AI and NIR Imaging in Modern Fruit Sorting Lines

Post-harvest fruit packing lines use near-infrared (NIR) hyperspectral imaging and artificial intelligence (AI) to non-destructively assess the internal quality of apples and avocados. By combining advanced light-sensing technology with machine learning algorithms, packing facilities can look beneath the skin of the fruit at high speeds. This process allows sorters to accurately detect hidden defects like internal browning and measure precise sugar levels, ensuring that only high-quality produce reaches consumers while significantly reducing food waste.

The Role of Near-Infrared Hyperspectral Imaging

Traditional cameras only capture visible light, which limits them to detecting surface-level defects on fruit. Near-infrared hyperspectral imaging goes a step further by projecting near-infrared light onto the apples and avocados as they move along the conveyor belt. This light penetrates the skin and interacts with the internal cellular structure and chemical bonds of the fruit.

As the light bounces back, specialized hyperspectral sensors capture hundreds of contiguous wavelength bands across the electromagnetic spectrum for every single pixel in the image. This creates a three-dimensional dataset known as a "hypercube," which contains both spatial information (the shape and size of the fruit) and spectral information (the internal chemical composition).

Measuring Sugar Content (Brix)

The sweetness of fruit, measured in degrees Brix, is determined by its soluble solids content, primarily sugars. Sugar molecules contain specific hydrogenbonds (such as C-H and O-H) that absorb near-infrared light at distinct wavelengths.

When the NIR light passes through an apple, the sensors measure the exact amount of light absorbed at those sugar-specific wavelengths. The more light absorbed in these bands, the higher the sugar concentration. This allows the system to map the sweetness distribution inside the fruit without cutting it open.

Detecting Internal Browning

Internal browning in apples and avocados is caused by cell membrane breakdown, oxidation, and chilling injuries. While the outside of the fruit looks completely healthy, the inside turns brown and spongy.

Healthy fruit tissue has intact cells filled with water and air spaces that scatter light predictably. When internal browning occurs, the cellular structure collapses, changing how the tissue scatters and absorbs NIR light. Hyperspectral imaging detects these subtle shifts in light scattering and water absorption bands, pinpointing the exact location and severity of internal rot or browning.

How Artificial Intelligence Processes the Data

Hyperspectral cameras generate massive amounts of complex data in milliseconds. To make this data useful on a fast-moving packing line, artificial intelligence and machine learning models—specifically Convolutional Neural Networks (CNNs) and Partial Least Squares Regression (PLSR)—are deployed.

The AI is first trained on a vast dataset of scanned fruits that have subsequently been cut open and lab-tested for actual sugar content and browning. Through this training, the AI learns to correlate specific spectral patterns with real-world internal traits.

During live operations, the AI performs the following steps in real time:

• Noise Reduction: It filters out interference caused by the fruit's varying skin thickness, surface moisture, or ambient temperatures.
• Feature Extraction: It isolates the exact wavelengths needed to evaluate sugar and tissue breakdown, ignoring irrelevant data to speed up processing.
• Classification: It assigns a quality grade to each individual apple or avocado based on pre-set commercial standards.

High-Speed Automated Sorting

The integration of NIR imaging and AI happens at speeds of up to 10 to 15 fruits per second per lane. As an apple or avocado rolls under the inspection hood, mirrors rotate the fruit so the cameras can capture a 360-degree internal view.

Within milliseconds of the scan, the AI processes the hypercube data and makes a decision. It sends a signal to mechanical ejectors further down the line. As the fruit reaches the sorting cups, pneumatic air jets or gentle mechanical arms divert it into specific chutes. Fruits with high sugar content and no browning are sent to premium retail packaging, slightly riper fruits are routed to local markets for quick sale, and internally browned fruits are diverted to processing plants or compost.