Multi-Object Tracking using Cadence Tensilica ConnX 220 DSP
March 15, 2024The Client
Cadence, a prominent player in electronic design, utilizes its computational software proficiency to provide comprehensive solutions in software, hardware, and IP that empower companies to develop cutting-edge electronic systems spanning various industries. Cadence also offers a wide selection of DSPs ranging from compact and low power to high performance optimized for radar, lidar, and communications applications in ADAS, autonomous driving, V2X, 5G/LTE/4G, wireless communications, drones, and robotics.
Challenge
To demonstrate the high performance ability of Cadence CONNX 220 DSPs for Real-Time Multi-Object Tracking on a real hardware platform from a leading Chinese semiconductor.
Solution
MulticoreWare delivered a fully optimized demonstration of RADAR based Multi-Object Tracking (MOT) workload on the Tensilica ConnX DSP family by implementing an optimized Statistical Kalman Filter based tracking pipeline, as elucidated in this analysis below:
MULTI OBJECT TRACKING FLOWCHART
- Raw Radar Data: The system initiates by capturing raw ADC data from the radar.
- Intelligent Clustering: The acquired data undergoes processing through an intelligent algorithm, efficiently grouping individual data points into clusters, particularly when objects are in close proximity. Each cluster signifies a potential object.
- Track Initialization: Utilizing these clusters, individual tracks are established for each potential object, enabling continuous monitoring of their location and identity.
- Optimized Multi-Object Tracking (MOT): Ultimately, the optimized MOT algorithm leverages information from both the current frame and historical data. It identifies new objects, tracks existing ones, and accurately associates detections with their respective tracks. This ensures reliable tracking even in intricate scenarios.
The optimized MOT modules that were deployed Tensilica DSP for improved performance are:
- Clustering
- Extended Kalman Filter Preparation
- Extended Kalman Filter Prediction
- Data Association
- Extended Kalman Filter Update
- Tracker Update
Impact
- Achieved maximum of 1640 Frames per second and less DSP Cycles.
- Our optimized Radar processing solutions deliver a staggering 30x improvement in performance.
Other Use Cases
ADAS/ Autonomous Mobility
Traffic Monitoring
Robotics & Automation
Intelligence Surveillance
Healthcare - Patient Monitoring
Sports Analytics
Conclusion
This case study effectively showcases MulticoreWare’s proficiency in DSP & Low-Power, High-Performance Software Optimization. By successfully implementing Multi-Object Tracking on the real hardware of the Tensilica ConnX 220 DSP, we have demonstrated our capability to deliver exceptional performance on demanding embedded hardware platforms.
Discover how we can help you achieve groundbreaking results in Autonomous Mobility and Robotics. Contact our DSP team at info@multicorewareinc.com