CN114863377A - Improved target detection system based on YOLOv5 - Google Patents
Improved target detection system based on YOLOv5 Download PDFInfo
- Publication number
- CN114863377A CN114863377A CN202210464185.6A CN202210464185A CN114863377A CN 114863377 A CN114863377 A CN 114863377A CN 202210464185 A CN202210464185 A CN 202210464185A CN 114863377 A CN114863377 A CN 114863377A
- Authority
- CN
- China
- Prior art keywords
- target detection
- yolov5
- circuit board
- detection system
- camera
- 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
Images
Classifications
-
- 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
-
- 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
- G06N3/045—Combinations of networks
-
- 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
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/10—Image acquisition
- G06V10/12—Details of acquisition arrangements; Constructional details thereof
-
- 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
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Image Processing (AREA)
Abstract
The invention relates to a YOLOv5 network-based improved target detection system, which mainly comprises an electronic display screen, a camera, a circuit board and a data storage processing device, wherein the electronic display screen is used for displaying video information processed by a target detection network, the camera is used for acquiring video information around an automobile so as to acquire image information, the data storage processing device is used for storing the video information, a network structure improved by YOLOv5, processing the video information and the like, and the circuit board is used for connecting all hardware to form a complete system. The software part of the invention takes a YOLOv5 network structure as a main algorithm, and improves a backhaul part, a Neck part and an NMS part of the software part respectively, so that a target detection system with high detection precision, strong real-time performance, high robustness and good generalization capability is obtained. Provides a more excellent system for users using automatic driving technology.
Description
Technical Field
The invention relates to the field of automatic driving of automobiles, and particularly provides an improved target detection system based on YOLOv 5.
Background
In the automobile automatic driving technology, the identification and the discrimination of a target object in a complex environment are a difficult challenge and are also one of key tasks to be solved. Target detection is not only an important branch in computer vision research, but also a critical link and task in an automatic driving system. Due to the complexity of the actual road conditions, the technical performance of the auxiliary automobile driving based on the traditional target detection is difficult to be greatly improved
Therefore, an improved target detection system based on YOLOv5 is designed, the device belongs to the application product of automobile automatic driving, and in the automatic driving technology, the identification and the discrimination of a target object in a complex environment are a difficult challenge and are one of the key tasks to be solved. The method comprises the steps of road and road edge recognition, lane line detection, vehicle recognition, vehicle type recognition, non-motor vehicle recognition, pedestrian recognition, traffic sign recognition, obstacle recognition and avoidance and the like. The system for identifying the target in the automobile observes the traffic environment around the automobile through computer vision, automatically identifies the target from a real-time video signal, and provides a judgment basis for automatic driving, such as starting, stopping, steering, accelerating, decelerating and the like. The system covers the detection and identification of various targets such as the detection and identification of roads and road edges around automobiles, lane line detection, vehicle type identification, non-motor vehicle identification, pedestrian identification, traffic sign identification, obstacle identification and the like, and the detection and identification results are displayed in real time through a display. The system has the advantages of high detection precision and strong real-time performance, promotes the development of the automobile automatic driving technology, and provides better user experience for people under the condition of using the automobile automatic driving function.
Disclosure of Invention
The method solves the problems of poor detection speed, insufficient detection precision of the shielded target, high missing detection rate of the small-scale target, insufficient detection precision in the complex environment and the like in the target detection in the automatic driving scene. An improved target detection system based on YOLOv5 is provided.
First, build a Yolov5 network
And secondly, a Coordinate Attention mechanism (Coordinate Attention) is added in a backbone network of YOLOv5, the Coordinate Attention mechanism embeds position information into channel Attention, the receptive field is increased, and the positioning capability of the model for the target is improved.
And thirdly, adding a small target detection layer in the detection layer of the YOLOv5 network for increasing the detection capability of the small-scale target.
And fourthly, replacing NMS with DIoU-NMS to improve the identification degree of the shielding target.
And fifthly, embedding the improved Yolov5 target detection algorithm into a chip, and assembling a camera and display hardware to enable the improved Yolov5 target detection algorithm to be applied to an automatic driving scene.
The system mainly comprises a machine body, a data processing device, a camera and an electronic display screen which are assembled.
The body is connected with a camera and used for acquiring image information outside the automobile.
An electronic display screen is embedded into the engine body and used for displaying image information acquired by the camera on the electronic display screen after processing, so that a driver can clearly and visually know information outside the vehicle.
And a data processing device is arranged in the machine body and used for processing the image information acquired by the camera and outputting the information to the electronic display screen.
The method has the advantages of high target detection precision, strong real-time property, high stability and the like, can be combined with other automatic driving technologies, is easy to expand, and can be suitable for most automatic driving scenes.
Drawings
FIG. 1 is a schematic structural diagram of a target detection system according to the present invention
FIG. 2 is a diagram of the improved YOLOv5 network structure according to the present invention
FIG. 3 is a flow chart of the system of the present invention
Description of reference numerals: 1. a body; 2. a circuit board; 3. a data storage processing device; 4. a camera; 5. an electronic display screen; 6. in order to obtain an improved structure diagram of the YOLOv5 network, 0-34 in the diagram are each layer network serial numbers.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all embodiments, and all other embodiments obtained by a person of ordinary skill in the art without creative efforts based on the embodiments of the present invention belong to the protection scope of the present invention.
Referring to fig. 1, the structural schematic diagram of the present invention is shown in fig. 1, which is a structural schematic diagram of a target detection system, wherein the target detection system includes a machine body 1, and a circuit board 2 is arranged inside the machine body 1 and used for connecting various hardware; the circuit board 2 is provided with a data storage processing device 3 which is used for processing, storing and uploading received data; the circuit board 2 is connected with a plurality of external cameras 4, and the cameras 3 are used for acquiring image information; the circuit board 2 is connected with an electronic display screen 5 on the machine body 1, and the electronic display screen 5 is used for displaying image information processed by a YOLOv5 target detection network.
Fig. 2 is a diagram of the improved YOLOv5 network architecture of the present invention.
The invention has the improvement part that a coordinate attention mechanism is added at the 5 th, 8 th and 11 th layer networks for improving the characteristic perception capability of a target detection system for a detected target, increasing the receptive field and improving the precision of the whole target detection.
The second improvement part of the invention, 31, 32 and 34, adds an up-sampling layer, a feature fusion layer and a detection layer compared with the original YOLOv5 network, and the whole of the three improvement points is equivalent to the small target detection layer mentioned above, and is used for improving the detection capability of the system for the small target.
The improvement part III of the invention is to improve a Detect layer, namely a detection layer, of the whole system, wherein the detection layer comprises a node called NMS (non-maximum suppression), and the DIoU-NMS is used for replacing the NMS part in the original system network, so that the operation can improve the detection capability of the whole target detection system on overlapped targets and shielded targets.
Referring to fig. 3, the flow chart of the object detection according to the present invention includes the following steps:
(1) capturing video from a camera
(2) Dividing the acquired video stream into one-frame and one-frame pictures
(3) Preprocessing each frame picture
(4) Sending the preprocessed picture into a target detection network
(5) Obtaining the picture processed by the target detection network
(6) Storage device for feeding picture-synthesized video stream into data storage processing device
(7) Sending the picture synthesized video stream to an electronic display screen for display
The speed of picture processing of this system is 73 sheets per second. The common video stream contains 24 frames of images, namely 24 pictures, every second, so that the system can completely meet the requirements of video stream image processing. The requirement for real-time performance in automatic driving technology can be met and exceeded.
The system also has the advantage of high detection precision, can achieve higher target detection precision, only few targets in the video can have the condition of false detection and omission, the automatic driving function requirement can be met to a great extent, and accidents caused by false detection and omission when the user uses the automatic driving function are avoided to a certain extent. The risk that the user suffers property loss and personal injury is reduced.
In conclusion, the method has the advantages of high detection precision, strong real-time performance, high robustness and good generalization capability. Has good promotion effect on the development of automatic driving technology
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art should be considered to be within the technical scope of the present invention, and the technical solutions and the inventive concepts thereof according to the present invention should be equivalent or changed within the scope of the present invention.
Claims (5)
1. Improved target detection system based on YOLOv5, comprising a body (1), and is characterized in that: the automobile image display device is characterized in that an electronic display screen (5) is arranged on the body (1) and used for displaying image information, a circuit board (2) is embedded into the body (1) and used for connecting hardware such as the electronic display screen (5), a camera (4) and a data storage processing device (3), the circuit board (2) is provided with the data storage processing device (3) and used for storing and processing data, and the circuit board (2) is connected with the camera (4) and used for acquiring the image information around an automobile.
2. The YOLOv 5-based improved target detection system according to claim 1, wherein: the machine body (1) is connected with an electronic display screen (5) and is used for displaying the processed video image information.
3. The YOLOv 5-based improved target detection system according to claim 1, wherein: the machine body (1) is embedded with a circuit board (2) for connecting all hardware into a whole.
4. The YOLOv 5-based improved target detection system according to claim 1, wherein: the circuit board (2) is provided with a data storage processing device (3) which is used for processing, storing and outputting the acquired data of the camera and storing the improved Yolov5 target detection network software information.
5. The YOLOv 5-based improved target detection system according to claim 1, wherein: the camera (5) is connected with the circuit board (2) and used for acquiring video image information around the automobile.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210464185.6A CN114863377A (en) | 2022-04-29 | 2022-04-29 | Improved target detection system based on YOLOv5 |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210464185.6A CN114863377A (en) | 2022-04-29 | 2022-04-29 | Improved target detection system based on YOLOv5 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114863377A true CN114863377A (en) | 2022-08-05 |
Family
ID=82634733
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210464185.6A Pending CN114863377A (en) | 2022-04-29 | 2022-04-29 | Improved target detection system based on YOLOv5 |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114863377A (en) |
-
2022
- 2022-04-29 CN CN202210464185.6A patent/CN114863377A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR101647370B1 (en) | road traffic information management system for g using camera and radar | |
| US11833966B2 (en) | Switchable display during parking maneuvers | |
| CN107273788B (en) | Imaging system for performing lane detection in a vehicle and vehicle imaging system | |
| US11425337B2 (en) | Camera system, event recording system and event recording method | |
| CN107133559B (en) | Mobile object detection method based on 360 degree of panoramas | |
| CN112163543A (en) | Method and system for detecting illegal lane occupation of vehicle | |
| CN110827197A (en) | Method and device for detecting and identifying vehicle all-round looking target based on deep learning | |
| JP3381351B2 (en) | Ambient situation display device for vehicles | |
| CN103646550A (en) | Intelligent vehicle license plate recognition system | |
| US11380111B2 (en) | Image colorization for vehicular camera images | |
| KR101985711B1 (en) | Augmented reality CCTV system, and control method for the same | |
| KR101561626B1 (en) | The Vehicle Black Box Capable of Real-Time Recognizing a License Number Plate for Moving Vehicle | |
| US11659154B1 (en) | Virtual horizontal stereo camera | |
| CN112990162B (en) | Target detection method and device, terminal equipment and storage medium | |
| JP2014089491A (en) | Picture processing system | |
| CN112937445B (en) | 360-degree vehicle safety auxiliary method and vehicle-mounted system | |
| CN111010545A (en) | Vehicle driving decision method, system, terminal and storage medium | |
| US20120189161A1 (en) | Visual attention apparatus and control method based on mind awareness and display apparatus using the visual attention apparatus | |
| CN111614931B (en) | Vehicle surrounding image synthesis method and system | |
| CN106864372A (en) | Outdoor scene internet is called a taxi accessory system and method | |
| CN114863377A (en) | Improved target detection system based on YOLOv5 | |
| CN214959899U (en) | Vehicle-mounted panoramic splicing system suitable for large vehicle | |
| CN216331763U (en) | Intelligent automobile electronic rearview mirror equipment of integrated panorama function and BSD function | |
| WO2022145054A1 (en) | Image processing device, image processing method, and recording medium | |
| JP2012222664A (en) | On-vehicle camera 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 |