Dragster Line Scan

Table of Contents

• Overview
  • Features
  • Supported pixel formats
• Image size and sensor AOI
  • FG_set_aoi()
  • FG_set_scan_param()
• Special features
• Dual Line sensors
  • Dual Line Monochrome sensors
  • Dual Line RGB sensors
    • RGB Demosaicing
    • Raw format
• Shading Correction
  • Recording a reference image
• Triggering the sensor
• Embedded line and trigger counters

Overview

The following AMS Dragster sensor models are available for the VisionCam XM:

Supported Dragster sensors

Sensor name	Type	eFG_CAMERA_TYPE	Number of pixels	Pixel size	Optical length
DR-2k-7	Monochrome	FG_CAMERA_TYPE_2048_1_IMAGO_VCx_DR2K7	1x 2048	7 µm	14.34 mm
DR-4k-7		FG_CAMERA_TYPE_4096_1_IMAGO_VCx_DR4K7	1x 4096	7 µm	28.67 mm
DR-4k-3.5		FG_CAMERA_TYPE_4096_1_IMAGO_VCx_DR4K35	1x 4096	3.5 µm	14.34 mm
DR-8k-3.5		FG_CAMERA_TYPE_8192_1_IMAGO_VCx_DR8K35	1x 8192	3.5 µm	28.67 mm
DR-2x2k-7	Dual Line Monochrome	FG_CAMERA_TYPE_2048_2_IMAGO_VCx_DR2X2K7	2x 2048	7 µm	14.34 mm
DR-2x4k-7		FG_CAMERA_TYPE_4096_2_IMAGO_VCx_DR2X4K7	2x 4096	7 µm	28.67 mm
DR-2x2k-7-RGB	Dual Line Color	FG_CAMERA_TYPE_2048_2_IMAGO_VCx_DR2X2K7_RGB	2x 2048	7 µm	14.34 mm
DR-2x4k-7-RGB		FG_CAMERA_TYPE_4096_2_IMAGO_VCx_DR2X4K7_RGB	2x 4096	7 µm	28.67 mm

The following image shows the functional bocks and pipeline stages:

Functional blocks for the Dragster sensor

Features

• Global Shutter with interleaved integration and readout
• Low noise pixels with true CDS (Correlated Double Sampling)
• ADC with adjustable gain: FG_set_gain()
• Internal pixel processing using 12 bit resolution
• Automatic black level correction by using dedicated black sensor pixels
• Gain Shading Correction
• Independent digital gain for R / G / B channels: FG_set_gain_rgb()
• Adjustable line delay for dual line sensors
• Mirrored readout
• AOI selection for scan rate increase: FG_set_aoi()
• Demosaicing for dual line RGB sensors
• Free-run mode with adjustable scan rate
• Hardware triggered mode (controlled by RTCC)
• Insertion of trigger and line counters for synchronization with software

Supported pixel formats

The supported pixel formats and maximum scan rates are listed below:

Supported pixel types

Sensor type	eFG_PIXEL_TYPE	Output format	AOI width	Max. scan rate
Monochrome	FG_PIXEL_TYPE_Y_8	1 byte per pixel	≤ 4704	80 kHz
			8192	46 kHz
	FG_PIXEL_TYPE_Y_16	2 bytes per pixel containing 10 valid bits	≤ 2352	80 kHz
			4096	46 kHz
			8192	23 kHz
Dual Line Monochrome	FG_PIXEL_TYPE_Y_8	Two lines with 1 byte per pixel	≤ 2352	80 kHz
			4096	46 kHz
	FG_PIXEL_TYPE_Y_16	Two lines with 2 bytes per pixel containing 10 valid bits	≤ 1176	80 kHz
			2048	46 kHz
			4096	23 kHz
Dual Line Color	FG_PIXEL_TYPE_Y_8	Raw mode: Two lines with 1 byte per pixel Even lines: B-G; Odd lines: G-R	≤ 2352	80 kHz
			4096	46 kHz
	FG_PIXEL_TYPE_Y_16	Raw mode: Two lines with 2 bytes per pixel containing 10 valid bits Even lines: B-G; Odd lines: G-R	≤ 1176	80 kHz
			2048	46 kHz
			4096	23 kHz
	FG_PIXEL_TYPE_RGB_24	Demosaicing mode: One line with 3 bytes per pixel containing 8 bit R/G/B	≤ 1560	80 kHz
			2048	61 kHz
			4096	30 kHz
	FG_PIXEL_TYPE_RGBX_32	Demosaicing mode: One line with 4 bytes per pixel containing 8 bit R/G/B/dummy	≤ 1176	80 kHz
			2048	46 kHz
			4096	23 kHz

Note

The scan rate can be increased by reducing the AOI size in x-direction. But 80 kHz is the highest value for all sensors.

Image size and sensor AOI

For line scan sensors, multiple sensor lines are concatenated to complete an image.

Selecting a smaller image height reduces the latency, because all image lines have to be in memory before the buffer is given to the user. But it will also increase the processing overhead if the height is very small.

FG_set_aoi()

With FG_set_aoi(), the x coordinates specify the sensor columns for readout. The parameter y1 has to be zero and y2 specifies the number of lines to acquire for each image. The default image height is 512 after initialization.

Example:

UINT32 x1, x2, y1, y2;
UINT32 image_width, image_height;
eFG_PIXEL_TYPE pixel_type;
 
FG_install_camera(FG_CAMERA_TYPE_X_X_IMAGO_Vxx_AUTO);
 
// read the default values to get the sensor width:
FG_get_aoi(&x1, &y1, &x2, &y2, &pixel_type);
 
// configure new image height (200) using the full sensor width:
image_height = 200;
image_width = x2 - x1 + 1;
FG_set_aoi(x1, y1, x2, image_height-1, pixel_type);

FG_set_scan_param()

The alternative function FG_set_scan_param() can also be used to specify the sensor AOI and number of scans per image (parameter image_scan_count). The parameter scan_height is normally one and image_scan_count is therefore equal to the image height. For Dual Line sensors with a pixel type of FG_PIXEL_TYPE_Y_8 or FG_PIXEL_TYPE_Y_16, the scan height can also be two. The image height is then twice the scan count value.

Example:

The following example configures an AOI width of 1024 at the sensor's center location:

UINT32 sensor_width, sensor_height;
UINT32 scan_x, scan_y, scan_width, scan_height, image_scan_count;
UINT32 image_width, image_height;
eFG_PIXEL_TYPE pixel_type;
 
FG_install_camera(FG_CAMERA_TYPE_X_X_IMAGO_Vxx_AUTO);
 
// read the default values to get the sensor size:
FG_get_scan_param(NULL, NULL, &sensor_width, &sensor_height, &image_scan_count, &pixel_type);
 
// the sensor AOI size:
scan_width = 1024;
scan_height = 1;
// set x offset to the sensor's center:
scan_x = (sensor_width - scan_width) / 2;
// use a scan count of 200 resulting in an image height of 200, because scan_height is 1:
image_scan_count = 200;
 
// configure AOI for line scan mode:
FG_set_scan_param(scan_x, scan_y, scan_width, scan_height, image_scan_count, pixel_type);
 
// calculate the image size (same as FG_get_image_size())
image_width = scan_width;
image_height = image_scan_count * scan_height;

Note

The maximum image size is limited to 8 MB if Linux kernel version < 4.19 is used.

Special features

The sensor supports the following special features:

Special features

Property name	Description	FGCamera.so
ScanPeriod	Sensor scan period for free-run mode in microseconds (12...6553). Please note that the exposure time must be at least 2 µs shorter than the readout period. The exposure time has priority and may increase the resulting readout period. The lowest value can also be limited by the pixel format and AOI width used.	≥ 1.1.1.0
Mirror	Readout direction of the sensor. 0: normal 1: reverse
CDS_Gain	Analog gain value for the CDS pixel stage. Only 1 and 4 are allowed values.
InsertLineCounters	Enables or disables insertion of trigger and sensor readout counters at the beginning of each line, see also Embedded line and trigger counters: 0: disable counters (default) 1: enable counters
SetGainCoeff	Sets the gain coefficients for each sensor pixel (see Shading Correction). The argument is a pointer to an array of float values containing the gain values for each pixel. The maximum gain value is 4.0. The length and arrangement of the array depends on the pixel format used. If argument is zero, the gain coefficients are reset to 1.0 for each pixel. Please note that the gain settings are lost after calling FG_install_camera() unless the SaveGainCoeff command is executed.
SaveGainCoeff	Stores the current gain coefficients permanently in flash memory (see SetGainCoeff). The values are loaded automatically when calling FG_install_camera() for the next time. The argument is ignored for this option.
ImageMode	Sets the image mode: 0 .. Video output 1 .. Test image mode: horizontal ramp using 10 bit
LineDelay	For dual line sensors only: Selects readout delay for one of the sensor lines. This setting can be used to compensate the physical displacement of both lines depending on the transport direction used in the application. 0: no line delay (default) 1: The first line is delayed by one readout period -1: The second line is delayed by one readout period

Additional features are described in the following sections:

• Hardware trigger
• Embedded counters (VisionCam XM / LM)

Dual Line sensors

Dual Line Monochrome sensors

The dual line sensors increase the effective line rate above the maximum scan rate of 80 kHz. Each trigger event and integration cycle will provide two scan lines within the image.

To avoid blurring in transport direction because of the increased object speed, make sure to reduce the integration time accordingly.

Depending on the transport direction, the line delay parameter can be set to 0 or +1. This will correct the line sequence in memory. Use FG_set_special_option() with the option "LineDelay".

Dual Line RGB sensors

The dual line color version is similar to the monochrome version, but it uses an RGB filter with the following pixel arrangement:

RGB color filter arrangement

RGB Demosaicing

When using an RGB output format with demosaicing in hardware, the following requirements must be met:

• The sensor has to be triggered for TDI mode (Time Delay and Integration). This means that each object position is captured by both sensor lines after each other, by using two consecutive integration cycles or trigger events.
• To obtain good results, the trigger signal should be synchronized with the speed of the moving object. This can be accomplished by using an encoder and the RTCC, see Triggering the sensor.
• The line delay has to be set to -1 or +1, depending on the transport direction of the moving object. This will map the BG-line and the GR-line on top of each other.
  Use FG_set_special_option() with the option "LineDelay".

The output of the demosaicing stage is one line with three RGB components for each pixel:

RGB Demosaicing output

While the G component contains the full resolution, every other R- and B component is interpolated by the demosaicing filter in horizontal direction (marked with *).

It's possible to store the output with 3 bytes per pixel (FG_PIXEL_TYPE_RGB_24) or with 4 bytes per pixel (FG_PIXEL_TYPE_RGBX_32).

Raw format

This format allows for higher line rates than the RGB Demosaicing format, because it doesn’t contain the interpolated color components and therefore uses less PCIe and memory bandwidth.

When using raw output format, each integration period results in two image lines. Even lines contain the B-G pixels and odd lines contain the G-R pixels within the returned image: The AOI y position and height must be a multiple of two when using the raw format.

Image contents for raw format

With the raw Bayer format, the sensor can be triggered in TDI mode or dual line mode. The difference is in how fast the sensor gets triggered compared to the object speed and how the output lines are treated by software:

• In TDI mode, each object position is captured by both lines using two consecutive integration cycles. Software handles pairs of odd and even lines as one physical object location.
  The adjustable line delay can be set to -1 or +1, depending on the transport direction. This will place the BG-line and the GR-line for the same object location next to each other in memory.
  The image contains G components with full resolution and R- / G-components with half resolution in horizontal direction.
• In dual line mode, both lines are treated as spatially separated object locations. The effective line frequency is doubled compared to the TDI mode, because there is no overlap of both lines between integration cycles. But note that the color resolution gets reduced.
  The line delay can be set to 0 or +1, depending on the transport direction.
  To avoid blurring in transport direction because of the increased object speed, make sure to reduce the integration time accordingly.

Please note:
By activating the mirror option, the sequence of the color components gets swapped within each line.
The first line in the image will always be the B-G line, regardless of line delay setting.

Shading Correction

The processing pipeline provides independent gain values for each sensor pixel. This can be used to compensate the following problems:

• Non-uniform illumination
• Lens vignetting
• Gain errors for individual pixels (PRNU - Photo Response Non-Uniformity)
• White balance for RGB sensors

The gain coefficients have to be calculated by the application. The gain values are configured using FG_set_special_option() with the option SetGainCoeff. The argument is a pointer to an array of float values containing the gain coefficients for each sensor pixel. The maximum supported gain value is 4.0 which should be enough for most situations.

The length and arrangement of the float array depends on the pixel format and AOI used:

Sensor type	eFG_PIXEL_TYPE	Number of values	Arrangement
Single line	FG_PIXEL_TYPE_Y_8 FG_PIXEL_TYPE_Y_16	AOI width	Gain values for each sensor pixel
Dual line	FG_PIXEL_TYPE_Y_8 FG_PIXEL_TYPE_Y_16	2x AOI width	Two lines of gain values in succession
	FG_PIXEL_TYPE_RGB_24 FG_PIXEL_TYPE_RGBX_32	3x AOI width	One line of R/G/B components for each pixel

Only gain coefficients for the configured AOI are given to the function. Gain values for sensor pixels outside of the AOI are automatically set to 1.0.

Recording a reference image

Before taking a reference image for calculation of the gain values, make sure that the following conditions are met:

• Reset gain coefficients to the default value (1.0):

  FG_set_special_option("SetGainCoeff", 0);

• Set digital RGB gain to the default value (255):

  FG_set_gain_rgb(255, 255, 255);

• Use a uniform target for capturing the reference image, e.g. a white sheet of paper.
• Adjust shutter and gain until middle to light grey brightness levels are reached, for example:

  FG_set_shutter_time(20);                // Shutter: 20 µs
  FG_set_special_option("CDS_Gain", 1);   // CDS gain: 1
  FG_set_gain(150);                       // ADC gain: 150 is the default setting

• Disable trigger and line counters (or ignore them during calculation of gain values):

  FG_set_special_option("InsertLineCounters", 0);

Example code for calculation and configuration of gain values

The example function below takes one acquired reference image to calculate optimal the gain values for each sensor pixel. The calculation of gain values for this example is based on the average pixel values for each column within the whole image. The code can be modified to use certain regions of the image only, for example a white horzontal region between printed lines of text. The function will limit gain values to the maximum allowed value of 4.0. This is the case for regions of the image which are too dark. int ShadingCorrection( FG_IMAGE *pFgImage, // pointer to reference image for calculating the shading coefficients bool save) // store coefficients permanently { eFG_PIXEL_TYPE pixelType; unsigned int x1, x2, y1, y2; unsigned int width, height; unsigned int channels, pixelSize; enum eFG_CAMERA_TYPE camera_type; bool dualLine; // Get AOI and pixel format FG_get_aoi(&x1, &y1, &x2, &y2, &pixelType); if (pixelType == FG_PIXEL_TYPE_Y_8 || pixelType == FG_PIXEL_TYPE_Y_16) channels = pixelSize = 1; else if (pixelType == FG_PIXEL_TYPE_RGB_24) channels = pixelSize = 3; else if (pixelType == FG_PIXEL_TYPE_RGBX_32) { channels = 3; pixelSize = 4; } else return -1; width = x2 - x1 + 1; height = y2 - y1 + 1; // Dual line sensor: we need to calculate column values for each line separately. // We handle the two image lines as one line with twice the width: FG_get_camera_type(&camera_type); dualLine = (camera_type == FG_CAMERA_TYPE_2048_2_IMAGO_VCx_DR2X2K7 || camera_type == FG_CAMERA_TYPE_2048_2_IMAGO_VCx_DR2X2K7_RGB || camera_type == FG_CAMERA_TYPE_4096_2_IMAGO_VCx_DR2X4K7 || camera_type == FG_CAMERA_TYPE_4096_2_IMAGO_VCx_DR2X4K7_RGB); if (dualLine && pixelType == FG_PIXEL_TYPE_Y_8) { width *= 2; height /= 2; } std::vector<float> gainCoeff(channels * width, 1.0f); // Resulting array of gain coefficients, default: 1.0 std::vector<unsigned int> colSum(channels * width, 0); // Sum for each column std::vector<double> colAvg(channels * width); // Average value for each column double maxVal = 0.0; // Highest average value // Calculate pixel sum for each column for (unsigned int y = 0; y < height; y++) { for (unsigned int x = 0; x < width; x++) { for (unsigned int color=0; color < channels; color++) { unsigned int pixel; if (pixelType == FG_PIXEL_TYPE_Y_16) { unsigned short *pSrc16 = (unsigned short *)pFgImage->pixel_ptr; pixel = pSrc16[pixelSize*(x + y * width) + color]; } else { unsigned char *pSrc8 = pFgImage->pixel_ptr; pixel = pSrc8[pixelSize*(x + y * width) + color]; } colSum[channels*x + color] += pixel; } } } // Calculate average column value and remember the highest average for all columns for (unsigned int x = 0; x < (channels*width); x++) { colAvg[x] = (double)colSum[x] / (double)height; if (colAvg[x] > maxVal) maxVal = colAvg[x]; } // Calculate gain coefficients for each column. // The column with the highest average gets a gain of 1.0. // The gain for darker columns is increased to reach the same brightness level. for (unsigned int x = 0; x < (channels*width); x++) { if (colAvg[x] > 0.0) { float coeff = maxVal / colAvg[x]; if (coeff > 4.0f) // limit gain value coeff = 4.0f; gainCoeff[x] = coeff; } } // Set coefficients FG_set_special_option("SetGainCoeff", (INT64)&gainCoeff[0]); // Save the new coefficients permanently if (save) FG_set_special_option("SaveGainCoeff", 0); return 0; }

Triggering the sensor

In hardware triggered mode, each trigger event results in one or two valid image lines (depending on sensor type). Multiple trigger events are required before the image is complete and returned to the user.

The RTCC can be used to adapt the the trigger signal to the desired scan rate, see Hardware trigger for examples.

Note

Software triggered mode is not supported.

Embedded line and trigger counters

After enabling the special feature InsertLineCounters, two counter values will be embedded at the beginning of each scan line (every other line for dual line sensors without demosaicing active):

• A 16 bit value counting the number of valid sensor scan cycles.
• A 16 bit value counting the number of trigger events received in hardware triggered mode. In free run mode, the counter increases according to the actual scan rate.

The first counter can be used for detection of dropped senor lines which are caused by insufficient acquisition buffers.

The second counter is used in hardware triggered mode for detection of ignored trigger events when the trigger signal is arriving too fast. The trigger event is properly synchronized with the associated lines received from the sensor.

The storage of the counter values depends on the used pixel format, see the example code below for reading the values.

Example code for reading line and trigger counters

int main() { FG_IMAGE imageList[NUM_BUFFERS]; UINT32 x1, y1, x2, y2; UINT32 scanLines; enum eFG_PIXEL_TYPE pixel_type; int checkCounters = 0; unsigned short lineCounter[2]; unsigned short trgCounter[2]; // install and configure the camera SetupCamera(); // configure customized hardware trigger SetupTrigger(); // get the image size for comparing the received counter values later: FG_get_aoi(&x1, &y1, &x2, &y2, &pixel_type); // the expected number of counter values is equal to the image height: scanLines = y2 - y1 + 1; // uncomment the following line when using dual line sensors without demosaicing active: // scanLines /= 2; // activate counters FG_set_special_option("InsertLineCounters", 1); // allocate image buffers for (UINT32 i = 0; i < NUM_BUFFERS; i++) FG_alloc_image(&imageList[i]); // start acquisition for (UINT32 i = 0; i < NUM_BUFFERS; i++) FG_append_image(&imageList[i]); // enter acquisition loop while (isRunning) { FG_IMAGE currentImage; unsigned short counterDiff; UINT32 res = FG_get_image(&currentImage, UINT_MAX); if (res == FG_ERROR_CODE_NoError) { // read the line and trigger counters for the first line in the image, depending on the pixel format: if (pixel_type == FG_PIXEL_TYPE_RGBX_32) { lineCounter[0] = ((unsigned short *)currentImage.pixel_ptr)[0]; trgCounter[0] = ((unsigned char *)currentImage.pixel_ptr)[2] + 256 * ((unsigned char *)currentImage.pixel_ptr)[4]; } else if (pixel_type == FG_PIXEL_TYPE_Y_16) { lineCounter[0] = (((unsigned short *)currentImage.pixel_ptr)[0] >> 2) + 256 * (((unsigned short *)currentImage.pixel_ptr)[1] >> 2); trgCounter[0] = (((unsigned short *)currentImage.pixel_ptr)[2] >> 2) + 256 * (((unsigned short *)currentImage.pixel_ptr)[3] >> 2); } else { lineCounter[0] = ((unsigned short *)currentImage.pixel_ptr)[0]; trgCounter[0] = ((unsigned short *)currentImage.pixel_ptr)[1]; } if (checkCounters) { // check the received sensor lines counterDiff = lineCounter[0] - lineCounter[1]; if (counterDiff != scanLines) printf("%d sensor lines dropped\n", counterDiff - scanLines); // check the trigger counter counterDiff = trgCounter[0] - trgCounter[1]; if (counterDiff != scanLines) printf("%d trigger events dropped\n", counterDiff - scanLines); } else { // delayed start of counter check checkCounters = 1; } // store values for the next cycle lineCounter[1] = lineCounter[0]; trgCounter[1] = trgCounter[0]; FG_append_image(&currentImage); } else if (res == FG_ERROR_CODE_BrokenImage) { FG_append_image(&currentImage); } else // Error during waiting return -1; } // abort image acquisition FG_stop_image(); // free buffers for (UINT32 i = 0; i < NUM_BUFFERS; i++) FG_free_image(&imageList[i]); // Close the camera FG_uninstall_camera(); return 0; } Please note that no error checking is performed in the examples in order to simplify the code.

Note

The alternative special features InsertImageCounter, InsertTriggerCounter have a similar meaning with some differences:

• The placement of the counter values within the image can be different, see example code.
• With InsertTriggerCounter, the trigger event is not synchronized with the sensor lines and therefore is not accurate.
• InsertLineCounters can be activated during acquisition.
• In free run mode, InsertLineCounters increments the counter with the actual scan rate.

InsertImageCounter, InsertTriggerCounter and InsertTimeStamp have priority and will overwrite counter values generated by InsertLineCounters.