Development of grain sorting algorithm in a flow with simultaneous determination of rate of each weevil.
The algorithm must determine the number of defective dots (pixels) inside the grain and remove seeds that do not meet pre-defined criteria: if the number of defective pixels greater than a given, it is considered defective grains. In this case, the algorithm must determine the contour of each weevil and calculate the optimal point on the path, which then have to direct the flow of air from the pneumatic valve, if this proves to be defective weevil.
From the linear image sensor to signal processor (DSP) come line (frame) image stream falling on sloping trough grains. Horizontal Frequency sensor about 4500 per second. Thus, each object is formed in DSP line. The processor DSP frames come in a monochromatic color grayscale 14 bits (16384).
1. The input data of the algorithm: The threshold (or region) to determine the brightness of defective pixels threshold (or region) to determine the brightness of the circuit grains, the number of defective pixels, and the parameter (or parameters) to ignore the possible dust in the flow of grain, other arbitrary parameters if necessary .
2. The algorithm should be linked to the processor DSP: ADSP-BF561 from Analog Devices. Data come from the image sensor to the parallel port PPI0 with 14-bit word length.
3. The processor must be able to process the flow lines image sensor with a frequency of at least 4500 lines per second and with a certain flow of grains (see attached photos). The number of pixels in a linear image sensor is 2048.
4. The algorithm should be able to simultaneously determine the velocity of each weevil for precise removal of defective grains (objects).
The accompanying (for example) displayed photographs stream grains (sunflower) with a small number of different contaminants (light objects). File size 2048x1000. The frame rate of 5000 per second. Accordingly, the 1000 lines - is 200 milliseconds. Gray scale is reduced to 8 bits.