Highly integrated visual navigation small pod

Abstract

The utility model discloses a highly integrated visual navigation small pod, which comprises a spherical pod and a connecting base, a lens, a core processing board card and a servo board card are arranged in the spherical pod, and the connecting base comprises a three-axis micro gyroscope, a magnetic encoder and a micro connector. The device is suitable for being used in various scenes, a user can conveniently install the device on an unmanned aerial vehicle, a ground vehicle or a fixed platform, the applicability of the device is improved, the high-performance computing power of the RK3588 core processor is adopted, real-time image processing and data analysis are supported, and the application range is wide. The position information of the target can be quickly transmitted to the upper computer through the wireless network, so that an operator can monitor the dynamic state of the target in real time and quickly make a decision, and the timeliness of combat and monitoring is enhanced.

Description

TECHNICAL FIELD

[0001] The utility model relates to target tracking investigation equipment technical field, especially in a kind of highly integrated visual navigation small pod. BACKGROUND

[0002] There is a demand for rapid positioning and continuous tracking of specific targets in military reconnaissance and related fields, which requires accurate discovery of targets and close tracking of their movements in complex environments, while also adapting to different terrain conditions. In actual combat or reconnaissance environment, the information obtained by sensors is complex and diverse, and the most critical for target tracking and positioning is the feature information of the target and the terrain data. Especially when GPS fails, how to quickly locate the target area becomes a problem to be solved.

[0003] Therefore, it is particularly important to develop a device that can still have efficient target positioning and tracking capability under GPS failure condition;

[0004] In addition, the existing device also has certain limitations in integration, weight and processing capacity, which limits its application in variable environment. INVENTION CONTENTS

[0005] In order to overcome the shortcomings of the prior art, the utility model is realized by the following technical solutions:

[0006] A highly integrated visual navigation small pod, comprising a spherical pod and a connecting base, the spherical pod is built-in lens, core processing board and servo board, the connecting base comprises three-axis micro gyroscope, magnetic encoder and J30J micro connector.

[0007] Further, the lens is 500 million high-definition 550nm narrowband M12 lens, the focal length is 20mm, supports 500 million pixel image or video capture, and optimizes the processing of 550nm wavelength light.

[0008] Further, the core processing board is a high-performance, low-power processor chip RK3588, which supports multiple AI frameworks and has neural network computing capability, suitable for intelligent visual computing and high-performance video encoding and decoding.

[0009] Further, the J30J micro connector is J30J-30 core, with excellent three-proof performance, suitable for high-density, miniaturized electrical circuit system interconnection.

[0010] Further, the RK3588 processor is built-in YOLOv8 as target detection algorithm, DSST++ as target tracking algorithm and template feature matching algorithm as terrain matching algorithm, which can realize target detection, tracking and terrain matching in complex environment.

[0011] Further, the YOLOv8 detection algorithm can detect 80 targets, the DSST++ algorithm optimizes the anti-occlusion ability, and the template feature matching algorithm integrates template matching and feature matching algorithms and can accurately find the area most similar to the template in the image.

[0012] The device takes RK3588 as a core processor, is built-in 20mm lens, a servo board card, three-axis micro gyroscope, magnetic encoder and J30J micro connector; the RK3588 processor is built-in yoloV8 as a detection algorithm, DSST++ as a tracking algorithm and a template feature matching algorithm as a terrain matching algorithm. YOLOv8 can detect 80 targets, DSST++ is optimized on the basis of DSST and can resist occlusion for a short time. The template feature matching algorithm integrates template matching and feature matching algorithms and can accurately find the area most similar to the template in the image.

[0013] Specific working process: first, the upper computer (or unmanned aerial vehicle) powers the small pod through the J30J micro connector, and the spherical pod obtains image data through the lens after starting. Then the image data is transmitted to the RK3588, and the RK388 calculates, analyzes and processes the position of the target through the built-in detection algorithm, tracking algorithm and terrain matching algorithm. Then, the RK3588 transmits the information to the servo board card, and the servo board card adjusts the three-axis micro gyroscope and magnetic encoder in the base to correct the attitude of the spherical pod so as to find the target and image in the center of the image. Finally, the calculation result is transmitted to the upper computer (or unmanned aerial vehicle) through the J30J micro connector.

[0014] In summary, the utility model has the following beneficial effects:

[0015] 1、The spherical design and lightweight structure (weight does not exceed 300g) of the highly integrated visual navigation small pod make the small pod convenient to carry and install, and it is suitable for use in various scenes, and users can conveniently install it on an unmanned aerial vehicle, ground vehicle or fixed platform, so that the applicability of the equipment is improved. The equipment can be powered by the upper computer or unmanned aerial vehicle, and automatically enters the working state after starting, so that the operation process is simplified, and the visual navigation small pod can be quickly deployed and adapt to rapidly changing task requirements, and is particularly suitable for military reconnaissance and emergency response scenes.

[0016] 2, high-performance computing capability of RK3588 core processor is adopted, real-time image processing and data analysis are supported, target position information can be rapidly transmitted to the upper computer through a wireless network, this characteristic enables the operator to monitor the target dynamic in real time, makes a decision quickly, and enhances the timeliness of combat and monitoring, the YOLOv8 target detection algorithm is adopted, up to 80 targets can be detected in real time, and the DSST++ tracking algorithm is combined, the speed and accuracy of target identification and tracking are improved significantly, so that the equipment can respond quickly in a complex environment, ensures the continuous tracking of the target in a dynamic scene, the 5 million high-definition M12 lens is optimized, is particularly suitable for processing 550nm wavelength light, effectively reduces the interference of other wavelength light, improves the precision and clarity of image capture, this design enables the equipment to maintain good working performance in strong light, shadow or other visual interference environment.

[0017] 3, the visual navigation small pod is not only suitable for military reconnaissance, but also can be widely applied to security monitoring, unmanned driving, unmanned aerial vehicle inspection and other fields, the advanced technical scheme and flexible application capability can meet the needs of various industries, and has good market prospect. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 is the overall three-dimensional structure schematic diagram of the utility model;

[0019] Figure 2 is the three-dimensional structure schematic diagram of another view of the utility model;

[0020] Figure 3 is the top view of the utility model;

[0021] Figure 4 is the internal circuit connection schematic diagram of the utility model.

[0022] In the figure, 1 is a spherical pod; 2 is a connecting base; 3 is a lens. DETAILED DESCRIPTION

[0023] The utility model will be further described in detail below in combination with the drawings.

[0024] Same parts are denoted by the same reference numerals. It should be noted that the words "front", "back", "left", "right", "up" and "down" used in the following description refer to the directions in the drawings, and the words "bottom surface" and "top surface", "inner" and "outer" refer to the directions towards or away from the geometric center of a particular component.

[0025] Referring to Figures 1-4As shown, it is a highly integrated visual navigation small pod in a preferred embodiment of the utility model, including spherical pod 1 and connecting base 2, the built-in lens 3, core processing board card and servo board card in the spherical pod, the connecting base includes three-axis micro gyroscope, magnetic encoder, J30J micro connector composition;The lens is 5 million high-definition 550nm narrowband M12 lens, the focal length is 20mm, supports 5 million pixel image or video capture, and optimizes the processing of 550nm wavelength light;The core processing board card is a high-performance, low-power processor chip RK3588, which supports multiple AI frameworks, has neural network computing capability, and is suitable for intelligent visual computing and high-performance video encoding and decoding;The J30J micro connector is J30J-30 core, has excellent three-proof performance, and is suitable for high-density, miniaturized electrical circuit system interconnection;YOLOv8 as target detection algorithm, DSST++ as target tracking algorithm and template feature matching algorithm as terrain matching algorithm are built in RK3588 processor, which can realize target detection, tracking and terrain matching in complex environment;The YOLOv8 detection algorithm can detect 80 targets, the DSST++ algorithm optimizes the anti-occlusion ability, the template feature matching algorithm integrates template matching and feature matching algorithms, and can accurately find the area most similar to the template in the image.

[0026] Embodiment 1

[0027] Detailed description of the unmanned aerial vehicle visual navigation small pod

[0028] I. Specific structure composition

[0029] Size: the overall height of the device is 120mm, the forward width is 65mm, the front and back width is 80mm, the weight is 300g, the diameter of the spherical pod is 65mm, the diameter of the lens is 15mm, and the focal length is 20mm.

[0030] Lens module: 5 million high-definition M12 lens, supports 550nm wavelength image capture, has optical image stabilization function, focal length 20mm, suitable for close-range target recognition.

[0031] Processor: RK3588 core processor, with powerful image processing capability, supports multitasking, built-in high-performance NPU, suitable for AI algorithm execution.

[0032] Power supply module: input voltage: DC 5V to 12V, supports unmanned aerial vehicle power supply system.

[0033] Communication module: integrated Wi-Fi and 4G module, used for data transmission and remote control.

[0034] Sensors: Three-axis micro-gyroscope (Inertial Measurement Unit) in the base for real-time monitoring of attitude and motion state; magnetic encoder module for position information processing.

[0035] II. Data Processing Process

[0036] The lens module captures image data in real time, which is transmitted to the RK3588 processor through a data bus (such as MIPI interface); the three-axis micro-gyroscope and magnetic encoder module are connected to the RK3588 through I2C or UART interface, providing real-time attitude and position information.

[0037] III. Algorithm Processing Process

[0038] Image preprocessing: The captured image is first preprocessed, including denoising, enhancement and scaling, to adapt to the subsequent target detection algorithm.

[0039] Target detection: RK3588 processor runs YOLOv8 algorithm:

[0040] Input: preprocessed image.

[0041] Output: detected target bounding box and its class (such as vehicle, pedestrian, etc.).

[0042] Processing process: YOLOv8 extracts features from images through convolutional neural network (CNN), generates feature maps, and outputs target location information and class probability through fully connected layer.

[0043] Target tracking: once the target is detected, the system automatically switches to DSST++ tracking algorithm:

[0044] Input: initial position of target (bounding box) and image of subsequent frame.

[0045] Output: continuous position and motion trajectory of target.

[0046] Processing process: DSST++ algorithm updates the position of the target in real time through modeling and matching of target features, and starts the anti-occlusion function when the target is short-term lost, to find the possible target position.

[0047] Data fusion: combine target detection results with IMU and GPS data for position correction:

[0048] Input: target position, IMU data (attitude information), GPS data (geographic location).

[0049] Output: accurate target position and attitude information.

[0050] Processing process: through Kalman filtering algorithm, the data of each sensor is fused to improve the accuracy of target positioning.

[0051] Result transmission: The processed target location information is transmitted in real time to the operator's host computer via Wi-Fi or 4G module for monitoring and decision-making.

[0052] IV. Circuit Connection Method

[0053] Power connection: The drone power supply provides power to the device through the J30J micro connector. The power module has a voltage regulator circuit inside to ensure stable operation of the processor and other components.

[0054] Data connection: The lens module uses a MIPI interface to connect the lens module's data output to the RK3588 processor.

[0055] The three-axis micro gyroscope module connects to the RK3588 via an I2C interface, providing real-time acceleration and angular velocity data.

[0056] The servo board module is connected to the RK3588 via an interface, providing control signals for lens rotation and movement.

[0057] The magnetic encoder module is connected to the RK3588 via a UART interface to provide position information.

[0058] Communication connectivity: The RK3588's Wi-Fi and 4G modules connect via an SPI interface, ensuring real-time data transmission to the operator's host computer. The communication module's output port connects to the drone's control system, enabling it to receive commands from the host computer.

[0059] V. Work Process

[0060] 1. Detection and tracking workflow

[0061] Real-time images are acquired through the camera. The YOLOv8 detection algorithm is activated to detect targets in real time. When a target is detected, the system automatically switches to tracking mode to track the target. If the target is briefly lost, the tracking algorithm automatically activates anti-occlusion functionality to search for the target near the lost area. If the target is lost for an extended period, the tracking algorithm stops, the detection algorithm is activated to search for the target, and the above functions are repeated. When the target is being tracked normally, the information is transmitted to the host computer via the J30J micro connector.

[0062] 2. Terrain and Landform Matching Process

[0063] The terrain template is transmitted to the visual navigation pod by the host computer, and the pod obtains real-time images through the lens. After obtaining the template and real-time images, the template feature matching algorithm is used to continuously search for the terrain image in the real-time image. When the template contains the terrain, the current real-time image is transmitted to the host computer through the J30J micro connector (the image is transmitted to the drone, which can be used for target area positioning when GPS fails).

[0064] Embodiment 2

[0065] Application of visual navigation pod in security monitoring field

[0066] I. Specific structure composition

[0067] Built-in 500 million high-definition 550nm narrowband M12 lens, focal length 20mm; Core processing board RK3588, with high-performance processing capability, supporting YOLOv8 and DSST++ algorithms; Three-axis micro gyroscope and magnetic encoder for attitude control; J30J micro connector for power supply and data transmission.

[0068] Workflow

[0069] System startup:

[0070] Install the pod on the security monitoring tower and connect it to the monitoring system for power supply through the J30J micro connector. After starting, the spherical pod immediately begins to obtain real-time images of the surrounding environment.

[0071] Real-time monitoring and target detection:

[0072] The RK3588 processor runs the YOLOv8 algorithm to monitor the monitoring area in real time and detect suspicious activities (such as intruders, abnormal gatherings, etc.); After detecting a suspicious target, it automatically switches to tracking mode.

[0073] Target tracking and data analysis:

[0074] The DSST++ tracking algorithm continuously tracks the suspicious target to ensure the accuracy of the monitoring.

[0075] If the target is lost for a short time, the system will start the anti-occlusion function to try to find the target near the lost area.

[0076] Alarm and information transmission:

[0077] Once suspicious activity is detected, the system will automatically trigger an alarm and transmit alarm information and real-time image data to the security personnel's terminal through the J30J micro connector.

[0078] Security personnel can view the monitoring screen in real time, judge the situation and take necessary measures.

[0079] Terrain matching and positioning:

[0080] The terrain topography template is transmitted to the small pod by the host computer to perform terrain matching.

[0081] The small pod acquires the current real-time image through a lens, uses a template feature matching algorithm to find terrain features similar to the template, and helps to locate suspicious areas.

[0082] The basic principle and main features of the utility model and the advantages of the utility model are shown and described. The skilled in the art should understand that the utility model is not limited by the above examples, and the above examples and descriptions in the specification are only to illustrate the principle of the utility model, and various changes and improvements of the utility model can be made without departing from the spirit and scope of the utility model. These changes and improvements all fall within the scope of the claimed utility model. The scope of protection of the utility model is defined by the appended claims and their equivalents.

Claims

1. A highly integrated visual navigation gondola, characterized in that, It comprises a spherical pod and a connecting base, the spherical pod is internally provided with a lens, a core processing board and a servo board, and the connecting base comprises a three-axis micro gyroscope, a magnetic encoder and a micro connector.

2. The highly integrated visual navigation pod of claim 1, wherein, The lens is a 500 million high-definition 550nm narrowband M12 lens, the focal length is 20mm, supports 500 million pixel image or video capture, and optimizes the processing of 550nm wavelength light.

3. The highly integrated visual navigation pod of claim 1, wherein, The core processing board is a high-performance and low-power processor chip RK3588, which supports multiple AI frameworks and has neural network computing capability, and is suitable for intelligent visual computing and high-performance video encoding and decoding.

4. The highly integrated visual navigation pod of claim 1, wherein, The micro connector is a J30J-30 core, which has excellent three-proof performance and is suitable for high-density and miniaturized electrical circuit system interconnection.

5. The highly integrated visual navigation pod of claim 3, wherein, The RK3588 processor internally provides YOLOv8 as a target detection algorithm, DSST++ as a target tracking algorithm and a template feature matching algorithm as a terrain matching algorithm, which can realize target detection, tracking and terrain matching in complex environment.

6. The highly integrated visual navigation pod of claim 5, wherein, The YOLOv8 detection algorithm can detect 80 targets, the DSST++ algorithm optimizes the anti-occlusion capability, and the template feature matching algorithm combines template matching and feature matching algorithms, which can accurately find the most similar area to the template in the image.

Images