Inspection of Construction Aggregate Particles with NI LabVIEW and NI Vision
Print
LabVIEW and iMAQ Vision are used to extract accurate 3D morphological data from aggregate samples.
Author(s):
Melvyn Smith - University of the West of England, Franchay Campus
Thorgeir Helgason - Petromodel Ltd
Industry:
Basic Materials - Steel/ Lumber/ Construction
Products:
LabVIEW, IMAQ Vision
The Challenge:
Developing a prototype benchtop system capable of optically acquiring accurate 3D data from gravel and rock particles to classify their morphological properties.
The Solution:
Using LabVIEW and IMAQ Vision tools to design a system for rapidly acquiring and analysing accurate 3D particle data.
"We developed code in LabVIEW to extract particle principal dimensions, including length, width, and height, and from this information we calculated the distribution of the different particle form categories. "
Aggregates such as sand, gravel, and crushed rock are used around the world in the construction of buildings, bridges, roads, railways, and other structures. Once quarried, samples are analysed and graded manually to determine their properties and suitability for differing applications. The 3D shape, size, and composition of aggregate materials are central to their performance in construction applications. Trained geologists conduct compositional analysis of aggregates in an expensive, time-consuming, and often subjective manual task. The development of robust, automated techniques that establish the properties of an aggregate sample offers several benefits, including better quality control through a more reliable, objective means of classification; business agility benefits from a reduction in the lead time required to determine the source of a particular aggregate; and cost-effectiveness and environmental benefits from more efficient use of aggregates in composites such as concrete and asphalt.
To operate in real time, speed was the major issue. The limiting factor in this case was camera frame rate. Standard analogue machine vision cameras acquire interlaced images at a maximum speed of 25 frames per second. We replaced these with JAI CV-A33 digital cameras, which together with a high-bandwidth Camera Link interface allow high frame rates at a similar resolution. We synchronised the cameras by using an NI RTSI bus to directly connect the two frame grabbers. This further enhanced performance because timing information was exchanged directly between the frame grabbers, thus avoiding the propagation delays and processing burdens associated with the exchange of such information through the computer’s PCI bus.
Hardware
We developed a benchtop device that uses laser triangulation (based on the Stoker-Yale 635 nm SNF series laser-line projector) to recover 3D data from aggregate particles as they passed along a moving conveyor. To meet the user specified requirement of analyzing at least 200 particles per hour, we combined high-speed digital cameras with a high-bandwidth interface to enable the acquisition of frame rates at up to 120 frames per second. The careful design of the system, incorporating the two registered cameras, minimised any data occlusion. A 10-bit encoder on the drive shaft of the aggregate transport conveyor was connected to an NI PCI-6601 counter/timer board in the PC. We synchronised the cameras and the encoder using an NI real-time bus, connecting the two frame grabbers to each other and to the counter/timer board. This precisely tailored the frame rate of the cameras to the speed of the belt, ensuring that each camera captured images every time the belt moved 0.1 mm. Timing information was exchanged directly between the frame grabbers, thus avoiding the propagation delays and processing burdens associated with the exchange of such information through the computer’s PCI bus. Now, a dense “cloud” of accurate 3D data points is recovered for each aggregate particle.
Connectivity
We connected the Camera Link cameras to the NI PCI-1428 image acquisition board, which in turn was connected to the PCI-6601 inside the system PC unit using the NI RTSI bus. We connected a shielded I/O connector block, the NI SCB-68, to the PCI-6601 using an NI SH6868 cable assembly. The particle conveyor encoder also connected through the SCB-68 connector block. This block provided 5 V and GND pins to power and ground the encoder. The encoder output was connected to the gate of the PCI-6601 counter/timer. The counter uses a synthetic source signal as a time base, so that it can be programmed in terms of periods of the source input. This way, output pulses are generated for every active gate signal edge. The VSync signal generated by Camera 1, the master camera, is the synchronization signal. The output signal from the counter passes directly to the bus, which triggers the cameras via the frame grabber. Hence, using this NI hardware configuration, we could bypass software and deliver accurate, high-speed triggering.
Algorithms and Software
We used novel techniques to classify the particle size and shape based on the acquired cloud of 3D data, which accurately represents the visible surfaces of the particles. To maximise the accuracy of the acquired geometric data, we used standard NI IMAQ Vision image smoothing and filtering operators combined with synchronized motion and acquisition.
We developed code in LabVIEW to extract particle principal dimensions, including length, width, and height, and from this information we calculated the distribution of the different particle form categories. Also, we developed an algorithm for measuring the angularity or roundness of the particles using a novel morphological opening operation with an ellipsoidal structuring element that is adaptively generated according to the size and aspect ratios of each individual particle. This effectively mimicked the natural wear process through which particles become rounded, thus providing a direct and geometrically meaningful measure of angularity. Additionally, we created a virtual sieving algorithm to provide a useful means of converting 3D particle sizes into the 1D particle sieve sizes widely used in industry.
Evaluation and Conclusions
Field trials show that typical frame rates of approximately 80 Hz could easily be achieved in practice, which, based on an average particle length of 16 mm, allowed the analysis of approximately 450 particles per hour at an accuracy of ±0.1 mm. Therefore, the system met the user throughput specification. The recovery of particle length, width, and height measures in themselves is of little use to aggregate producers and consumers. However, particle size distributions, recovered using the virtual sieve, are critical from an industrial perspective, and the ability to automatically recover sieve size without undertaking manual sieving offers significant commercial value. This first prototype, built at UWE in Bristol, UK, as part of a EUREKA project, has now been upgraded and supplied with an automatic particle feeder.
Related Case Studies
NI Vision Tools Meet the Challenges of Real-Time InspectionSylvania Lighting Develops Flexible Machine Control System Using NI Vision, Motion, Intelligent DAQ, and LabVIEW
NI LabVIEW and Vision Products Monitor Pharmaceutical Processes
Diagnostic Sonar Acquires and Processes Ultrasonic Phased Array Image Data with NI LabVIEW and PXI
Industrial Machine Vision Based on National Instruments Technology Goes Further in Complex Human Task Automation
|
|