CN117125020A - Pedestrian classification and identification-based active energy absorption device for automobile and control method thereof - Google Patents
Pedestrian classification and identification-based active energy absorption device for automobile and control method thereof Download PDFInfo
- Publication number
- CN117125020A CN117125020A CN202311106575.7A CN202311106575A CN117125020A CN 117125020 A CN117125020 A CN 117125020A CN 202311106575 A CN202311106575 A CN 202311106575A CN 117125020 A CN117125020 A CN 117125020A
- Authority
- CN
- China
- Prior art keywords
- automobile
- pedestrian
- air bag
- child
- adult
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 55
- 238000010521 absorption reaction Methods 0.000 title claims description 5
- 238000013145 classification model Methods 0.000 claims abstract description 24
- 238000001514 detection method Methods 0.000 claims abstract description 21
- 230000008569 process Effects 0.000 claims abstract description 17
- 230000001133 acceleration Effects 0.000 claims description 24
- 230000005540 biological transmission Effects 0.000 claims description 19
- 239000006096 absorbing agent Substances 0.000 claims description 16
- 238000011176 pooling Methods 0.000 claims description 16
- 238000003860 storage Methods 0.000 claims description 15
- 238000012544 monitoring process Methods 0.000 claims description 12
- 230000004927 fusion Effects 0.000 claims description 11
- 238000004590 computer program Methods 0.000 claims description 6
- 230000006378 damage Effects 0.000 abstract description 11
- 238000004880 explosion Methods 0.000 abstract description 5
- 208000027418 Wounds and injury Diseases 0.000 abstract description 4
- 208000014674 injury Diseases 0.000 abstract description 4
- 238000012549 training Methods 0.000 description 13
- 230000006870 function Effects 0.000 description 8
- 238000012545 processing Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 230000004913 activation Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 230000002776 aggregation Effects 0.000 description 2
- 238000004220 aggregation Methods 0.000 description 2
- 239000003086 colorant Substances 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000001595 flow curve Methods 0.000 description 2
- 238000009863 impact test Methods 0.000 description 2
- 230000001965 increasing effect Effects 0.000 description 2
- 238000002372 labelling Methods 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 206010019196 Head injury Diseases 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 235000019879 cocoa butter substitute Nutrition 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000003116 impacting effect Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60R—VEHICLES, VEHICLE FITTINGS, OR VEHICLE PARTS, NOT OTHERWISE PROVIDED FOR
- B60R21/00—Arrangements or fittings on vehicles for protecting or preventing injuries to occupants or pedestrians in case of accidents or other traffic risks
- B60R21/34—Protecting non-occupants of a vehicle, e.g. pedestrians
- B60R21/36—Protecting non-occupants of a vehicle, e.g. pedestrians using airbags
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60R—VEHICLES, VEHICLE FITTINGS, OR VEHICLE PARTS, NOT OTHERWISE PROVIDED FOR
- B60R21/00—Arrangements or fittings on vehicles for protecting or preventing injuries to occupants or pedestrians in case of accidents or other traffic risks
- B60R21/01—Electrical circuits for triggering passive safety arrangements, e.g. airbags, safety belt tighteners, in case of vehicle accidents or impending vehicle accidents
- B60R21/013—Electrical circuits for triggering passive safety arrangements, e.g. airbags, safety belt tighteners, in case of vehicle accidents or impending vehicle accidents including means for detecting collisions, impending collisions or roll-over
- B60R21/0134—Electrical circuits for triggering passive safety arrangements, e.g. airbags, safety belt tighteners, in case of vehicle accidents or impending vehicle accidents including means for detecting collisions, impending collisions or roll-over responsive to imminent contact with an obstacle, e.g. using radar systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
- G06V10/449—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters
- G06V10/451—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters with interaction between the filter responses, e.g. cortical complex cells
- G06V10/454—Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
- G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/103—Static body considered as a whole, e.g. static pedestrian or occupant recognition
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/16—Anti-collision systems
- G08G1/166—Anti-collision systems for active traffic, e.g. moving vehicles, pedestrians, bikes
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60R—VEHICLES, VEHICLE FITTINGS, OR VEHICLE PARTS, NOT OTHERWISE PROVIDED FOR
- B60R21/00—Arrangements or fittings on vehicles for protecting or preventing injuries to occupants or pedestrians in case of accidents or other traffic risks
- B60R2021/003—Arrangements or fittings on vehicles for protecting or preventing injuries to occupants or pedestrians in case of accidents or other traffic risks characterised by occupant or pedestian
- B60R2021/006—Type of passenger
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mechanical Engineering (AREA)
- Biomedical Technology (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Databases & Information Systems (AREA)
- Remote Sensing (AREA)
- Radar, Positioning & Navigation (AREA)
- Human Computer Interaction (AREA)
- Biodiversity & Conservation Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Air Bags (AREA)
Abstract
The invention discloses an automobile active energy absorbing device based on pedestrian classification and identification and a control method thereof, wherein the automobile active energy absorbing device is of an adult/child airbag structure, and the method comprises the following steps: acquiring a real-time image of a road; inputting the road real-time image into a pedestrian classification model constructed based on a YOLOv5 network for classification detection to obtain a pedestrian classification result, namely identifying whether pedestrians exist or not and whether the pedestrians are adults or children; detecting the distance between a pedestrian and an automobile, acquiring the current vehicle state, and pre-judging that collision occurs based on the distance and the current vehicle state, and ejecting a corresponding adult type air bag or child type air bag by the automobile active energy absorbing device according to the pedestrian classification result. The technical scheme of the invention can avoid secondary injury and insufficient inflation of children caused by insufficient inflation, can not effectively protect the head of an adult, and simultaneously realizes the whole airbag point explosion and flick process before people and vehicles really contact, thereby striving for precious time for pedestrian protection.
Description
Technical Field
The invention belongs to the technical field of automobile energy absorbing devices, and particularly relates to an automobile active energy absorbing device based on pedestrian classification and identification and a control method thereof.
Background
The birth of the automobile greatly changes the travel mode of people and brings a plurality of non-negligible problems. Pedestrians are a weak group in a traffic environment, and safety problems thereof are attracting a great deal of attention. In order to prevent the occurrence of pedestrian-automobile collision accidents, combining the pedestrian detection technology with an automobile active energy absorption device is a key technology for reducing the risk of pedestrians, has important significance for improving the active safety protection capability of an automobile, and has practical significance for reducing extra expenses such as automobile maintenance.
Presently, the Volvo motor company E-Bergeniham et al, issued a "pedestrian protection airbag for a vehicle" patent in which the airbag deploys and expands to cover the windshield and/or the A-pillar when a sensor system within the vehicle detects a potential collision risk of a pedestrian or rider, such that the pedestrian/rider would hit the soft airbag.
Wu Taide et al disclose a vehicle and pedestrian protection system that deploys an external airbag to block and/or cushion contact between a vehicle carrying the external airbag and a pedestrian, animal, other vehicle, and/or stationary object. During a collision, pedestrian and vehicle protection systems activate one or more airbags by using one or more sensor controllers to determine when the airbags should deploy, using sensor feedback to detect a dangerous condition, and determining the severity of the dangerous condition.
However, in the two solutions described above, the first solution is to design a pedestrian protection airbag device, and no specific pedestrian classification recognition and targeted processing are realized. For pedestrians such as children, if the ejected pedestrian protection air bags have larger inflation quantity, secondary damage to the children can be caused; for pedestrians such as adults, if the inflated quantity of the ejected pedestrian protection air bag is smaller, the protection performance of the safety air bag cannot be realized; the second scheme uses one or more sensors to detect pedestrians or other collision objects, the sensors can receive collision information after collision, then analyze and process the collision information according to the received information, and finally the automobile control module determines whether the air bag is triggered according to the analysis result, which causes higher requirement on the air bag ejecting time, has hysteresis and still causes larger risk to the pedestrians.
Disclosure of Invention
Aiming at the problems existing in the prior art, the invention provides an automobile active energy absorbing device based on pedestrian classification and identification and a control method thereof. The method divides pedestrians into adults and children, and two different inflatable air bags are established through different objects so as to avoid secondary injury and insufficient inflation of the children caused by over-sufficient inflation, and the head of the adults cannot be effectively protected. In addition, the technical scheme of the invention introduces a YOLOv5 network to construct a pedestrian classification model for classification detection, so that early warning judgment is made on dangerous conditions before collision of people and vehicles, precious time is strived for pedestrian protection, and the whole airbag point explosion and bullet opening process is realized before people and vehicles really contact.
For this purpose, the technical scheme of the invention is as follows:
on one hand, the invention provides a control method of an automobile active energy absorbing device based on pedestrian classification and identification, wherein the automobile active energy absorbing device is arranged on an automobile head and is of an adult/child airbag structure, and the method comprises the following steps:
shooting by using a vehicle-mounted camera to obtain a road real-time image;
inputting the road real-time image into a pedestrian classification model constructed based on a YOLOv5 network for classification detection to obtain a pedestrian classification result, wherein the pedestrian classification result is used for identifying whether pedestrians exist in the road real-time image and whether the pedestrians are adults or children;
detecting the distance between a pedestrian and an automobile by using an on-board sensing device, acquiring the current automobile state, and pre-judging whether collision between the pedestrian and the automobile occurs or not based on the distance between the pedestrian and the automobile and the current automobile state;
if collision is predicted, ejecting a corresponding adult type air bag or child type air bag by the automobile active energy absorbing device according to the pedestrian classification result; otherwise, continuing monitoring.
Further optionally, the YOLOv5 network comprises an input end, a backbone network back, a neck network back and an output end head which are sequentially connected;
the SPP module in the traditional backbone network backup is replaced by a spatial pyramid pooling SPPCSPC, so that the network robustness is enhanced, the target recognition anti-interference level of children is improved, the spatial pyramid pooling SPPCSPC is divided into two transmission paths, and the first transmission path is sequentially provided with three convolution layers CBS, a maximum pooling layer, a feature fusion module and two convolution layers CBS; the second transmission path is provided with a convolution layer CBS, and the output of the last convolution layer CBS on the first transmission path and the output of the last convolution layer CBS on the second transmission path are subjected to feature fusion C and then input into the convolution layer CBS.
The technical scheme of the invention considers that: because pedestrians can be affected by light brightness, clothing, shielding and other factors, the detection effect of the network on pedestrians is reduced, and false detection is easy to occur. Meanwhile, on a picture, adults and children have different scales, the resolution ratio of the small target pedestrians is generally low, the pixel area is small, the pixels are easy to interfere with noise points in the image, and therefore the network cannot accurately position the pedestrians. In order to make the network not easy to be interfered, an SPPCSPC module is introduced to replace an SPP module, so that the robustness of the network is enhanced.
Further optionally, the YOLOv5 network comprises an input end, a backbone network back, a neck network back and an output end head which are sequentially connected;
the method comprises the steps of leading out another output path on a first C3 module on a backbone network backup, leading out another output path on a second C3 module on the backbone network backup, merging the two paths, sequentially passing through a Focus layer Focus, a convolution layer CBS and a feature merging module, and then accessing the traditional transmission path from the second C3 module to the neck network ck.
The technical scheme of the invention considers that: in road traffic environment, pedestrians are in dynamic state, so that the sizes of adults are different or the shapes of children are smaller than those of adults, and the small target objects can slowly lose the characteristic information of the pedestrians in the images along with deepening of the network depth, the identifiability is gradually reduced, only the rough outline of the pedestrians is seen, the pedestrians are more and more blurred, the pedestrians cannot be distinguished from the background, and finally the feature images extracted by the SPPCSPC module can not identify the outline of the pedestrians. In order to enhance the information acquisition capability of the network to the small target pedestrian object, a path is newly added.
From the above, in order to improve the recognition capability of the targets of adults and children, the technical scheme of the invention improves the network according to the characteristics of the real-time image of the road.
Further optionally, the current vehicle state at least includes a current vehicle speed, and the process of predicting whether a collision between a pedestrian and an automobile occurs based on the distance and the current vehicle state is as follows:
if the pedestrian is an adult, the current vehicle speed exceeds a preset vehicle speed threshold V1, and the distance is smaller than a preset distance threshold L1, the adult and the vehicle are considered to collide;
if the pedestrian is a child and the current vehicle speed exceeds a preset vehicle speed threshold V2 and the distance is smaller than a preset distance threshold L2, the child is considered to collide with the vehicle;
wherein the value ranges of the preset vehicle speed threshold V1 and the preset vehicle speed threshold V2 are 5.6 m/s-11.1 m/s;
the range of the preset distance threshold L1 and the preset distance threshold L2 is 6.7 m-13.3 m.
Further alternatively, the relationship between the gas volume of the adult-type airbag or the child-type airbag and the head acceleration and the head mass is as follows:
wherein P is the gas pressure; v is the gas volume; n is the gas quantity; r is a gas constant; t is the gas temperature; m is the head mass; s is the stress area of the head shape in the direction of the speed vector, and a is the head acceleration;
and determining the gas volumes corresponding to the expansion of the adult type air bag and the child type air bag according to the relational expression, the head type mass and the head type acceleration which are approximately equivalent to those of the adult and the child.
In two aspects, the technical scheme of the invention also provides a control system based on the method, which comprises the following steps:
the image acquisition module is used for acquiring real-time images of roads;
the pedestrian classification module is used for inputting the road real-time image into a pedestrian classification model constructed based on a YOLOv5 network for classification detection to obtain a pedestrian classification result, and the pedestrian classification result is used for identifying whether pedestrians exist in the road real-time image and whether the pedestrians are adults or children;
the collision judging module is used for acquiring the distance between the pedestrian and the automobile and the current automobile state, and predicting whether the collision between the pedestrian and the automobile occurs or not based on the distance between the pedestrian and the automobile and the current automobile state;
the control module is used for popping up a corresponding adult type air bag or child type air bag according to the pedestrian classification result if the collision is judged to occur in advance; otherwise, continuing monitoring.
In three aspects, the technical scheme of the invention provides the automobile active energy absorbing device based on the method, which comprises a child type air bag and a adult type air bag, or comprises an air bag and air bag inflating equipment, wherein the air bag inflating equipment controls the gas volume in the air bag according to the child type and the adult type to obtain the child type air bag or the adult type air bag.
In a fourth aspect, the vehicle-mounted system based on the method provided by the technical scheme of the invention at least comprises: the system comprises an automobile controller, an automobile active energy absorbing device, an on-board camera and on-board sensing equipment;
the vehicle active energy absorbing device, the vehicle-mounted camera and the vehicle-mounted sensing equipment are all connected with the vehicle controller, and the vehicle controller calls or loads a pedestrian classification model constructed based on a YOLOv5 network;
the vehicle-mounted camera transmits the shot road real-time image to the automobile controller;
the automobile controller utilizes the pedestrian classification model to carry out classification detection on the road real-time image to obtain a pedestrian classification result;
the vehicle-mounted sensing equipment detects the distance between a pedestrian and an automobile, acquires the current state of the automobile and transmits the current state of the automobile to the automobile controller;
the automobile controller pre-judges whether collision of the pedestrian and the automobile occurs or not based on the distance between the pedestrian and the automobile and the current automobile state, if the collision is pre-judged, the automobile active energy absorption device pops up a corresponding adult type air bag or child type air bag according to the pedestrian classification result; otherwise, continuing monitoring.
In a fifth aspect, a computer readable storage medium is provided in the present invention, and stores a computer program, where the computer program is called by a processor to implement:
acquiring a real-time image of a road;
inputting the road real-time image into a pedestrian classification model constructed based on a YOLOv5 network for classification detection to obtain a pedestrian classification result, wherein the pedestrian classification result is used for identifying whether pedestrians exist in the road real-time image and whether the pedestrians are adults or children;
acquiring the distance between a pedestrian and an automobile and the current automobile state, and predicting whether collision between the pedestrian and the automobile occurs or not based on the distance and the current automobile state;
if collision is predicted, ejecting a corresponding adult type air bag or child type air bag by the automobile active energy absorbing device according to the pedestrian classification result; otherwise, continuing monitoring.
Advantageous effects
Compared with the prior art, the invention has the advantages that:
1. according to the control method provided by the technical scheme of the invention, on one hand, pedestrians are divided into adults and children, and aiming at the adults and the children, the adults and the children respectively pop up the adult type air bags and the children type air bags, so that secondary injury of the children and insufficient inflation caused by over-sufficient inflation are avoided, and the heads of the adults cannot be effectively protected. In the two aspects, the technical proposal of the invention introduces the YOLOv5 network to construct a pedestrian classification model for classification detection, so that early warning judgment is carried out on dangerous conditions before collision of people and vehicles, precious time is strived for pedestrian protection, the whole airbag point explosion and flick process is realized before the real contact of people and vehicles, the collision hazard is more effectively reduced,
2. according to the technical scheme, the traditional YOLOv5 network is optimized, in order to enhance the acquisition capability of the network to the small target feature information, the deep features and the shallow features of the YOLOv5 network are enhanced by utilizing the feature information of the shallow network, so that the feature layer output by the first C3 module and the feature layer output by the second C3 module are fused. In order to fuse the feature layers with different scales, the invention adopts a Focus layer Focus, and the aim is to realize double downsampling under the condition of not losing feature information. The invention introduces spatial pyramid pooling SPPCSPC to replace SPP modules in the original YOLOv5 network. In spatial pyramid pooling SPPCSPC, SPP contains a plurality of different MaxPools, so that the network has a plurality of receptive fields to distinguish target objects with different sizes; in CSP, the feature layer is divided into two parts, one part is subjected to conventional processing such as convolution, standardization and activation functions, the other part is subjected to SPP structure processing, and finally the two parts are subjected to superposition operation by using Concat, so that the calculation amount is reduced by half, and the speed and the precision are improved. Since pedestrians are affected by various factors such as colors, forms and environments, the invention also introduces a CA attention mechanism for extracting the characteristics which are more beneficial to network training, and aims to tell the network which contents and which positions are more needed to be focused.
3. According to the technical scheme, the expansion volume of the air bag or the gas quantity of the air bag is controlled, and the acceleration peak value of the head impact air bag is approximately controlled, so that damage to the head is reduced.
Drawings
FIG. 1 is a schematic airbag ejection view of an automotive active energy absorber;
FIG. 2 is a schematic diagram of a YOLOv5 network;
FIG. 3 is an architecture diagram of a portion of the modules of a Yolov5 network;
FIG. 4 is a mass flow curve of an adult-type airbag;
FIG. 5 is a mass flow curve of a child safety air bag;
FIG. 6 is an analysis result of an adult head type impact test, wherein, a and b are respectively a schematic diagram of the front end of an adult type impact vehicle without an airbag and a schematic diagram corresponding to an adult head type HIC; c. d, respectively representing a schematic diagram of the front end of an adult type automobile impacted with the adult type safety airbag and a schematic diagram of corresponding adult head type HIC; e. f, respectively representing a schematic view of the front end of an adult type automobile impacted with a child type safety airbag and a schematic view corresponding to the adult head type HIC;
FIG. 7 is a schematic view of a front end of a child-type impact vehicle without an airbag, and a schematic view of a corresponding child-type HIC, respectively; c. d, respectively representing a schematic view of the front end of the automobile with the adult-type safety airbag in a child-type collision mode and a schematic view of the corresponding child head-type HIC; e. and f, respectively representing a schematic view of the front end of the automobile provided with the child safety airbag in the child type impact and a schematic view of the corresponding child head HIC.
Detailed Description
The invention provides an automobile active energy absorber based on pedestrian classification and identification and a control method thereof, wherein the pedestrian classification is firstly proposed, corresponding air bags are popped up in a targeted manner aiming at adults and children, secondary injury and insufficient inflation of the children caused by insufficient inflation are avoided, and the head of the adults cannot be effectively protected. In addition, a YOLOv5 network is introduced to construct a pedestrian classification model for classification detection, so that early warning judgment on dangerous conditions before collision of people and vehicles is possible, and the whole airbag point explosion and flicking process is realized before the people and vehicles are in real contact. The invention will be further illustrated with reference to examples.
As shown in fig. 1, the invention is provided with an automobile active energy absorber, in particular an adult/child airbag structure, on the front automobile of the automobile, the airbag after ignition can cover dangerous areas such as the rear edge of an engine cover, the front edge of a windshield, an A column and the like, and the head of a pedestrian can collide on the soft airbag, so that the risk of head injury of the pedestrian is reduced. In some embodiments, the automotive active energy absorber comprises a child-type air bag, an adult-type air bag. Namely, two air bags are arranged, and corresponding adult air bags are popped up in a targeted manner. In some embodiments, the active energy absorber comprises an airbag and an airbag inflation device, wherein the airbag inflation device controls the gas volume in the airbag according to the type of child and adult to obtain a child type airbag or an adult type airbag, namely, the airbag inflation device controls the gas volume in the airbag to construct a human type airbag or a child type airbag at best.
Based on an automobile active energy absorbing device, the control method of the automobile active energy absorbing device based on pedestrian classification and identification provided by the technical scheme of the invention comprises the following steps:
shooting by using a vehicle-mounted camera to obtain a road real-time image;
inputting the road real-time image into a pedestrian classification model constructed based on a YOLOv5 network for classification detection to obtain a pedestrian classification result, wherein the pedestrian classification result is used for identifying whether pedestrians exist in the road real-time image and whether the pedestrians are adults or children;
detecting the distance between the pedestrian and the automobile by using the vehicle-mounted sensing equipment, acquiring the current automobile state, and pre-judging whether collision between the pedestrian and the automobile occurs or not based on the distance and the current automobile state;
if collision is predicted, ejecting a corresponding adult type air bag or child type air bag by the automobile active energy absorbing device according to the pedestrian classification result; otherwise, continuing monitoring.
As shown in fig. 2, the YOLOv5 network architecture of the pedestrian classification model provided by the technical scheme of the invention is as follows:
the improved YOLOv5 network comprises an input end, a backbone network backbone, a neck network neg and an output end head. The input end preprocesses the input road real-time image, and in the embodiment of the invention, a Mosaic data enhancement method is adopted, 4 pictures in the data set are arbitrarily selected in the training process to perform operations such as rotation scaling and the like, and are spliced in a random manner to form a new training set, so that the robustness and generalization capability of the network are enhanced. In other possible embodiments, the preprocessing technique is not limited to the above examples, and the preprocessing means for enhancing the image features can be considered as falling within the scope of the present invention.
The backbone network backup uses a traditional architecture, performs network optimization according to the characteristics of real-time images of roads and the characteristics of adults and children, improves the recognition capability of adults and children, and specifically comprises a Focus layer Focus, a convolution layer CBS (Conv2D_BN_SiLU), a C3 module, a CA attention module and a spatial pyramid pooling SPPCSPSPC. In this embodiment, the Focus layer Focus, 2 convolution layer CBSs, the first C3 module, the convolution layer CBS, the second C3 module, the convolution layer CBS, the third C3 module, the convolution layer CBS, the CA attention module, the spatial pyramid pooling SPPCSPC, and the fourth C3 module are sequentially connected. According to the technical scheme, the SPP module of the traditional backbone network backup is replaced by the spatial pyramid pooling SPPCSPC.
The Focus layer Focus realizes slicing operation, and downsampling operation is carried out on an input feature layer, namely, a picture is sampled every other pixel point to obtain 4 independent feature layers, so that RGB 3 channels of an original image are expanded by 4 times, and meanwhile, the height and width of the original image are reduced by 1/2 of the original image. And the C3 module maps the image features into two parts, one part carries out residual convolution, the other part carries out convolution layer CBS processing, and then the two parts carry out Concat superposition operation, so that the features of the shallow layers are separated, and the gradient disappearance problem is relieved.
Since pedestrians are affected by various factors such as colors, forms and environments, the invention also introduces a CA attention mechanism for extracting the characteristics which are more beneficial to network training, and aims to tell the network which contents and which positions are more needed to be focused. The CA attention mechanism firstly carries out global average pooling on the height and the width respectively, uses rolling and splicing to carry out dimension reduction processing on an input image, then carries out dimension increasing to obtain attention weights in two directions, and finally carries out multiplication weighting operation on the obtained weights and the input feature images.
Traditional YOLOv5 uses SPP modules that employ 3 different max pooling layers to enhance the network receptive field. In order to enable the network to extract more reliable characteristic information and improve the detection precision of the network, in the CSP, firstly, a characteristic layer is divided into two parts, wherein one part is subjected to conventional processing such as convolution, standardization and activation functions, the other part is subjected to SPP structure processing, and finally, the two parts are subjected to Concat superposition operation, so that the method is beneficial to reducing half of calculated amount, and the speed and the precision are improved. The invention introduces spatial pyramid pooling SPPCSPC to replace SPP modules in the original YOLOv5 network. The space pyramid pooling SPPCSPC is divided into two transmission paths, wherein the first transmission path is sequentially provided with three convolution layers CBS, a maximum pooling layer, a feature fusion module and two convolution layers CBS; the second transmission path is provided with a convolution layer CBS, and the output of the last convolution layer CBS on the first transmission path and the output of the last convolution layer CBS on the second transmission path are subjected to feature fusion C and then input into the convolution layer CBS. Among these, the convolutional layer CBS is a network basic component that consists of convolutional, normalization, and activation functions.
The neck network neg, also known as an enhanced feature extraction network, employs a feature pyramid and path aggregation network architecture. The feature pyramid structure fuses high-level semantic information with shallow semantic information through upsampling, so that the learning capability of the network on the features is enhanced. The path aggregation network structure is provided with a feature fusion path from bottom to top on the feature pyramid structure, so that shallow information can be better transferred to a high layer, and the detection performance of the network is further improved. The technical scheme of the invention optimizes the traditional neck network neg, and the second C3 module optimizes the transmission path of the traditional neck network neg, specifically: another output path is led out from the first C3 module, another output path is led out from the second C3 module, the two paths are fused, and then the two paths sequentially pass through the Focus layer Focus, the convolution layer CBS and the feature fusion module and then are connected into the traditional neck network neg. It should be noted that, as the depth of the network deepens, the size of the feature map output is continuously reduced after multiple convolution operations, which finally causes the loss of the small target feature information contained in the feature map in the transmission process. Therefore, in order to enhance the acquisition capability of the network to the small target feature information, the invention utilizes the feature information of the shallow network to enhance the fusion of the deep features and the shallow features of the YOLOv5 network, so that the feature layer output by the first C3 module and the feature layer output by the second C3 module are fused. In order to fuse the feature layers with different scales, the invention adopts a Focus layer Focus, and the aim is to realize double downsampling under the condition of not losing feature information. And converting 160 x 160 feature layers into 80 x 80 feature layers by utilizing Focus layer Focus, then adopting standard convolution to enable the channel number of the feature layers to be consistent with the output channel number of the second C3 module, and finally utilizing Concat superposition operation to complete feature fusion.
The output end obtains the classification result, and the present embodiment adopts the GIOU function as the loss function of the bounding box, and uses non-maximum suppression to screen the multi-objective box.
The embodiment of the invention preferably adopts an improved YOLOv5 network to construct a pedestrian classification model, wherein the input data of the pedestrian classification model is a road real-time image, and the output result is a pedestrian recognition result and a pedestrian category (adult and child). In order to perform pedestrian recognition by using the pedestrian classification network, network training needs to be performed in advance, and the following will exemplify a training process:
1. a sample dataset is prepared. The Labelimg software is used for selecting, modifying and remarking classification (pedestrian annotation and pedestrian category annotation) on the basis of the existing target data set (composed of road images), and the remarked categories are classified into adults and children. The data set is 14000 in total, wherein 21070 for adult labeling samples and 6600 for child labeling samples comprise diversified pedestrian picture samples under different geographic environments, weather, light and shade changes, different sexes and the like.
2. Again, the YOLOv5 model was trained on the PC side. And inputting the manufactured data set, carrying out iterative updating on network parameters through propagation, calculating a loss function, and stopping training when the training times are reached or the set threshold requirement is met.
3. After training, judging a training result according to the evaluation index. In the training process, setting the IOU value of the YOLOv5 to be 0.5, and when the threshold exceeds 0.5, representing that the current target is judged correctly, and reserving a prediction frame; if the threshold value is not more than 0.5, the judgment of the current target is wrong, and the prediction frame is not reserved. In YOLOv5, the higher the accuracy value, the fewer false positives. The higher the recall, the less missed the test.
As shown in fig. 3, the training process of the YOLOv5 network is a prior art implementation process, and the loss function related to the training process may refer to the prior art, which is not specifically limited in the present invention. The trained pedestrian classification model is loaded into the automobile controller.
And inputting the acquired road real-time image into a pedestrian classification model to obtain a pedestrian classification result, and synchronously or later, detecting the distance between the pedestrian and the automobile by utilizing vehicle-mounted sensing equipment and acquiring the current automobile state, so as to predict whether the collision between the pedestrian and the automobile occurs or not based on the distance and the current automobile state.
In the embodiment, if the current pedestrian is detected to be an adult, the speed exceeds 8.52m/s and the distance is smaller than 10m, the automobile control module sends a command to the pedestrian protection module, so that the adult type air bag is exploded and sprung; if the current pedestrian is detected to be a child, the speed of the vehicle exceeds 8.52m/s and the distance is smaller than 10m, the automobile control module sends an instruction to the pedestrian protection module to make the child type air bag point burst and pop open; if the distance between the adult/child and the automobile is more than 10m and the speed is less than 8.52m/s, the pedestrian air bag cannot be shot by point explosion.
Regarding the design of the adult-type airbag and the child-type airbag, the present invention has found that the relationship between the volume of the gas for the deployment of the adult-type airbag or the child-type airbag, the head acceleration, and the head mass is as follows:
wherein P is the gas pressure; v is the gas volume; n is the gas quantity; r is a gas constant; t is the gas temperature; m is the head mass; s is the force bearing area of the head model in the direction of the speed vector, and a is the head model acceleration.
And determining the gas volumes corresponding to the expansion of the adult type air bags and the child type air bags according to the relational expression, the head type mass and the head type acceleration which are approximately equivalent to those of the adult and the child, and further designing/selecting the corresponding air bags to effectively protect the head of the adult and the head of the child. I.e. the equivalent head mass can be determined from the average level of children/adults, the approximately equivalent head acceleration is set according to the application requirements, which can reduce the risk.
The invention adopts a uniform pressure method to simulate the active energy absorber of the automobile, and the method adopts a volume control method to describe the gas generated by the gas generator through mass flow and temperature. In the uniform pressure method model, it is assumed that the ideal gas is filled in the air bag and the air bag is insulated from the outside, so that there are:
wherein P is 2 Is the pressure of the air bag; v (V) 2 Is the volume of the air bag; m is m 2 Is the mass of the gas in the air bag; t (T) 2 Is the temperature of the gas in the air bag; r is an ideal gas state constant; ρ is the gas density; e is gas energy; k is the heat capacity ratio constant;
the rate of change of the gas mass inside the control volume is determined by the mass of gas flowing across the boundary at time t, namely:
wherein m is min For the gas mass flow into the gas-bag, i.e. the gas mass flow m of the gas generator into the gas-bag m12 ;m mout For the mass flow of gas out of the bladder, i.e. the mass flow m out through the vent m23 And the mass flow rate m 'of the leakage' m23 And (3) summing.
According to the above formula, the state parameters such as the volume, the pressure and the like at each moment in the air bag unfolding process can be determined, as shown in fig. 4 and 5, so that the acceleration peak value of the head impact air bag can be approximately controlled by controlling the air bag unfolding volume or the air quantity of the air bag, and the damage to the head is reduced. After determining the state parameters such as the volume, the pressure and the like at each moment in the air bag unfolding process, the same thought can also control the state parameters by controlling the air bag volume according to the safety requirements of other state parameters, so that the damage is reduced.
Impact test analysis: from the simulation results, fig. 6 shows that: in the adult model, as shown in a and b of fig. 6, when the head-shaped impactor impacts the front end structure of the automobile without the active energy absorbing device, the peak acceleration is about 280.72g, the HIC value of the head is as high as 1682.67, and serious damage is caused to the head of the adult; as shown in c and d in fig. 6, when the adult head type impactor impacts the active energy absorbing device of the adult automobile, the peak acceleration is 115.87g, the head HIC value is 820.895, and the peak acceleration and HIC value are respectively reduced by 51.2% and 58.7%; as shown in e and f of fig. 6, when the adult-head type impactor impacts the active energy absorbing device of the child type car, the peak acceleration is 238.54g, the head HIC value is 1181.06, and the peak acceleration and HIC value are reduced by 15.0% and 29.8%, respectively. Compared with the active energy absorber of the children type automobile, the active energy absorber of the adult type automobile can greatly reduce the damage to the head of the adult and can bring better protection to the head of the adult.
From the simulation results, fig. 7 shows that: in the child model, as shown in fig. 7 a and b, when the child head impactor impacts the front end structure of the automobile without the active energy absorber, the peak acceleration is about 196.38g, and the head HIC value is 761.01; as shown in fig. 7 c and d, when the child head impactor impacts the adult-sized vehicle active energy absorbing device, a peak acceleration of 167.22g is generated. However, since the adult type automobile active energy absorber cannot effectively absorb the impact energy of the head type impactor of the child, the acceleration is slowly reduced in the descending process after reaching the peak value, the reduced acceleration value is still higher than the acceleration when impacting the non-active energy absorber, and finally the HIC value of the head is increased to 1320.07, so that the head of the child is more seriously damaged; as shown in e and f of fig. 7, when the child head impactor impacts the child car active energy absorber, the peak acceleration is 91.93g, the head HIC value is 522.76, and the peak acceleration and HIC value are reduced by 53.2% and 31.3%, respectively. Compared with an adult type automobile active energy absorber, the child type automobile active energy absorber can reduce damage to the head of a child and can bring better protection to the head of the child.
In some embodiments, the technical solution of the present invention provides a system based on the above control method, at least including: the system comprises an image acquisition module, a pedestrian classification module, a collision judgment module and a control module.
The image acquisition module is used for acquiring real-time images of roads; the pedestrian classification module is used for inputting the road real-time image into a pedestrian classification model constructed based on a YOLOv5 network for classification detection to obtain a pedestrian classification result, and the pedestrian classification result is used for identifying whether pedestrians exist in the road real-time image and whether the pedestrians are adults or children; the collision judging module is used for acquiring the distance between the pedestrian and the automobile and the current automobile state, and pre-judging whether the collision between the pedestrian and the automobile occurs or not based on the distance and the current automobile state; the control module is used for popping up a corresponding adult type air bag or child type air bag according to the pedestrian classification result if the collision is judged to occur in advance; otherwise, continuing monitoring.
It should be noted that, the foregoing embodiments are stated to describe the technical solution of the present invention in terms of functional modules, and it should be understood that the implementation process of each module may be stated with reference to the foregoing method, where the foregoing functional module is merely a division of logic functions, and there may be other division manners in actual implementation, for example, multiple units or components may be combined or may be integrated into another system, or some features may be omitted, or not implemented. Meanwhile, the integrated units can be realized in a hardware form or a software functional unit form.
In some embodiments, the technical solution of the present invention provides a system based on the above method for manufacturing a buckle pavilion, at least comprising: the system comprises an automobile controller, an automobile active energy absorbing device, an on-board camera and on-board sensing equipment;
the vehicle active energy absorbing device, the vehicle-mounted camera and the vehicle-mounted sensing equipment are all connected with the vehicle controller, and the vehicle controller calls or loads a pedestrian classification model constructed based on a YOLOv5 network;
the vehicle-mounted camera transmits the shot road real-time image to the automobile controller; the automobile controller utilizes the pedestrian classification model to carry out classification detection on the road real-time image to obtain a pedestrian classification result; the vehicle-mounted sensing equipment detects the distance between a pedestrian and an automobile, acquires the current state of the automobile and transmits the current state of the automobile to the automobile controller; the automobile controller predicts whether collision between a pedestrian and an automobile occurs or not based on the distance and the current automobile state, and if the collision is predicted, the automobile active energy absorber ejects a corresponding adult type air bag or child type air bag according to the pedestrian classification result; otherwise, continuing monitoring.
It should be noted that, the embodiments of the present invention are described in terms of hardware devices.
In some embodiments, the present disclosure further provides a computer readable storage medium storing a computer program, the computer program being invoked by a processor to implement:
acquiring a real-time image of a road;
inputting the road real-time image into a pedestrian classification model constructed based on a YOLOv5 network for classification detection to obtain a pedestrian classification result, wherein the pedestrian classification result is used for identifying whether pedestrians exist in the road real-time image and whether the pedestrians are adults or children;
acquiring the distance between a pedestrian and an automobile and the current automobile state, and predicting whether collision between the pedestrian and the automobile occurs or not based on the distance and the current automobile state;
if collision is predicted, ejecting a corresponding adult type air bag or child type air bag by the automobile active energy absorbing device according to the pedestrian classification result; otherwise, continuing monitoring.
For a specific implementation of each step, please refer to the description of the foregoing method.
The readable storage medium is a computer readable storage medium, which may be an internal storage unit of the controller according to any one of the foregoing embodiments, for example, a hard disk or a memory of the controller. The readable storage medium may also be an external storage device of the controller, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card) or the like, which are provided on the controller. Further, the readable storage medium may also include both an internal storage unit and an external storage device of the controller. The readable storage medium is used to store the computer program and other programs and data required by the controller. The readable storage medium may also be used to temporarily store data that has been output or is to be output.
Based on such understanding, the technical solution of the present invention is essentially or a part contributing to the prior art, or all or part of the technical solution may be embodied in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned readable storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, randomAccess Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
It should be emphasized that the examples described herein are illustrative rather than limiting, and that this invention is not limited to the examples described in the specific embodiments, but is capable of other embodiments in accordance with the teachings of the present invention, as long as they do not depart from the spirit and scope of the invention, whether modified or substituted, and still fall within the scope of the invention.
Claims (9)
1. A control method of an automobile active energy absorbing device based on pedestrian classification and identification is characterized by comprising the following steps of: the automobile active energy absorbing device is arranged on an automobile head and is of an adult/child airbag structure, and the method comprises the following steps:
shooting by using a vehicle-mounted camera to obtain a road real-time image;
inputting the road real-time image into a pedestrian classification model constructed based on a YOLOv5 network for classification detection to obtain a pedestrian classification result, wherein the pedestrian classification result is used for identifying whether pedestrians exist in the road real-time image and whether the pedestrians are adults or children;
detecting the distance between a pedestrian and an automobile by using an on-board sensing device, acquiring the current automobile state, and pre-judging whether collision between the pedestrian and the automobile occurs or not based on the distance between the pedestrian and the automobile and the current automobile state;
if collision is predicted, ejecting a corresponding adult type air bag or child type air bag by the automobile active energy absorbing device according to the pedestrian classification result; otherwise, continuing monitoring.
2. The method according to claim 1, characterized in that: the YOLOv5 network comprises an input end, a backbone network back, a neck network back and an output end head which are sequentially connected;
the SPP module in the traditional backbone network backup is replaced by the spatial pyramid pooling SPPCSPC, so that the network robustness is enhanced, and the target recognition anti-interference level of children is improved; the space pyramid pooling SPPCSPC is divided into two transmission paths, and the first transmission path is sequentially provided with three convolution layers CBS, a maximum pooling layer, a feature fusion module and two convolution layers CBS; the second transmission path is provided with a convolution layer CBS, and the output of the last convolution layer CBS on the first transmission path and the output of the last convolution layer CBS on the second transmission path are subjected to feature fusion C and then input into the convolution layer CBS.
3. The method according to claim 1, characterized in that: the YOLOv5 network comprises an input end, a backbone network back, a neck network back and an output end head which are sequentially connected;
the method comprises the steps that a first C3 module on a backbone network backup is led out of another output path, a second C3 module on the backbone network backup is led out of another output path, two paths are fused, the two paths sequentially pass through a Focus layer Focus, a convolution layer CBS and a feature fusion module, and then the second C3 module is connected to a traditional transmission path of a neck network neg so as to improve feature capturing capability of a child target.
4. The method according to claim 1, characterized in that: the current vehicle state at least comprises a current vehicle speed, and the process of predicting whether collision of pedestrians and automobiles occurs based on the distance and the current vehicle state is as follows:
if the pedestrian is an adult, the current vehicle speed exceeds a preset vehicle speed threshold V1, and the distance is smaller than a preset distance threshold L1, the adult and the vehicle are considered to collide;
if the pedestrian is a child and the current vehicle speed exceeds a preset vehicle speed threshold V2 and the distance is smaller than a preset distance threshold L2, the child is considered to collide with the vehicle;
wherein the value ranges of the preset vehicle speed threshold V1 and the preset vehicle speed threshold V2 are 5.6 m/s-11.1 m/s;
the range of the preset distance threshold L1 and the preset distance threshold L2 is 6.7 m-13.3 m.
5. The method according to claim 1, characterized in that: the relation between the volume of the gas expanded by the adult type air bag or the child type air bag and the head type acceleration and the head type mass is as follows:
wherein P is the gas pressure; v is the gas volume; n is the gas quantity; r is a gas constant; t is the gas temperature; m is the head mass; s is the stress area of the head shape in the direction of the speed vector, and a is the head acceleration;
and determining the gas volumes corresponding to the expansion of the adult type air bag and the child type air bag according to the relational expression, the head type mass and the head type acceleration which are approximately equivalent to those of the adult and the child.
6. A control system based on the method of any one of claims 1-5, characterized in that: comprising the following steps:
the image acquisition module is used for acquiring real-time images of roads;
the pedestrian classification module is used for inputting the road real-time image into a pedestrian classification model constructed based on a YOLOv5 network for classification detection to obtain a pedestrian classification result, and the pedestrian classification result is used for identifying whether pedestrians exist in the road real-time image and whether the pedestrians are adults or children;
the collision judging module is used for acquiring the distance between the pedestrian and the automobile and the current automobile state, and predicting whether the collision between the pedestrian and the automobile occurs or not based on the distance between the pedestrian and the automobile and the current automobile state;
the control module is used for popping up a corresponding adult type air bag or child type air bag according to the pedestrian classification result if the collision is judged to occur in advance; otherwise, continuing monitoring.
7. An automotive active energy absorber based on the method of any one of claims 1-5, characterized in that: the automobile active energy absorbing device comprises a child type air bag and a adult type air bag, or the automobile active energy absorbing device comprises an air bag and air bag inflating equipment, and the air bag inflating equipment controls the air volume in the air bag according to the child type and the adult type to obtain the child type air bag or the adult type air bag.
8. An in-vehicle system based on the method of any one of claims 1-5, characterized in that: at least comprises: the system comprises an automobile controller, an automobile active energy absorbing device, an on-board camera and on-board sensing equipment;
the vehicle active energy absorbing device, the vehicle-mounted camera and the vehicle-mounted sensing equipment are all connected with the vehicle controller, and the vehicle controller calls or loads a pedestrian classification model constructed based on a YOLOv5 network;
the vehicle-mounted camera transmits the shot road real-time image to the automobile controller;
the automobile controller utilizes the pedestrian classification model to carry out classification detection on the road real-time image to obtain a pedestrian classification result;
the vehicle-mounted sensing equipment detects the distance between a pedestrian and an automobile, acquires the current state of the automobile and transmits the current state of the automobile to the automobile controller;
the automobile controller pre-judges whether collision of the pedestrian and the automobile occurs or not based on the distance between the pedestrian and the automobile and the current automobile state, if the collision is pre-judged, the automobile active energy absorption device pops up a corresponding adult type air bag or child type air bag according to the pedestrian classification result; otherwise, continuing monitoring.
9. A computer-readable storage medium, characterized by: a computer program is stored, which is called by a processor to implement:
acquiring a real-time image of a road;
inputting the road real-time image into a pedestrian classification model constructed based on a YOLOv5 network for classification detection to obtain a pedestrian classification result, wherein the pedestrian classification result is used for identifying whether pedestrians exist in the road real-time image and whether the pedestrians are adults or children;
acquiring the distance between a pedestrian and an automobile and the current automobile state, and predicting whether collision between the pedestrian and the automobile occurs or not based on the distance and the current automobile state;
if collision is predicted, ejecting a corresponding adult type air bag or child type air bag by the automobile active energy absorbing device according to the pedestrian classification result; otherwise, continuing monitoring.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311106575.7A CN117125020A (en) | 2023-08-30 | 2023-08-30 | Pedestrian classification and identification-based active energy absorption device for automobile and control method thereof |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311106575.7A CN117125020A (en) | 2023-08-30 | 2023-08-30 | Pedestrian classification and identification-based active energy absorption device for automobile and control method thereof |
Publications (1)
Publication Number | Publication Date |
---|---|
CN117125020A true CN117125020A (en) | 2023-11-28 |
Family
ID=88862582
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311106575.7A Pending CN117125020A (en) | 2023-08-30 | 2023-08-30 | Pedestrian classification and identification-based active energy absorption device for automobile and control method thereof |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117125020A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113784877A (en) * | 2019-03-05 | 2021-12-10 | 塞瓦技术公司 | Communication interface for an external inflatable pedestrian safety structure equipped on a vehicle, associated inflatable structure-vehicle communication protocol and safety module |
-
2023
- 2023-08-30 CN CN202311106575.7A patent/CN117125020A/en active Pending
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113784877A (en) * | 2019-03-05 | 2021-12-10 | 塞瓦技术公司 | Communication interface for an external inflatable pedestrian safety structure equipped on a vehicle, associated inflatable structure-vehicle communication protocol and safety module |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US10007854B2 (en) | Computer vision based driver assistance devices, systems, methods and associated computer executable code | |
CN111507174B (en) | Method and device for predicting height and weight of passenger | |
CN104828005B (en) | collision detecting device | |
CN206561825U (en) | A kind of pedestrian protection system | |
CN117125020A (en) | Pedestrian classification and identification-based active energy absorption device for automobile and control method thereof | |
US20090268947A1 (en) | Real time environment model generation system | |
US9430708B2 (en) | Method of detecting pedestrians using integral image of feature information, computer readable recording medium, and terminal device for the same | |
CN104709215B (en) | The method of security system and the security system for operating vehicle | |
EP1709568A2 (en) | Method and apparatus for object tracking prior to imminent collision detection | |
Kashevnik et al. | Seat belt fastness detection based on image analysis from vehicle in-abin camera | |
Op den Camp et al. | Cyclist target and test setup for evaluation of cyclist-autonomous emergency braking | |
CN109733317A (en) | A kind of automobile multistage intelligent anti-collision system and control method | |
KR20180127245A (en) | Method for Predicting Vehicle Collision Using Data Collected from Video Games | |
Östling et al. | Predicting crash configurations in passenger car to passenger car crashes to guide the development of future passenger car safety | |
JP2020514918A (en) | Method and apparatus for minimizing damage in an accident situation in a motor vehicle | |
Vamsi et al. | In-vehicle occupancy detection and classification using machine learning | |
JP2023522131A (en) | Safety systems, vehicle safety systems and devices, methods, and media for improving road compatibility of vehicles | |
CN115279643A (en) | On-board active learning method and apparatus for training a perception network of an autonomous vehicle | |
KR101050687B1 (en) | Pedestrian recognition device and method using posterior probability and pedestrian protection device using it | |
TWM626569U (en) | car accident severity prediction device | |
Kuroe et al. | Exploratory study of an airbag concept for a large touring motorcycle: Further research second report | |
CN102043941B (en) | Dynamic real-time relative relationship identification method and system | |
CN112052829A (en) | Pilot behavior monitoring method based on deep learning | |
Kang et al. | High performance and fast object detection in road environments | |
CN116691591A (en) | Airbag device and airbag system |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |