Next Article in Journal
Assessing the Ability of Luojia 1-01 Imagery to Detect Feeble Nighttime Lights
Next Article in Special Issue
An Integrated Thermopile-Based Sensor with a Chopper-Stabilized Interface Circuit for Presence Detection
Previous Article in Journal
Development of Open-Tubular-Type Micro Gas Chromatography Column with Bump Structures
Previous Article in Special Issue
A Fully-Differential Switched-Capacitor Dual-Slope Capacitance-To-Digital Converter (CDC) for a Capacitive Pressure Sensor
Open AccessArticle

An FPGA-Based Ultra-High-Speed Object Detection Algorithm with Multi-Frame Information Fusion

by Xianlei Long 1,2, Shenhua Hu 1,2, Yiming Hu 1,2, Qingyi Gu 1,2,* and Idaku Ishii 3
1
The Research Center of Precision Sensing and Control, Institute of Automation, Chinese Academy of Sciences, Beijing 100190, China
2
The School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing 101408, China
3
The Robotics Laboratory, Department of System Cybernetics, Hiroshima University, Hiroshima 739-8527, Japan
*
Author to whom correspondence should be addressed.
Sensors 2019, 19(17), 3707; https://doi.org/10.3390/s19173707
Received: 28 June 2019 / Revised: 17 August 2019 / Accepted: 21 August 2019 / Published: 26 August 2019
(This article belongs to the Special Issue Advanced Interface Circuits and Systems for Smart Sensors)
An ultra-high-speed algorithm based on Histogram of Oriented Gradient (HOG) and Support Vector Machine (SVM) for hardware implementation at 10,000 frames per second (FPS) under complex backgrounds is proposed for object detection. The algorithm is implemented on the field-programmable gate array (FPGA) in the high-speed-vision platform, in which 64 pixels are input per clock cycle. The high pixel parallelism of the vision platform limits its performance, as it is difficult to reduce the strides between detection windows below 16 pixels, thus introduce non-negligible deviation of object detection. In addition, limited by the transmission bandwidth, only one frame in every four frames can be transmitted to PC for post-processing, that is, 75% image information is wasted. To overcome the mentioned problem, a multi-frame information fusion model is proposed in this paper. Image data and synchronization signals are first regenerated according to image frame numbers. The maximum HOG feature value and corresponding coordinates of each frame are stored in the bottom of the image with that of adjacent frames’. The compensated ones will be obtained through information fusion with the confidence of continuous frames. Several experiments are conducted to demonstrate the performance of the proposed algorithm. As the evaluation result shows, the deviation is reduced with our proposed method compared with the existing one. View Full-Text
Keywords: ultra-high-speed vision; object detection; field-programmable gate array; histogram of oriented gradient; multi-frame information fusion model ultra-high-speed vision; object detection; field-programmable gate array; histogram of oriented gradient; multi-frame information fusion model
Show Figures

Figure 1

MDPI and ACS Style

Long, X.; Hu, S.; Hu, Y.; Gu, Q.; Ishii, I. An FPGA-Based Ultra-High-Speed Object Detection Algorithm with Multi-Frame Information Fusion. Sensors 2019, 19, 3707.

Show more citation formats Show less citations formats
Note that from the first issue of 2016, MDPI journals use article numbers instead of page numbers. See further details here.

Article Access Map by Country/Region

1
Back to TopTop