Sherlock8 can help you develop the best vision for visual inspection

With the rapid development of artificial intelligence (AI) technology, the machine vision industry is ushering in a new round of technological changes . As the main battlefield of artificial intelligence applications, machine vision is a key core technology for realizing industrial automation and intelligence, and an important technical support for promoting the development of Industry 4.0 and intelligent manufacturing.

Industrial production involves many processes and complicated procedures. From a small part to the final assembly into a finished product, and then to packaging and delivery, there is a crucial link, namely product quality inspection. Traditional manual inspection is often inefficient, has a high false positive rate, and is difficult to ensure consistency and stability. At the same time, it requires high experience and professional skills of quality inspectors.

With the development of visual technology and the deepening of application scenarios, machine vision has become the standard in the manufacturing industry. In specific applications, there are many companies that use traditional visual algorithms. Compared with manual inspection, traditional visual algorithms have obvious advantages in inspection efficiency and cost optimization, but it is difficult to make a qualitative leap in defect detection.

Faced with complex and ever-changing production links and a variety of inspection objects, many unexpected scenes often appear in the inspection link. For example, when defects are difficult to distinguish from the background; when the defects or flaws of bad products are only slightly or minutely different from those of good products; when the background texture is complex and produces noise interference to the detection algorithm, etc. These situations will touch the "knowledge blind spots" of traditional visual algorithms, thereby affecting its judgment and processing.

For large-scale and complex inspection and recognition tasks, letting the machine learn by itself is the best solution . Introducing deep learning algorithms into machine vision systems effectively solves the pain points of the above-mentioned traditional vision algorithms. Well-trained AI visual inspection systems can not only detect defects, but also have higher efficiency and accuracy, greatly improving the flexibility of the manufacturing industry and attracting more and more attention from the industry.

So, how does machine vision powered by AI achieve fast, efficient and accurate detection?

AI-based visual inspection relies primarily on two major strengths of artificial intelligence: computer vision and deep learning. AI can adapt to a variety of environments through deep learning, making it applicable to a wide range of industries. It has unlimited potential and can be developed quickly to meet the needs of manufacturers.

The reason why AI visual detection can be more efficient than the human eye is that the AI ​​"brain" stores a large amount of information, and the powerful computing power can quickly parse the captured available data. The system can classify objects in photos and videos and perform complex visual perception tasks such as searching images and captions, detecting objects, identifying and classifying.

For example, for visual inspection in the automotive industry, a deep learning-based algorithm needs to be developed specifically for the industry and trained with examples of defects it must detect until it has enough data to support its neural network to detect defects autonomously without any additional instructions.

From a technical perspective, the implementation of AI in industry requires not only deep learning and image processing technology, but also the support of supporting software algorithms to help AI inspection systems efficiently and accurately respond to the defect characteristics of different industries and products, shorten the development cycle of algorithm models, and quickly deploy them. It can be said that the software algorithm platform behind computer vision and deep learning is the core of the entire visual inspection system.

In terms of the integration of machine vision and software algorithms, Teledyne DALSA, a world-renowned visual solution provider, has rich industry know-how in the field of AI visual inspection, combining years of technical accumulation and implementation experience. Sherlock8 is the latest generation of AI (deep learning) visual inspection software launched by Teledyne DALSA. The high precision of AI inspection requires the supplement of rich traditional visual algorithms, and the powerful Sherlock8 visual platform provides perfect support for AI inspection.

Technical advantages

1

Sherlock8 has a flexible modular platform and multi-threaded parallel processing, supporting line scan, area array, 3D and infrared cameras. It has one of the few visual platforms with the richest visual algorithms in the industry, is compatible with almost all industrial communication standards, and supports VB, VC and C++.

2

Sherlock8 includes the Astrocyte AI Trainer, which has multiple deep learning neural network architectures for classification, anomaly, detection, and segmentation, and is suitable for detection in different scenarios.

1

Functional advantages

● Effective anomaly training can be performed with very few images;

● New images can be trained continuously directly in Sherlock8, and previously trained images do not need to be trained again, which greatly saves time;

● The segmentation framework has a wealth of drawing tools, which is simple and time-saving;

● Automatically generate hyper parameter configurations to facilitate inexperienced users;

● Detailed diagnostic reports to simplify problem finding;

● The flexible Sherlock8 platform enhances program flexibility and control, improves detection accuracy, and avoids the quagmire of over-training.

Application Cases

Combining multi-directional lighting with advanced software algorithms removes surface background effects such as noise or color and produces an image focused on the features most relevant for inspection. This image can then be inspected using standard Sherlock vision tools.

01

Extract raised characters from tire sidewall

02

Abstract stamped characters from metal gears

03

Extracting Braille markings from packaging surfaces

04

Extract date and batch characters from barcodes

As a rapid application development tool, Sherlock8 has helped vision systems be widely used in various industries such as 3C manufacturing, automobiles, logistics, printing and textiles, and new energy, and has comprehensively promoted the development of intelligent manufacturing to a higher level. In the future, Teledyne DALSA will continue to empower the industry with its advanced algorithm capabilities and integrated hardware and software solutions.