-processed environmental data is fused using a fusion algorithm based on multi-dimensional weight optimization. The fusion algorithm based on multi-dimensional weight optimization introduces a multi-dimensional weight optimization mechanism to dynamically adjust the weights of different sensor data, monitor the temperature of new energy vehicles, and also monitor the license plate color of new energy vehicles to achieve accurate data fusion.

[0028] 3. The environmental data after noise filtering is stabilized using a data enhancement algorithm based on a multi-level recursive convolutional network. The data enhancement algorithm combines multi-level recursive convolution with a temporal neural network to perform multi-level prediction and correction on the environmental data after noise filtering to ensure the stability of the data in dynamic scenes. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 A flow chart of an automatic control method for an unmanned vehicle of the present invention;

[0030] Figure 2 This is a rendering of the intelligent system installed in the present invention. DETAILED DESCRIPTION

[0031] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0032] S1. Collect and pre-process the environmental data of the new energy vehicle on fire, the vehicle environmental data includes temperature sensing data and license plate color data, perform data fusion processing on the pre-processed vehicle environmental data to obtain fused vehicle environmental data, and perform adaptive denoising processing on the fused vehicle environmental data to obtain noise-filtered vehicle environmental data;

[0033] The specific implementation process is as follows:

[0034] S1, specifically includes:

[0035] S11. Using a near-infrared depth camera, obtain a car license plate image, and obtain the vertical height of each pixel point and the near-infrared depth camera and the RGB data of each pixel point according to the car license plate;

[0036] S111. Obtain the saturation S of the pixel according to the brightness V and the minimum RGB data;

[0037] S112. Obtain the hue H of the pixel according to the RGB data, the minimum RGB data, and the brightness V;

[0038] S113. Obtain HSV data according to the brightness V, saturation S, and hue H of the pixel point.

[0039] S12. Convert RGB data into HSV data;

[0040] S13. Determine a threshold condition according to the HSV data, and binarize the car license plate image according to the threshold condition to obtain a binary image;

[0041] S14. Identify the boundary contour line of the car license plate image in the binary image to obtain the boundary contour area;

[0042] Specifically, the near-infrared depth camera is used to illuminate and identify the car license plate image, and the original data group (X, Y, Z, G) is output; wherein X represents the horizontal coordinate point of the car license plate image pixel, Y represents the vertical coordinate point of the car license plate image pixel, Z represents the vertical height between the car license plate image pixel and the near-infrared depth camera, and G represents the intensity value of the car license plate image pixel in the green channel.

[0043] Convert RGB data to HSV data.

[0044] This step is used to convert the RGB data of each pixel into HSV data.

[0045] The first column data X, the second column data Y, and the fourth column data G are extracted from the original data group (X, Y, Z, G) to form an intermediate data group (X, Y, G).

[0046] Traverse the RGB data of each pixel of the car license plate image to obtain the maximum RGB data and the minimum RGB data of the pixel.

[0047] For each pixel, compare the intensity values ​​of the three channels R, G, and B to find the maximum and minimum values, which are used to calculate the brightness V and saturation S.

[0048] The brightness V of the pixel is determined based on the maximum RGB data; the brightness (V) in the HSV color space is determined by the maximum value of the RGB value. For this, the brightness V formula is: Where N i is the number of pixels.

[0049] According to the brightness V and the minimum RGB data, the saturation S of the pixel is obtained.

[0050] This embodiment calculates the saturation based on the calculated brightness and the minimum RGB data:

[0051] According to the RGB data, the minimum RGB data and the brightness V, the hue H of the pixel is obtained.

[0052] According to the brightness V, saturation S and hue H of the pixel point, HSV data is obtained.

[0053] According to the above steps, HSV data of the HSV color space can be calculated, and the HSV data is combined with X and Y into a color data group (X, Y, H, S, V).

[0054] The threshold condition is determined according to the HSV data, and the automobile license plate image is binarized according to the threshold condition to obtain a binary image; this embodiment traverses the HSV data of each pixel point, and the lower limit (0, 0, 200) of the HSV data of the pure green color on the surface of the automobile license plate image and the upper limit (180, 25, 255) of the HSV data are preferably used as the threshold condition, and the automobile license plate is binarized to form a binary black and white image, thereby reducing the influence of other colors on the recognition of green. When the brightness and saturation of the license plate are ensured, the green range occupies more than 60% of the overall range of the license plate, and the vehicle can be identified as a new energy vehicle, thereby reducing the influence of license plate obstructions on license plate recognition and improving recognition accuracy.

[0055] like Figure 1-2 As shown in the figure, the original environmental data is transformed into multiple layers in the time and frequency domain to obtain sub-band signals at different levels. Multi-level time and frequency domain analysis is achieved through the combination of short-time Fourier transform and wavelet transform.

[0056] x is (t,f)=NKFT(MT(x(t)))[i,s]

[0057] Among them, x(t) represents the original environmental data in the time-frequency domain, t is the time variable, which represents the time point of data collection; f is the frequency variable, which represents the frequency point of data collection; x is (t, f) represents the subband signal of the sth level of sensor i at time t and frequency f; NKFT represents short-time Fourier transform; MT represents wavelet transform. The purpose of this is to decompose the original environmental data into different frequency components for subsequent processing.

[0058] Next, the sub-band signals of each level are subjected to preliminary denoising filtering and processed using the improved non-local means algorithm. The specific steps are as follows:

[0059] The non-local mean algorithm is used for preliminary denoising filtering, and the calculation formula is as follows:

[0060]

[0061] in, is the subband signal of the sth level after preliminary denoising filtering at time t and frequency f of sensor i; h is is the similarity parameter; k is the neighborhood index in the non-local means algorithm, indicating the kth neighborhood point in the search window; Ω is is the search window, which represents the neighborhood range in the non-local means algorithm; Z is (t, f) is the normalization factor; x is (t+k, f) represents the subband signal of the sth level after preliminary denoising filtering of sensor i at time t+k and frequency f.

[0062] Through the above processing steps, the pre-processed environmental data is obtained.

[0063] Furthermore, the pre-processed environmental data is fused using a fusion algorithm based on multi-dimensional weight optimization. The fusion algorithm based on multi-dimensional weight optimization dynamically adjusts the weights of different sensor data by introducing a multi-dimensional weight optimization mechanism, monitors the temperature of new energy vehicles, and also monitors the license plate color of new energy vehicles to achieve accurate data fusion. The specific implementation process is as follows:

[0064] Calculate the multidimensional weight of each sensor data. The specific formula is as follows:

[0065]

[0066]

[0067] Among them, w is (t, f) is the weight of the preprocessed environmental data of sensor i on the sub-band signal of the sth level at time t and frequency f; γ is (t, f) is the weight adjustment factor; L is (t, f) is the similarity score, which measures the similarity between the sub-band signal of sensor i at the sth level and other sensor data at time t and frequency f; S is the total number of sub-band signals, indicating the number of sensors involved in data fusion; α is an adjustment parameter used to control the sensitivity of the weight adjustment factor; w is an adjustment parameter used to determine the threshold of the sensor data quality score; Q is(t, f) is the sensor data quality score, which represents the sensor data quality of the subband signal at the sth level of sensor i at time t and frequency f. is The value of (t, f) controls the weight distribution of each sensor.

[0068] Next, data fusion is performed based on multi-dimensional weights to integrate the data from each sensor and obtain fused environmental data. The calculation formula is as follows:

[0069]

[0070] Among them, x f (t, f) is the fused environmental data, is the subband signal of the sth level after preliminary denoising and filtering of sensor i at time t and frequency f, and N is the total number of sensors. Through multi-dimensional weight optimization, the information of each sensor is integrated to ensure the accuracy and consistency of the data.

[0071] The environmental data after noise filtering is stabilized using a data enhancement algorithm based on a multi-level recursive convolutional network. The data enhancement algorithm combines multi-level recursive convolution with a temporal neural network to perform multi-level prediction and correction on the environmental data after noise filtering to ensure the stability of the data in dynamic scenes. The specific implementation process is as follows:

[0072] Perform multi-level recursive convolution processing on the noise-filtered environmental data to extract spatiotemporal features. The calculation formula is as follows:

[0073]

[0074] in, is the spatiotemporal convolution feature of the lth convolutional layer at time t, is the activation function, which is selected based on expert experience. is the convolution kernel, x′ f (t, f) is the environmental data after noise filtering, is the bias term of the lth convolutional layer at time t, is the feature weight coefficient of the lth convolutional layer at time t to the mth convolutional layer at time t-1, is the number of convolutional layers, is the spatiotemporal convolution feature of the mth convolutional layer at time t-1.

[0075] Next, the existing time series neural network is used to predict and correct the multi-level convolution features, capture the dependencies in the time series, adapt to dynamically changing scenarios, and ensure data stability. The calculation formula is as follows:

[0076]

[0077] in, is the hidden state of the l-th layer of the temporal neural network at time t, is the weight matrix from the input to the hidden state of the l-th layer of the temporal neural network, is the weight matrix from hidden state to hidden state of the l-th layer of temporal neural network, is the bias term of the l-th layer of the temporal neural network, is the weight matrix from the hidden state to the output of the l-th layer of the temporal neural network, is the output bias term of the l-th layer of the temporal neural network, is the activation function, is the output data of the lth layer of the time series neural network after correction.

[0078] Through the processing of the data enhancement algorithm based on the multi-level recursive convolutional network, the environmental data with enhanced stability is obtained.

[0079] The existing feature engineering technology is used to extract features of the environmental data with enhanced stability to obtain environmental feature information.

[0080] Based on the environmental data and environmental feature information with enhanced stability, in-depth analysis and identification are performed to obtain obstacle recognition results, and various obstacles are classified based on the obstacle recognition results to achieve environmental perception analysis for intelligent driving.

[0081] S2. Generate a preliminary driving path for the new energy vehicle based on the basic information data after intelligent processing, optimize the preliminary driving path based on the preliminary driving path and the environmental information data after intelligent processing to obtain a predictive driving path, formulate a driving strategy based on the predictive driving path, and generate operating instructions; convert the operating instructions into vehicle control instructions to control the actual actions of the vehicle.

[0082] At the same time, the burning car will make an intelligent prediction of its driving trajectory during the driving process. During the path optimization process, the result of the node cost function will be used as input, which will directly affect the objective function of the intelligent multi-layer path optimization algorithm and integrate multiple levels of environmental adaptability.

[0083] The first layer fitness is calculated as follows:

[0084]

[0085] in, It is the first-level fitness of path P, which is used to measure the matching degree between the path and the environmental information data after intelligent processing and the smoothness of the path; It is the standardized error of the environmental information data after intelligent processing, which is used to evaluate the path and time point Environmental information data after intelligent processing The degree of deviation; S is the number of time steps, which means the total number of time steps from the initial moment to the final moment; Indicates time point Environmental information data after intelligent processing; It is the mean value of the environmental information data after intelligent processing, indicating the expected value of the environmental information data; It is the standard deviation of the environmental information data after intelligent processing, indicating the degree of dispersion of the environmental information data; is the smoothness of the path, which represents the rate of change of the path point P(t) between time t and t+1; P(t) is the path point at time t, which represents the position of the vehicle at time t; Δt is the time interval between adjacent time points, which is used to calculate the rate of change between path points; is a regularization parameter used to balance the weight between path smoothness and the normalized error of environmental information data.

[0086] The second layer fitness is calculated as follows:

[0087]

[0088] in, is the second-layer fitness of path P; Represents the sum of the costs of all nodes in the path, directly using the result of the node cost function; M is the total number of nodes.

[0089] Define a multi-objective optimization function; integrate multiple levels of fitness to construct a multi-objective optimization function. The formula of the multi-objective optimization function is as follows:

[0090]

[0091] in, is the total fitness of path P; ω 1 ,ω 2 It is the multi-objective optimization weight, which is used to balance the fitness of the first layer and the fitness of the second layer.

[0092] Each path segment contains continuous path points. The location information and time point are extracted for each path point. Based on the extracted location information and time point, the speed and acceleration between each path point are calculated. Based on the speed and acceleration information, a specific driving strategy is formulated, which may include steering angle, target speed, acceleration and deceleration instructions; for example, when the acceleration is greater than the preset threshold, an acceleration operation is planned; when the acceleration is less than the preset threshold, a deceleration operation is planned; when the acceleration is equal to the preset threshold, the speed remains unchanged. According to the driving strategy, an operation instruction is generated, and the operation instruction is converted into a vehicle control instruction to control the actual action of the vehicle and realize the automatic control of the unmanned vehicle.

[0093] Obtaining basic road information data through data sources such as traffic management centers, cloud servers, and vehicle sensors. The basic road information data may include current road traffic conditions, road closure information, lane information, and traffic flow, etc.; performing preprocessing such as data cleaning and missing value filling on the obtained basic road information data to obtain preprocessed basic information data;

[0094] The pre-processed basic information data is processed using a comprehensive road information intelligent processing algorithm. The specific implementation process is as follows:

[0095] Performing data fusion on the preprocessed basic information data, weighted fusion of the preprocessed basic information data from different sources to obtain fused basic information data;

[0096] Perform nonlinear processing on driving trajectory data to generate comprehensive automobile road information data:

[0097]

[0098] Among them, N k is the k-th type of comprehensive road information data, which is the data after nonlinear processing; It is Class fusion data, such as lane information or traffic flow after fusion; α is an adjustment parameter used to balance the influence of the square and logarithmic parts in the formula; m is the total number of categories of fused data, indicating the number of data types involved in the fusion; β is an adjustment parameter used to adjust the weight of the exponential part in the nonlinear processing; γ is an adjustment parameter used to balance the influence of the square root and sine parts in the formula; δ is an adjustment parameter used to adjust the weight of the sine part in the nonlinear processing.

[0099] Perform multi-level processing and optimization on the comprehensive road information data to generate the final basic information data after intelligent processing:

[0100] Processing and optimization:

[0101]

[0102] Among them, V a It is the basic information data of the a-th category after the final intelligent processing, and is the comprehensive road information data after multi-level processing and optimization; ω is the adjustment parameter used to adjust the overall weight of each part in the formula; λ is the adjustment parameter used to adjust the weight and influence of the exponential part in the formula; u represents the total number of data categories involved in the processing.

[0103] In order to further ensure the safety and stability of path selection, nonlinear constraints are added to the controller during the solution of the objective function, and control is performed through the tolerance coefficient and connectivity factor of the link. The specific formula is:

[0104]

[0105] Among them, k e (t+1) is the tolerance coefficient of link e at time t+1, which indicates the tolerance of link e to load increase and can be obtained by the ratio of the maximum load in the history of the link to the current load; ζ e (t+1) is the connectivity factor at time t+1, which reflects the complexity of the network topology and can be obtained by the inverse of the number of links connected to the node; θ e is the angle parameter of link e, that is, the angle between the two end nodes of link e on the plane, which is used to describe the layout complexity of the physical link; C max It is a global resource constraint. The controller determines whether the path meets the tolerance requirements of the current network based on the nonlinear constraints. If a path does not meet the constraints, the controller will automatically select an alternative path or dynamically adjust the traffic on other paths to avoid network bottlenecks or security risks; global resource constraints ensure that the network will not suffer a decline in overall performance due to excessive pursuit of a single optimization goal.

[0106] The order of the embodiments of the invention is for description only and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0107] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0108] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. An intelligent path analysis method for a new energy vehicle on fire, characterized in that: The following steps are involved: S1. Collect and pre-process the environmental data of the new energy vehicle on fire, wherein the environmental data of the new energy vehicle on fire includes temperature sensing data and license plate color data, perform data fusion processing on the pre-processed vehicle environmental data to obtain fused vehicle environmental data, and perform adaptive denoising processing on the fused vehicle environmental data to obtain noise-filtered vehicle environmental data; S2. Generate a preliminary driving path for the new energy vehicle based on the basic information data after intelligent processing, optimize the preliminary driving path based on the preliminary driving path and the environmental information data after intelligent processing, obtain a predictive driving path, formulate a driving strategy based on the predictive driving path, and generate an operation instruction; Convert operating instructions into vehicle control instructions to control the actual actions of the vehicle.

2. An intelligent path analysis method for collision and fire of new energy vehicles according to claim 1, characterized in that: S1, specifically includes: S11. Using a near-infrared depth camera, obtain a car license plate image, and obtain the vertical height of each pixel point and the near-infrared depth camera and the RGB data of each pixel point according to the car license plate; S12. Convert RGB data into HSV data; S13. Determine a threshold condition according to the HSV data, and binarize the car license plate image according to the threshold condition to obtain a binary image; S14. Identify the boundary contour line of the car license plate image in the binary image to obtain the boundary contour area.

3. An intelligent path analysis method for collision and fire of new energy vehicles according to claim 2, characterized in that: Specifically include: S111. Obtain the saturation S of the pixel according to the brightness V and the minimum RGB data; S112. Obtain the hue H of the pixel according to the RGB data, the minimum RGB data, and the brightness V; S113. Obtain HSV data according to the brightness V, saturation S, and hue H of the pixel point.

4. An intelligent path analysis method for collision and fire of new energy vehicles according to claim 1, characterized in that: S1, specifically includes: The preprocessing algorithm for preliminary denoising and filtering is used to preprocess the environmental data by combining short-time Fourier transform and wavelet transform to achieve multi-level time-frequency domain analysis.

5. An intelligent path analysis method for collision and fire of new energy vehicles according to claim 2, characterized in that: S1, specifically includes: In the process of implementing the preprocessing algorithm based on multi-layer time-frequency domain transformation and preliminary denoising filtering, sub-band signals at different levels are obtained by performing multi-layer time-frequency domain transformation on the original environmental data; an improved non-local mean algorithm is introduced to perform preliminary denoising filtering on the sub-band signals.

6. An intelligent path analysis method for collision and fire of new energy vehicles according to claim 4, characterized in that: S1, specifically includes: The pre-processed environmental data is fused using a fusion algorithm based on multi-dimensional weight optimization. The fusion algorithm based on multi-dimensional weight optimization introduces a multi-dimensional weight optimization mechanism to dynamically adjust the weights of different sensor data, monitor the temperature of new energy vehicles, and also monitor the license plate color of new energy vehicles.

7. An intelligent path analysis method for collision and fire of new energy vehicles according to claim 1, characterized in that: S1, specifically includes: The environmental data after noise filtering is enhanced by a data enhancement algorithm based on a multi-level recursive convolutional network; the environmental data after noise filtering is predicted and corrected by combining multi-level recursive convolution with a temporal neural network.

8. An intelligent path analysis method for collision and fire of new energy vehicles according to claim 1, characterized in that: S2, specifically includes: The driving trajectory of the new energy vehicle that caught fire is intelligently estimated during driving, and multi-level fitness calculations are performed during the path optimization process.

9. An intelligent path analysis method for collision and fire of new energy vehicles according to claim 1, characterized in that: S2, specifically includes: The driving trajectory data is processed nonlinearly to generate comprehensive automobile road information data.

10. An intelligent path analysis method for collision and fire of new energy vehicles according to claim 9, characterized in that: S2, specifically includes: According to the vehicle road information data, the new energy vehicle that caught fire automatically goes to a road section with fewer vehicles and people, and reports the location information of the new energy vehicle that caught fire to the police.

Images