Littro MicroBox——Ultra-small terminal deep learning recognition module
[Copy link]
Introduction
Littro MicroBox supports on-device artificial intelligence recognition, and can push the recognized video stream and display the results in real time. MicroBox supports multiple online object recognition algorithms, including pedestrians, vehicles, and special designated object types, and supports video inputs of multiple resolutions and frame rates in multiple bands of infrared and visible light, and multiple types of inputs on the interface to meet the needs of different users. The product also supports auxiliary functions such as video encoding and decoding, video compression, and online storage.
As a general artificial intelligence platform on the terminal, the product is used in security, drone and robot integration, industrial inspection and control, unmanned driving and other industries.
download attach
save to album
2021-4-6 08:12 上传
2. Product interface
Model: MOLEX 5031542090
Interface definition:
BI_4-
|
BI_4+
|
BI_3-
|
BI_3+
|
RX_D2 -
|
RX_D2+
|
TX_D1-
|
TX_D1+
|
GND
|
GND
|
RX-
|
RX+
|
TX-
|
TX+
|
green
|
Green and White
|
orange
|
Orange White
|
1
|
3
|
5
|
7
|
9
|
11
|
13
|
15
|
17
|
19
|
2
|
4
|
6
|
8
|
10
|
12
|
14
|
16
|
18
|
20
|
12v
|
12v
|
GND
|
GND
|
RS485B
|
RS485A
|
GND
|
UART0_TX
|
UART0_RX
|
PGND
|
Wiring and accessories
Matching connector cable.
Extension interface:
Littro MicroBox has multiple external expansion interfaces, including video input and data output interfaces, which facilitates secondary development and product integration. The movement adopts wide voltage power supply, low power consumption and small size, and has a wide operating temperature to meet the needs of common industrial environments.
Power requirement: +7V~+19V
Working temperature: -40~+85℃
2.1 Video input interface
Littro MicroBox supports common USB cameras as video acquisition devices, and also supports cameras with Ethernet interfaces. The specific interface information is as follows:
Video input interface
|
Interface Type
|
parameter
|
Analog infrared thermal imager
|
PAL
|
720*576@25Hz
|
Visible light camera
|
MIPI
|
1080P
|
2.2 Information output interface
Littro MicroBox supports high-definition HDMI video output. The specific interface information is as follows:
Output Interface
|
Connector Type
|
parameter
|
|
|
network
|
MOLEX 5031542090
|
100Mbps supports RTSP/Onvif protocol TCP/IP, UDP protocol
|
|
|
|
|
RS485
|
Private protocol 9600/115200 baud rate
|
|
|
TF card interface
|
Module internal installation
|
Support high-speed TF memory card
|
|
|
3. AI software
Littro MicroBox supports video inputs of multiple resolutions and frame rates in multiple bands of infrared and visible light, and the interface supports multiple types of inputs to meet the needs of different users. The product integrates common functions of video encoding and decoding, compression and online storage, and can accurately identify specific targets (people, cars, gas, aircraft, ships, etc.) based on input image data and deep learning models. It can be used as a general artificial intelligence platform on the terminal for security, drone and robot integration, industrial detection and control, unmanned driving and other industries. It can also support data training and model establishment of user-defined target types.
Littro MicroBox can realize the real-time operation of multiple deep learning algorithms and supports multiple neural networks. Based on the unique infrared enhancement module, it uses infrared images as the video source for target recognition. Compared with traditional infrared recognition algorithms, it has a breakthrough improvement and can achieve high-accuracy detection and recognition in scenarios where the target is obstructed, or the shape and scale change.
Deep Learning Algorithms
|
Function
|
Serve
|
Infrared thermal imaging recognition
|
Based on deep learning methods, the accuracy rate of identifying common targets such as people and cars is higher than 90%.
|
Customizable
|
Visible light recognition
|
Recognize multiple objects based on deep learning methods
|
Customizable
|
Dual-band synchronous recognition
|
Day and night adaptive detection
|
Customizable
|
|