Illegal whistle capture system based on NVIDIA Jetson TX2 uses advanced sound localization algorithm, which can accurately locate vehicle whistle sounds, ignoring interference noises such as car engines, brakes, special vehicle tweets and electric horn car sounds. At the same time, the whistle capture system Generate punishment basis pictures that meet the requirements of GAT832-2014 "Technical Specifications for Image Forensics of Road Traffic Safety Violations", and generate 2s "sound" videos, and store photos, sounds, and other evidence of the illegal whistle process in real time to the center's back office for further verification All the snap management data can be seamlessly connected to the traffic police illegal monitoring system, providing an authoritative basis for whistle snap capture and evidence collection.
The dynamic portrait deployment control retrieval system is based on the latest research results in the field of artificial intelligence. It uses a combination of a front-end embedded smart camera and a central high-performance portrait comparison engine (based on NVIDIA Jetson TX2) to capture pedestrian images in real time and compare them with the list of deployment controls. Yes, real-time alarm. The system has the characteristics of high accuracy, low false alarm rate, small network resource occupation, and strong combat performance. It is widely applicable to trains, bus stations, checkpoints, checkpoints, and large-scale events, which need to control the entry and exit of personnel. Place.
By adopting the NVIDIA Jetson TX2 Module to build the world's leading all-scene image supervision device "Smart Sky Eye", it can realize 7x24-hour monitoring and form complete image evidence based on the vehicle's driving trajectory. Electronic payment, realizing the management mode of unmanned tolling, and the parking experience of "inductive payment".
The high-precision passenger flow sensing device directly obtains the stream information of the webcam, decodes the stream information, combines the NVIDIA Jetson TX2 Module to perform pedestrian learning based on deep learning, and then uses the association information between consecutive multiple frames to generate multiple This kind of hypothetical trajectory calculates the a priori probability of each hypothesis, and combines the appearance and motion information to make predictions. At the same time, it introduces color information and depth information to match, improves the correlation accuracy, realizes dynamic multi-target tracking, and outputs target tracking trajectory. Through the analysis of motion trajectory, statistical parameters such as the flow of people, speed, flow direction, density in the area or through the boundary line can be accurately obtained. Finally, the obtained data information is uploaded to the database, and the update frequency of 1 second is used to ensure the security of the data. Combine various data resources to establish a big data analysis platform centered on passenger flow, and provide comprehensive services for passenger flow security management, risk prevention, efficiency improvement, and business value mining.
Adopted NVIDIA Jetson TX2 Module deep learning platform, which greatly improved the accuracy, computing time and efficiency of detected pedestrian data, and obtained passenger flow calculation accuracy greater than or equal to 96%; passenger flow density detection accuracy greater than or equal to 95%; average passenger flow speed The detection accuracy is greater than 92%; the accuracy of pedestrian tracking trajectory is greater than or equal to 95%.