Custom CCA - 2023.2 English

The custom CCA function takes a binary image as input which contains a fruit on a conveyer belt (black background) and returns the total fruit pixels minus defect, total defect pixels and defect image which has the defects marked as ‘255’. This function is a custom made solution for defect detection in fruit, which ideally works with other pre-processing functions.

The custom CCA algorthm works in two-passes. The first pass includes labelling the background, foreground and defect in forward and reverse raster-scan. The second pass to perform an ‘&’ operation over the forward and reverse partial output data.

API Syntax

template <int HEIGHT, int WIDTH>
void ccaCustom(
uint8_t* in_ptr1,
uint8_t* in_ptr2,
uint8_t* tmp_out_ptr1,
uint8_t* tmp_out_ptr2,
uint8_t* out_ptr,
int& obj_pix,
int& def_pix,
int height,
int width)

Parameter Descriptions

The following table describes the template and the function parameters.

Table 479 Table accumulate Parameter Description
Parameter	Description
HEIGHT	Maximum height of input and output image.
WIDTH	Maximum width of input and output image.
in_ptr1	Input 8-bit image pointer for forward pass, binary 8-bit image (‘0’ and ‘255’)
in_ptr1	Input 8-bit image pointer for the parallel computation of reverse pass, binary 8-bit image (‘0’ and ‘255’)
tmp_out_ptr1	8-bit pointer to store and read from the temporary buffer in DDR for the forward pass. This memory must be allocated before the kernel call.
tmp_out_ptr2	8-bit pointer to store and read from the temporary buffer in DDR for the reverse pass. This memory must be allocated before the kernel call.
out_ptr	Output 8-bit image pointer for the which contains the defects image. Defect pixels are marked as ‘255’.
obj_pix	output - no. of object/foreground pixels without the count of defect pixels.
def_pix	output - no. of defect pixels in the object/foreground.
height	Height of the input image
Width	Width of the input image

Resource Utilization

The following table summarizes the resource utilization for custom CCA, generated using Vitis HLS 2021.1 tool for the xczu9eg-ffvb1156-1-i-es1, to process a FHD (1080x1920) image.

Table 480 Table customCCA Function Resource Utilization Summary
Operating Mode	Operating Frequency (MHz)	Utilization Estimate
Operating Mode	Operating Frequency (MHz)	BRAM_18K	DSP_48Es	FF	LUT	CLB
1 Pixel	300	10	10	11166	7556	1757

The following table summarizes the resource utilization for custom CCA, generated using Vitis HLS 2021.1 tool for the xczu9eg-ffvb1156-1-i-es1, to process a 4K image.

Table 481 Table 16. customCCA Function Resource Utilization Summary
Operating Mode	Operating Frequency (MHz)	Utilization Estimate
Operating Mode	Operating Frequency (MHz)	BRAM_18K	DSP_48Es	FF	LUT	CLB
1 Pixel	300	12	10	11199	7804	1748

Performance Estimate

The following table summarizes the performance for custom CCA, as generated using Vivado HLS 2019.1 tool for the xczu9eg-ffvb1156-1-i-es1, to process a FHD (1080x1920) image.

Table 482 Table 17. customCCA Function Performance Estimate Summary
Operating Mode	Latency Estimate
Operating Mode	Max Latency (ms)
1 pixel operation (300 MHz)	14