Cost-Effective Urine Sediment Examination with CNN-Based Vision System
What is the CNN-based embedded vision system for urine sediment analysis all about?
In a urine sediment analysis, laboratory employees examine the solid components of a urine sample under microscope to diagnose diseases of the kidney or urinary organs. However, manual examinations under microscope generally carry a risk of errors and take up valuable time. Many laboratories therefore use automated systems. These differ according to the required performance, functionality, and design size, which also means they vary in terms of their integrated vision system. The CNN-based embedded vision system represented below meets the requirements of cost-effectiveness and high-performance.
What are the challenges in urine sediment analysis?
Lab analyses increasingly have to be provided quickly and inexpensively, of course without compromising on reliability. This is possible with a CNN-based analysis algorithm on the right embedded vision system. The decisive factor is the integration of the right hardware and software components—namely those that are coordinated with each other. This is best achieved with an experienced partner.
The solution: a reliable and affordable CNN-based vision system
The demo offers a reliable and inexpensive CNN-based vision system suitable for tasks such as the above-mentioned automated microscopic examination of urine sediment. We carry all required hardware and software components in our extensive portfolio.
The system hardware is composed of a dart color camera module with 5 MP resolution and MIPI CSI2 interface, a compatible data cable, and the Embedded Vision Processing Board developed by Basler. The board provides relevant communication and control interfaces that make it suitable for integration as a central control unit in small benchtop devices. The core element of our price/performance-optimized board is the i.MX 8M Plus from NXP®.
The system software includes the pylon Camera Software Suite, which is used for the camera configuration and subsequent image capture. The dart camera module used has an ISP (Image Signal Processor) that completes all the image pre-processing steps. These include, for example, debayering and denoising as well as setting an ROI (Region of Interest) or scaling the sensor pixels to the target resolution. Since not all dart camera modules have an integrated ISP, the integrated ISP of the i.MX 8M Plus in the Embedded Vision Processing Board can be used as an alternative. The CNN’s subsequent inference is performed on a special processor of the i.MX 8M Plus SoC (System on Chip), known as the NPU (Neural Processing Unit). Using hardware acceleration, the NPU performs the typically intensive processing tasks for the CNN, which makes the execution particularly efficient. Following the inference by the NPU, additional processing steps take place on the CPU (Central Processing Unit) of the i.MX 8M Plus SoC, such as the visualization of bounding boxes and the creation of output statistics about the classified urine sediment components.
The benefits of a CNN-based embedded vision system in urine sediment analysis
Fully self-sufficient embedded vision processing board—from a single source
Board suitable for prototyping as well as series production of your complete system
Expertly produced vision technology with long-term availability and price/performance optimization
Seamless integration of classic and CNN-based image processing and analysis