Surveillance systems rely heavily on embedded vision technology to enable faster deployment across a wide range of applications. These systems are used in various fields such as traffic monitoring, security, intelligence gathering, and business analytics. The diverse nature of these applications brings several challenges that system designers must address effectively. Key challenges include:
- **Multi-camera vision** – the ability to connect both homogeneous and heterogeneous sensors.
- **Computer vision** – the use of advanced libraries like OpenCV and OpenVX.
- **Machine learning** – the implementation of inference engines using frameworks such as Caffe.
- **Higher resolution and frame rates** – which increase the data processing demands per image.
Figure 1: Example application (top: face detection and classification, bottom: optical flow)
Heterogeneous systems like the All Programmable Zynq®-7000 and Zynq® UltraScale+™ MPSoC are increasingly being used in surveillance systems. These devices combine programmable logic (PL) with high-performance ARM® cores, offering a powerful platform for real-time image processing. The tight integration between PL and PS improves responsiveness, reconfigurability, and energy efficiency compared to traditional CPU/GPU-based solutions.
Traditional SoCs often suffer from memory bottlenecks, where images need to be transferred between processing stages, leading to increased latency and power consumption. With Zynq-7000 or Zynq UltraScale+ MPSoC devices, image processing pipelines can be implemented directly in the PL, enabling a true parallel pipeline where each stage feeds into the next. This leads to deterministic response times, lower latency, and better power efficiency.
The flexible I/O interface of the PL also supports a wide range of industry-standard interfaces, including MIPI, Camera Link, and HDMI, making it easier to integrate different camera types and legacy systems. Additionally, multiple cameras can be connected in parallel, enhancing scalability.
However, the key challenge remains implementing complex algorithms without rewriting them in hardware description languages like Verilog or VHDL. This is where the **reVISIONâ„¢ stack** comes into play.
Figure 2: Comparison of a traditional CPU/GPU solution with the Zynq-7000/Zynq UltraScale+ MPSoC
The **reVISION stack** provides a comprehensive toolchain for developing computer vision and machine learning applications on Zynq devices. It is structured into three layers:
1. **Platform Development**: This layer forms the foundation for building the rest of the stack, providing definitions for the SDSoC tool.
2. **Algorithm Development**: This layer supports the implementation of image processing and machine learning algorithms, enabling their transfer to the programmable logic.
3. **Application Development**: This top layer offers support for industry-standard frameworks, allowing developers to build applications that leverage the platform and algorithm layers.
In the algorithm layer, image processing functions are developed using OpenCV, with support for accelerating features such as OpenVX kernels. For machine learning, the layer includes predefined hardware functions for implementing inference engines. These algorithms are then accessed by the application layer to create final applications that support frameworks like OpenVX and Caffe.
Figure 3: reVISION stack
The reVISION stack provides all the necessary tools to develop high-performance surveillance systems. One of its key advantages is the acceleration of OpenCV functions, divided into four categories:
- **Calculations**: Includes operations like absolute deviation, pixel operations, gradients, and integrals.
- **Input Processing**: Supports bit-depth conversion, channel operations, histogram equalization, and resizing.
- **Filtering**: Offers filters such as Sobel, Gaussian, and custom convolution.
- **Others**: Includes edge detection, thresholds, and classifiers like SVM and HoG.
These features are tightly integrated with OpenVX and the application layer, allowing developers to create efficient algorithmic pipelines in the programmable logic.
Machine Learning in reVISION
reVISION integrates with Caffe to simplify the implementation of machine learning inference engines. At the algorithm and application layers, Caffe provides a rich set of pre-trained models and functions that allow users to build and train networks efficiently. Developers can reuse models from the Caffe model zoo, reducing development time significantly.
By providing a prototxt file, the framework handles the rest, configuring the processing system and hardware optimization libraries in the programmable logic. Functions such as Conv, ReLU, and Pooling are implemented in the PL, resulting in high-performance and low-power solutions.
Figure 4: Caffe Process Integration
The use of fixed-point representations, such as INT8, plays a crucial role in improving performance. Fixed-point arithmetic is more efficient than floating-point, allowing for faster execution and reduced power consumption. The reVISION stack supports INT8 in the PL, enabling efficient use of DSP blocks for multiply-accumulate operations.
In conclusion, the reVISION stack empowers developers to leverage the full potential of Zynq-7000 and Zynq UltraScale+ MPSoC devices. Even experts can implement complex algorithms in programmable logic using industry-standard frameworks, reducing development time and enabling the creation of highly responsive, reconfigurable, and power-efficient systems.
Characristic
â—High permeability
â—Low loss
â—Magnetostriction coefficient is close to zero
â—Curie temperature is high
Application fields
Filter inductance is suitable for energy storage and filtering inductors in switching power supply, because of its high BS value and low loss.Compared with iron powder core and ferrite with the same volume and magnetic permeability, it has higher energy storage capacity. More widely used in AC inductors, output inductors, rotary transformers, pulse transformers, power factor positive circuits.
Filter Inductance,New Design Filter Inductance,High Performance Filter Inductance,Cost-Optimal Filter Inductance
Anyang Kayo Amorphous Technology Co.,Ltd. , https://www.kayoamotech.com