Video Occupancy Sensor, Blackfin
The Video Occupancy Sensor library for the Blackfin processor provides a suite of functions that use input from Vision Sensors to detect occupancy or presence of humans or vehicles in outdoor spaces such as parking lots.
Top
Overview
Features and Benefits
- Cache: Both Instruction and Data Cache enabled
- MDMA: One channel Memory DMA is used
- Input Format:
- UYVY422 in ITU-R BT.656 format or YUV420 planar format or UYVY422 raw progressive or Y only.
- Output format:
- Number of Detected Objects
- Features of Detected Objects, such as Area, Bounding Rectangle Box, Centroid.
- Features of Tracked Objects, such as Area, Bounding Rectangle Box, Centroid and Tracking Trajectory over last 100 frames/images.
- Output frames:
- Reference/Background frame
- Difference frame:
- MOG Background/Foreground Segmented Binary Image
- Algorithm:
- Mixture of Gaussian (MOG/GMM) used as background modeling function
- Tracking Algorithm
- Blob Merging based on size and spatial location. The code for merging is implemented in the application and hence completely configurable.
- Code compatible across the Blackfin Processor Family ADSP-BF7xx
- Release Format: Real time and File I/O application code in C, Library, C and ASM Sources of the library
- C and ASM Sources of VOS Primitives
Product Details
Video Occupancy Sensor is designed to detect occupancy or presence of humans or vehicles in outdoor spaces such as parking lots using Vision Sensors. The detection is used for smart switching off/on lights and lighting control for energy savings. This product will do occupancy sensing for outdoor spaces such as outdoor parking lots etc. It detects occupancy when vehicles and/or people come into the field of view of the sensor and rejects other motion such as swaying trees, rain/snow and other such movements.
Licensing
Each module supports the Analog Devices, Inc. (ADI) Blackfin or SHARC Processor family and is a licensed product that is available in object code format. Recipients must sign or accept a license agreement with ADI prior to being shipped or downloading the modules identified in the license agreement.
Performance Metrics
MIPS summary:
| Memory (Bytes) | Performance |
||||||
| Test |
Width/Height | Code (KiB) |
Data RAM (KiB) |
Constant Data Tables (KiB) |
Frame Buffer |
MIPS (Average) |
MIPS (Peak) |
| YUV420, 25 fps |
352 x 288 |
39.39 |
25.48 | 0.26 |
1380.1 | 189 |
208 |
| UYVY, Aptina Sensor Input on PPI |
640 x 480 |
39.39 | 25.48 | 0.26 | 1380.1 | 189 |
205 |
- MIPS measured using optimal memory layout, on ADZS-BF707 BLIP2 platform @ 30 fps. All code, data were placed in L1 memory where possible and rest was placed in cache. The frame state buffers were placed in L3 or L2 and stack was placed in L1scratch memory. 16KiB L1 instruction memory was configured as instruction cache and 32KiB of L1 data memory was configured as data cache.
- "Data RAM" for one instance, includes Stack, Scratch, Instance/State for VGA resolution. “Constant Data Tables” for one instance, includes constants and constant tables. “Frame Buffer” for one instance of VGA resolution includes frame buffers (larger buffers in L3 memory) which are both state and scratch memory.(1 KiB = 1024 Bytes)
- Data Cache and Instruction Cache are enabled. The cache is set in “write back” and “long cache” (DCBS=1) mode. Memory DMA is used with 64 bit DMA for ADSP-BF707.
- Measurements done with CCLK=384 MHz, SCLK=192 MHz on ADSP-BF707.
Systems Requirements
- Windows 7 Professional/Enterprise/Ultimate (32 and 64-bit).
- Minimum of 2 GHz single core processor, 3.3 GHz dual core is recommended.
- Minimum of 1 GB memory (RAM), 4 GB is recommended.
- Minimum of 2 GB hard disk (HDD) space is required.
- CrossCore® Embedded Studio for Analog Devices Processors.