Final Activity Report Summary - INTARWELD (Intelligent Augmented Reality Welding Helmet)
The aim of the project was to build personal protection equipment (PPE) system for the manual welding application. The darkening of the welders view shall cover only the places where glaring light sources respectively the welding arc and its reflections occur. The idea is to extract the glares by a video camera system and using that information to darken pixels on a see-through graphical LCD (GLCD). This idea needs a video acquisition system for welding, a calibration procedure to calculate the transformation between the camera and user view and an adequate darkening by a GLCD. This project mainly elaborated a system for the video observation of high dynamic contrast scenes with the application of welding.
The development of a SADF system has been achieved as a feasibility study due to non-existent adequate hardware. The insufficient darkening of the prototype of a 2D GLCD specialised for welding gave the opportunity to build a mock-up system, to proof the concept. The lack of a defined helmet fixation to remain a valid calibration needs an alternative approach in order to avoid tedious calibrations. The problems of having on the one side a camera for high dynamic contrast but only eight bits to record the greyscale is a drawback which bounds the quality of the raw image. One step to overcome with such limitations is to use the toggle merging with two images containing information of two different brightness ranges.
It ended up in an image with more than 256 greyscale values, which has to be aligned to eight bit information for displaying on a monitor. This high dynamic range increase produces the full range of the available 256 steps of greyscale for the next algorithmic step of contrast enhancement. After widening the input data range of the image, the second most important question arose related to enhancing the image information. The question of what is the information of the image and how can it be enhanced ended up at the VBSAHE algorithm after examining different contrast enhancement approaches.
This algorithm individually enhances the segmented information of the image. It takes the eight bit greyscale and enhances the local contrast, so that globally seen, the information becomes better visible as information by small greyscale differences can be enhanced without overrunning the limitations of 256 greyscale values of the output image. Related to the application of welding the problem of synchronising the camera onto the welding process gave a milestone in video quality for the human. The video sequence becomes stabilized by fixing a system immanent flickering white stripe to a fixed position without depending on any machine interface.
The optical trigger with automatic alignment of the sensor system to the brightness condition makes this system robust for a wide range of applications related to welding method, working distance and conditions. With the integration of the trigger, the flashing of the scene with high power LED became feasible. Although the effect of this technique is limited to medium welding currents, it produces a lighting of the areas which are lying in the shadow of the glaring welding arc.
Side effects of this research were the examination of the morphological top-hat operator and the application of the Wiener filter for de-noising the image. In the future the aspects of a SADF system need to be observed regarding available GLCDs which may darken sufficiently the bright spots and follow movements with an adequate frame rate; but such a development by the industry is not expected as the market does not need it, yet. An alternative suggestion is to re-think the idea of a video see-through welding helmet were the user sees the scene by a video see-through HMD while the scene is captured by a camera system; a camera system like the one researched here.
The development of a SADF system has been achieved as a feasibility study due to non-existent adequate hardware. The insufficient darkening of the prototype of a 2D GLCD specialised for welding gave the opportunity to build a mock-up system, to proof the concept. The lack of a defined helmet fixation to remain a valid calibration needs an alternative approach in order to avoid tedious calibrations. The problems of having on the one side a camera for high dynamic contrast but only eight bits to record the greyscale is a drawback which bounds the quality of the raw image. One step to overcome with such limitations is to use the toggle merging with two images containing information of two different brightness ranges.
It ended up in an image with more than 256 greyscale values, which has to be aligned to eight bit information for displaying on a monitor. This high dynamic range increase produces the full range of the available 256 steps of greyscale for the next algorithmic step of contrast enhancement. After widening the input data range of the image, the second most important question arose related to enhancing the image information. The question of what is the information of the image and how can it be enhanced ended up at the VBSAHE algorithm after examining different contrast enhancement approaches.
This algorithm individually enhances the segmented information of the image. It takes the eight bit greyscale and enhances the local contrast, so that globally seen, the information becomes better visible as information by small greyscale differences can be enhanced without overrunning the limitations of 256 greyscale values of the output image. Related to the application of welding the problem of synchronising the camera onto the welding process gave a milestone in video quality for the human. The video sequence becomes stabilized by fixing a system immanent flickering white stripe to a fixed position without depending on any machine interface.
The optical trigger with automatic alignment of the sensor system to the brightness condition makes this system robust for a wide range of applications related to welding method, working distance and conditions. With the integration of the trigger, the flashing of the scene with high power LED became feasible. Although the effect of this technique is limited to medium welding currents, it produces a lighting of the areas which are lying in the shadow of the glaring welding arc.
Side effects of this research were the examination of the morphological top-hat operator and the application of the Wiener filter for de-noising the image. In the future the aspects of a SADF system need to be observed regarding available GLCDs which may darken sufficiently the bright spots and follow movements with an adequate frame rate; but such a development by the industry is not expected as the market does not need it, yet. An alternative suggestion is to re-think the idea of a video see-through welding helmet were the user sees the scene by a video see-through HMD while the scene is captured by a camera system; a camera system like the one researched here.