RU2818487C1 - Local video analytics device - Google Patents

Abstract

FIELD: computer engineering.

SUBSTANCE: device includes a housing with connectors and a cover with a silicone seal, as well as a motherboard on which a system-on-module with a neural processor are installed, temperature sensor module, system start-up device, connectors for connecting a cooler, input power, connecting to an Ethernet network, USB ports, a slot for connecting an SD card, microUSB connector, power converters, protection chips, supervisors, power stabilizers, debugging interfaces and connectors for connecting peripherals, wherein the cover with the silicone seal simultaneously serves as a passive cooling radiator, the system-on-module includes software that enables to decode and analyse the incoming video stream, motherboard is made with possibility of self-diagnosis, reading data from temperature sensor, SD-card contains software providing automatic realization of algorithms of data collection and processing from video camera, at that system-on-module input is equipped with DC line.

EFFECT: possibility of identification of abnormal events, as well as integration with warning systems of a higher hierarchical level.

1 cl, 8 dwg

Description

The proposed invention relates to the field of intelligent video surveillance systems and can be used for video surveillance of equipment and employees of production premises in order to prevent the occurrence of emergency situations associated with violation of industrial safety requirements.

A local video analytics device is known (US 2011/0050947 A1, applicant VIDEOIQ INC, pub. 2011-03-03 - L.1), including a circuit design solution, a software solution and a design solution that allows you to perform tasks of recognizing video images incoming from external cameras in industrial conditions of use with a built-in algorithm for issuing event cards to a higher-level control system.

The disadvantage of the known device is its high power consumption requirements, as well as the high complexity of detachable connections to the motherboard.

The prototype of the proposed invention is the NVIDIA Jetson platform, a leading platform for autonomous vehicles and robots, as well as other embedded applications (https://www.nvidia.com/ru-ru/autonomous-machines/embedded-systems/). It includes compact, high-performance Jetson modules, the NVIDIA JetPack™ SDK for software acceleration, and an ecosystem of partners offering sensors, SDKs, services and products to accelerate development. Jetson is compatible with the same AI software and cloud technologies found on other NVIDIA platforms. It provides performance and energy efficiency for building software-defined infrastructure for autonomous machines.

The fundamental difference between the local video analytics device ETRTs.466459.001 and the prototype is the orientation of the design, circuitry and software solution towards video analytics with the ability to accumulate, structure data and create event cards based on the processed information in order to prevent negative factors.

In addition, the difference from NVIDIA Jetson is:

1. Use of video analytics ETRTs.466459.001 in a local device as a loader for open Linux-based solutions (rejection of proprietary software solutions);

2. Local video analytics device ETRTs.466459.001 is a complete technical solution, which provides, among other things, structural protection from external influences through the use of protective plugs for connectors and an aluminum cover with a silicone seal, and eliminates overheating of the electronics.

The objectives of the proposed invention are to increase the safety of the production process, reduce the risk of downtime, and increase the speed of response to emergency situations.

The technical result is the creation of the ability to recognize abnormal events, as well as integration with warning systems of a higher hierarchical level.

The specified technical result is achieved by a local video analytics device, including a housing with connectors and a cover with a silicone seal, as well as a motherboard on which a system-on-module with a neural processor, a temperature sensor module, a system launcher, connectors for connecting a cooler, and input power are installed. , Ethernet network connections, USB ports, SD card slot, microUSB connector, power converters, protection chips, supervisors, power stabilizers, debugging interfaces and connectors for connecting peripherals, while the cover with a silicone seal simultaneously serves as a passive cooling radiator, the system-on-module includes software that allows decoding and analysis of the incoming video stream, the motherboard is designed with the ability to self-diagnostics, reading data from the temperature sensor, the slot for connecting an SD card includes software that ensures the automatic implementation of algorithms for collecting and processing data from the video camera , while the system-on-module input is equipped with a DC volt line.

The invention is illustrated by drawings, where in FIGS. 1 shows a general diagram of the device; Fig. 2 – general view of the local video analytics device ETRTs.466459.001, in Fig. 3 – protective plug, in Fig. 4 – aluminum cover with silicone seal, in fig. 5 – DIN rail, in fig. 6 – printed motherboard, in Fig. 7 – block diagram of the system-on-module board, Fig. 8 – block diagram of the motherboard.

The local video analytics device ETRTs.466459.001 is a hardware and software complex for processing video information from CCTV cameras and is intended to work as part of the Video Analytics software package, which provides automation of receiving and processing data from a video camera for identifying and recognizing various objects and events, as well as organizing transferring information about recognized objects and events to the business logic server.

The local video analytics device ETRTs.466459.001 consists of a housing 1 with installed connectors 2, 3, 4 with protective plugs 5 and an aluminum cover with a silicone seal 6, which simultaneously serves as a passive cooling radiator. Connector 2 – RJ45 – 1 pc, connector 3 – OL1312S2 – 1 pc, connector 4 YU-USB3-JSX-01-001 – 2 pcs. To mount housing 1, two DIN holders are provided, which are attached to the bottom of housing 1 with screws and fixed on DIN rail 7.

Printed motherboard 8 is located in housing 1. Motherboard 8 is designed to house a set of components and modules necessary to ensure the overall performance of the local video analytics device ETRTs.466459.001, as well as to communicate between all elements connected to it and coordinate their work.

The printed circuit board 8 is equipped with a temperature sensor module 9 (operating range from -55 to +125 °C), a system launcher 10, a connector for connecting a cooler 11, a connector for connecting input power 12, a connector for connecting to an Ethernet network 13 with a function passive PoE, two USB 3.0 ports 14 – one dual connector, SD card slot 15, microUSB connector 16.

Also installed on the printed circuit board 8 is a system-on-module 17 (SoM). SoM module 17 is the main computing node of the entire local video analytics device ETRC.466459.001, provides processing of all data from the periphery and issues the corresponding tasks to the connected executive devices.

The basis of the SoM module 17 installed on the motherboard 8 is the neural processor 18 i.MX8M+ from Solidrun, which provides the following main characteristics for the software:

• 4x Arm® Cortex-A53 - processor with four cores and frequency up to 1.8 GHz;

• 1x Arm® Cortex-M7 - one processor core for working with vectors and mathematical operations with floating point numbers with a frequency of up to 800 MHz;

• Cadence® Tensilica® HiFi4 DSP - processor for processing digitized signals, including voice and natural language with a frequency of up to 800 MHz;

• Neural Processing Unit - Neural processor 2.3 TOPS.

The SoM module 17 requires only one 5 volt direct current (VDC) line at its input, which simplifies the power supply system for the motherboard 8. The motherboard 8 also contains power converters 19, debug interfaces 16, and connectors for connecting peripherals 20.

The local video analytics device ETRTs.466459.001 is powered in two ways: via an Ethernet channel using Passive PoE technology, or via a 12 VAC line. When connected using Passive PoE technology, the power is up to 71.3 W, when connected via 12 VAC lines - up to 100 W (including possible additional peripheral debugging interfaces). The input voltage complies with the PoE standard and is 36-57 VAC. The maximum power consumption without connecting peripherals is up to 45 W.

After the input power 12, on the motherboard 8 there are protection chips and supervisors (current/voltage), as well as power stabilizers for all consumers: +5 V (main power bus), +3.3 VAC (main power bus of logic chips), +12 VAC stabilized (for RS485).

The device works as follows.

The local video analytics device ETRTs.466459.001 receives video streams from cameras and tasks from business logic via Ethernet connector 13. Automatic implementation of algorithms for collecting and processing data from a video camera occurs through the software of the local video analytics device, which is located on an SD card in slot 15. For account of the software of the local video analytics device, the SoM module 17 of the board 8 decodes and analyzes the video stream. When analyzing a video stream, neural networks can be used, the operation of which is optimized for the neural processor located on SoM module 17. Based on the results of working with the stream, event cards are generated in accordance with the connected analytics modules. The generated event is transmitted back to the business logic server via Ethernet connector 13. Motherboard 8 performs self-diagnosis by reading data from temperature sensor 9.

The software located on the SD card in slot 15 for video analytics LU is designed to ensure automatic implementation of algorithms for collecting and processing data from a video camera and transferring the processed information to the business logic server.

The product's firmware has the following functionality:

- Receiving video stream from video cameras;

- Video stream decoding;

- Generating data on incidents detected at the stage of primary analysis;

- Integration of software detectors - analytics modules;

- Perimeter intersection detection: setting zones on the frame, detecting the location and intersection of these zones by any objects or specified ones (people or cars);

- The operation of the following typical video analytics algorithms: motion detection: identifying areas of the frame with moving objects in a given area of the video stream, detecting movement of only specified objects;

- Object tracking: allows you to track the movement of an object and record the time of its appearance in the frame;

- Construction of a heat map based on human movement and the presence of general movements in the marked area of the frame;

- Determining the direction of movement of an object in the frame;

- Calculate the time a person spends in a designated area or in a frame;

- Overlaying visualization of analytics results on a video image of an object with time stamps and other accompanying information about the recorded event;

- Transmission of messages to the central server of consolidated video analytics by generating data packets consisting of visual (static or video) data and metainformation that fully describes the event within the API format of the central server message protocol.

The proposed invention allows complex calculations of neural network algorithms, including a system calculated radiator system for heat removal. Device protection class IP55.

Claims (1)

Local video analytics device, including a housing with connectors and a cover with a silicone seal, as well as a motherboard on which a system-on-module with a neural processor, a temperature sensor module, a system launcher, connectors for connecting a cooler, input power, and network connections are installed Ethernet, USB ports, SD card slot, microUSB connector, power converters, protection chips, supervisors, power stabilizers, debugging interfaces and connectors for connecting peripherals, while the cover with a silicone seal simultaneously serves as a passive cooling radiator, system-on- The module includes software that allows decoding and analysis of the incoming video stream, the motherboard is designed with the ability to self-diagnostics, reading data from the temperature sensor, the SD card contains software that ensures the automatic implementation of algorithms for collecting and processing data from the video camera, while the input system The on-module is equipped with a DC volt line.