جميع المنتجات
-
(غريغ بلايدز)1أفضل خدمة وأفضل سعر 2آمل أن نتمكن من القيام بمزيد من الأعمال معا في المستقبل. 3بما أن خدمتك جيدة جداً، سأبث كلمة جيدة عن (شيشيان فارورد) بين جماعة (نانتشانغ سي جي 6) -
(أرشاد سليم)1.لن أتصل بأي شركة مباشرة، ستكون مصدر عملي في الصين. 2أنتم رائعون، لدينا تعاون ناجح جداً هذه هي نتائج جهودنا المشتركة
Ascend 310B High-Computing-Power Image Processing Board with Heterogeneous Architecture and Rich Interface Resources for Edge AI Applications
| مكان المنشأ | الصين |
|---|---|
| اسم العلامة التجارية | Forward |
| رقم الموديل | فود-310ب |
| الحد الأدنى لكمية | 1 |
| وقت التسليم | 5-8 أيام عمل |
| شروط الدفع | تي/تي |
| القدرة على العرض | 10000/سنة |
تفاصيل المنتج
| إبراز | Ascend 310B image processing board,edge AI heterogeneous architecture board,thermal imaging monocular with rich interfaces |
||
|---|---|---|---|
منتوج وصف
Ascend 310B High-Computing-Power Image Processing Board for Infrared Cameras / Thermal Imagers
Product Overview
This intelligent image processing board is designed for edge-side applications, featuring a heterogeneous architecture combining the Huawei Ascend 310 series (ZXA41D20) processor and JFM7K230T FPGA. Characterized by high integration, availability, and scalability, it delivers efficient intelligent inference computing while providing general-purpose CPU and GPU processing capabilities with specialized image processing functions and abundant interface resources.
Applications
This board serves as a core component integrated into equipment such as seekers, electro-optical (EO) pods, and gimbals/turrets for target observation, aiming, and reconnaissance through deployment of various application software or algorithmic programs.
Key Features
- Heterogeneous Architecture: CPU+NPU+GPU multi-core processing architecture meeting edge AI inference demands while supporting CPU and GPU functionalities
- Rich Interface Resources: Wide variety of video and communication interfaces for diverse scenario requirements
- Dedicated Image Processing Unit: Hardware-accelerated image preprocessing and video encoding/decoding functions
- Broad Framework Support: Supports mainstream neural network models and programming frameworks
- Optimized SWaP-C: High integration, ultra-compact form factor, and cost-effectiveness
- 100% Indigenous Component Sourcing: Utilizes 100% domestic components for independent controllability and supply-chain security
Key Performance Specifications
Functions
- SDI Video Capture: Video source input and acquisition via SDI interfaces
- CameraLink Image Acquisition: Image capture from standard CameraLink industrial cameras
- MIPI CSI Camera Interfacing: Direct image acquisition from MIPI CSI camera modules
- Serial Communication: Robust serial communication capabilities (UART/RS-232/RS-422/RS-485)
- Ethernet Connectivity: High-speed Ethernet communication capabilities
- Model Inference Runtime: Optimized environment for AI inference supporting YOLO, ResNet, ViT, LSTM, MobileNet, and VGG networks
- Video Encoding: Real-time hardware encoding in H.265 (HEVC) and H.264 (AVC) formats
Core Hardware Configuration
| Configuration | Specifications |
|---|---|
| CPU | Ascend 310B, 4x TAISHAN V200M, 1.6 GHz |
| NPU | 1x DaVinci V300 AI Core @ 1.0 GHz INT8: 20 TOPS FP16: 10 TFLOPS |
| GPU | ARM Mali G52 GPU Core |
| FPGA | JFM7K230T type |
| RAM | 12GB LPDDR4X (96-bit) |
| Built-in Storage | 64MB SPI Flash |
| Encoding/Decoding | H.264/H.265 decoding: 40-channel 1080P@30FPS, 4-channel 4K@75FPS H.264/H.265 encoding: 20-channel 1080P@30FPS, 3-channel 4K@50FPS |
Hardware Interfaces
| Interface | Specifications | Description |
|---|---|---|
| Power Supply | DC 12V ±1V | Max current: 2.5A; typical: ~1.5A; micro-rectangular connector |
| Video Input | 2× SDI | 3G-SDI, HD-SDI; up to 1920×1080@60fps; SMA RF connectors |
| Video Input | 1 × CameraLink | 1x Full or 2x Base; max 1280x1024@100fps; SDR26 connectors |
| Video Input | 2× MIPI CSI | Two 4-lane MIPI CSI interfaces; micro-coaxial connectors |
| Video Output | 2× Ethernet Ports | Gigabit auto-negotiating Ethernet; RJ45 interfaces |
| USB | 1 × USB 3.0 | Type-C |
| Serial Port | 6 × RS422 | 6-Channel 310B RS422 Serial Port |
| Serial Port | 4× RS422 | 4-Channel FPGA RS422 Serial Port |
| Serial Port | 2 × UART | Reuse 2 of the 6 310B RS422 serial ports |
Board Specifications
Installation dimensions: 140 mm (L) × 95 mm (W) (excluding connectors). The schematic diagram of board dimensions is shown in Figure 1. Hardware interfaces are shown in Figure 2.
Environmental Adaptability
Operating Temperature: -40℃ to +60℃
Software Development Support
Driver Software
All core peripheral drivers have been adapted and verified on the openEuler operating system. Users can conduct application-layer development using standard system calls without complex underlying driver concerns.
- Core Processor Driver: Complete adaptation of Ascend 310B system drivers including power management, clock management, and interrupt control
- Video Input Drivers: Adapted SDI, CameraLink, and MIPI CSI interfaces with stable acquisition and processing capabilities
- Network Interface Driver: Adaptive Gigabit Ethernet supporting network video streaming and remote command control
- Communication Interface Drivers: Adapted UART, I2C, and GPIO interfaces for external device control and data interaction
Video Capture
The board integrates three types of video input interfaces to meet different sensor connectivity requirements:
- SDI: Provides standard V4L2 interface for image frame data acquisition
- CameraLink: Supports high-speed industrial camera video data input with frame grabber SDK
- MIPI: Provides standard V4L2 device nodes for video capture and image data acquisition
Algorithm Deployment
Leveraging the CANN intelligent heterogeneous computing architecture, the board delivers highly efficient and stable inference capabilities supporting mainstream AI frameworks including MindSpore, PyTorch, PaddlePaddle, and TensorFlow with MindX application enablement components.
The platform efficiently handles vision tasks such as target detection, object classification, and target tracking, providing comprehensive AI inference runtime environment and development interfaces.
Sample algorithm running projects for reference:
- Target Detection: Deployment example based on YOLO series models for real-time video stream analysis
- Target Tracking: Integrated tracking algorithm examples for closed-loop control based on miss distance data
- Video Stream Integration: Complete video AI processing pipeline with real-time OSD results visualization
Note: Examples are for engineering implementation reference only; we are not responsible for the performance or efficacy of the algorithms themselves.
Product Delivery Checklist
| NO. | Substance | QTY |
|---|---|---|
| 1 | Image Processing Card | 1 set |
| 2 | Heat Sink | 1 pcs |
| 3 | Power Cable | 1 set |
| 4 | Video Input (SDI) Cable | 2 pcs |
| 5 | Video Input (MIPI) Cable | 2 pcs |
| 6 | Video Output (Network) Cable | 1 pcs |
| 7 | RS422 Serial Cable | 1 pce |
| 8 | External Sync Cable | 1 pcs |
| 9 | User Manual | 1 |
| 10 | Certificate of Conformity | 1 |
المنتجات الموصى بها
-
VIDEO
-