In the agricultural sector, the adoption of automation has been on the rise, with an increasing number of companies investing in industrial cameras and smart software (AI) for optical quality control and sorting of vegetables, fruits, and plants. The company GeT Cameras, soon to be VA Imaging, offers industrial cameras that serve as visualization tools for machines. These cameras essentially function as the “eyes” of a robot or computer. By replacing human actions with an optical vision system, machine vision, enhances the speed and reliability of operations. This ensures continuous and consistent quality.

The majority of quality control tasks in this field rely on 2D area scan camera technology, though other advanced technologies like 3D cameras, line scan cameras, and hyperspectral imaging are becoming increasingly prominent. This article takes you in 5 steps through the selection process of the most suitable camera, lens, lighting and software.

Camera selection

Selecting the appropriate camera is crucial for any vision task. Key factors to consider include:

• Interface Type (USB3, GigE, or 10GigE)
• Colour or Monochrome Camera
• Global Shutter or Rolling Shutter Technology
• Resolution (number of pixels in the sensor)

Interface type (USB3, GigE or 10GigE)

Machine vision cameras require a connection to a computer, and the camera interface forms this connection. While engineers often have a preferred interface, it’s important to consider the distance between the camera and the computer. If this distance is less than 4.6 meters, a USB3 camera is recommended. For greater distances, a GigE camera is preferred. GeT Cameras often recommends cameras with Power over Ethernet (PoE), which provides both power and data communication through a standard Ethernet cable, simplifying the installation process.

Colour or monochrome camera

Monochrome cameras are frequently used in machine vision applications where tasks such as counting products, checking the presence of objects, or making measurements require good contrast. Monochrome sensors are up to three times more light-sensitive and produce sharper images compared to colour sensors.

However, when colour information is necessary, such as in inspecting vegetables, fruits, and plants, a colour camera (RGB) is essential. For instance, detecting different colour spots or defects on produce, which can vary in colour, requires an RGB camera. Additionally, colour cameras are a necessity for deep learning software since they provide extra information from colour images.

In scenarios where optical quality control is required, and defects might have varying colours, a colour camera is the appropriate choice.

Global Shutter or Rolling Shutter camera

When capturing images of moving objects, a Global Shutter camera is recommended. This type of camera reads all lines or pixels simultaneously, preventing image distortion. If both the camera and the object are stationary, a Rolling Shutter camera can be used. However, if the object is in motion, a Rolling Shutter camera may produce distorted images due to its line-by-line readout process.

In cases where fruits and vegetables are on a continuously moving conveyor belt during quality control and sorting, a Global Shutter camera is necessary to avoid image distortion. For more information, we advise you to read the following article, Rolling Shutter vs Global Shutter.

Resolution (number of pixels sensor)

Determining the correct resolution for a camera involves considering the smallest detail to be inspected and the area (Field of View) that needs to be covered. GeT Cameras typically recommends three pixels per smallest detail for a stable vision system, though two pixels can suffice depending on the software’s capability.

For example, if defects on fruits and vegetables need to be detected with 1mm accuracy on an 80-centimeter-wide conveyor belt, a camera with a resolution of 2400 x 1800 pixels is suitable. The MER2-503-23GC-P (IMX264) 5MP camera, which offers 2448 x 2048 pixels, meets these requirements.

Lens selection for IMX264

Choosing the correct lens is also essential for completing the computer vision system. Most machine vision cameras from GeT Cameras use a c-mount, requiring a c-mount lens. Factors such as the field of view, working distance (distance from camera/lens to the object), and sensor size of the selected camera must be considered.

For instance, with a horizontal field of view of 800mm and a working distance of 735mm, a focal length of 8mm is required, making the LCM-5MP-08MM-F1.4-1.5-ND1 lens suitable based on these calculations. This is seen in the lens calculator. Based on this calculation and the camera specifications, the LCM-5MP-08MM-F1.4-1.5-ND1 lens is suitable.

Lighting for inspecting fruits, vegetables and plants

Selecting the right industrial machine vision light can be challenging. GeT Cameras offers expert support to help you choose the most suitable machine vision lighting solution for your specific requirements. With a team of experienced professionals, GeT Cameras assists in designing, implementing, and troubleshooting your machine vision lighting system. For personalized guidance, feel free to reach out through this link and the best solution will be found for you.

For this application, proper lighting is crucial for accurate inspection of this quality control system for fruits, vegetables, plants and more. Two bar lights are often used to illuminate produce on a conveyor belt, which is typically enclosed to eliminate ambient light and create a diffuse lighting environment. The lights are positioned transversely above the belt, covering its entire width and illuminating the objects from both sides to minimize reflection and shadows. The bar lights can be placed in the preferred angle.

Software for quality control and sorting of vegetables, fruits and plants

In addition to the hardware, computer vision software is required for automatic defect recognition. Customers can either write their own software or use existing vision software licenses, such as Zebra Aurora Vision Studio. Aurora Vision is a powerful machine vision software that allows users to design vision programs easily through a graphical interface, without needing programming knowledge. For more complex tasks, a deep learning add-on is available to tackle intricate detection and recognition challenges.

Aurora Vision Lite, a free demo version, allows users to design their own vision programs using all the tools available in the studio version. GeT Cameras also offers the service of creating sample programs based on customer-provided images and specifications.

Machine vision applications

These guidelines should assist in selecting the right hardware and software for computer vision applications in quality control and sorting of vegetables, fruits, and plants.

For further inquiries or technical support, GeT Cameras’ extensive experience and knowledge in the machine vision industry are always available to assist customers.