Details about Vision sensor xSOFTip?

[mon2]

Hi (Namaste :) Sudha. Perhaps these questions are directed to yourself.

Thanks for your posted GitHub posts for the Vision IP. The sample code is working well for tracking of the car.

For the noted project, can the frame rate be increased ? What steps should be taken to improve the tracking frame rates to be more fluid (real-time) ?

Also, you have a Vision Sensor project on GitHub posted but so far with no documentation.

Is that project complete ?

Are more details available ?

How is the sensor attached to the XMOS ? Which sensor was used ? Is it WDR capable ? Have you reviewed such sensors ?

We attempted a while ago to interface with the OV7670 but got lost in the reading from the sensor and exporting the sensor data to the attached LCD panel. Had limited time before for the review so we wish to revisit that project (machine vision) again in the near future.

Could the CMOS sensor project serve as the front end interface to offer realtime tracking of objects using the XMOS core(s) ?

Thanks again for your excellent and interesting project posts !

Kumar

[sudha]

Hi Kumar,

Thanks for your interest in vision sensor projects.

The object tracking demo currently stores a sequence of image frames on an external SDRAM. Since there is a limitation in the number of frames to be stored (only 12 frames are stored), we have deliberately introduced delay between successive frames while displaying on LCD. This is for the purpose of demo.

In real scenario, the object tracking is done on the image frames from a camera. In that case, we will have only double frame buffers to store two successive frames at a time for processing. The frame rate depends on the processing time and we have achieved processing around 8 frames in a second.

We have a working image sensor slice and it is yet to be released. The demo at sc_image_sensor displays the image frames captured by the camera in the slice on the LCD. The sensor is Aptina's MT9V034.

-Sudha