The key to the bean color sorter’s 99.9% recognition accuracy and 3-15 tons of processing capacity per hour lies in its highly efficient and coordinated automated sorting system, encompassing four key steps: feeding and mixing → image acquisition → intelligent analysis → precise rejection. Each step relies on specific hardware components and technical algorithms to precisely eliminate off-color particles and impurities from the beans.
1. Core Components: The Hardware Foundation of a Color Sorter
Before understanding the workflow, it’s important to first understand the four core hardware components that support color sorting. These components are like the “limbs” and “brain” of a color sorter, each essential:
Feeding and Distributing System: This system consists of a hopper, vibrating feeder, and a hopper. Its core function is to ensure that beans enter the sorting area evenly, stably, and orderly, preventing bean accumulation and congestion that could affect identification accuracy.
Optical Detection System: The “eyes” of the color sorter, consisting of a high-resolution CCD image sensor (charge-coupled device), a low-power LED light source, and an optical lens.
AI Intelligent Analysis System: The “brain” of the color sorter, consisting of an industrial-grade main control chip and proprietary AI recognition algorithms. The chip receives image data transmitted by the optical system. The algorithm, through deep learning from a vast array of bean samples (including normal beans, off-color beans, insect-damaged beans, stones, and grass seeds), builds a characteristic model for “qualified beans” and “rejected beans/impurities.” It instantly compares the differences between the currently captured bean image and the model to determine whether it meets the sorting criteria.
The rejection execution system, the “hand” of the color sorter, primarily consists of high-frequency solenoid valves and air nozzles. The solenoid valves have a response speed of up to 0.02 seconds. The nozzles are arranged in an array below the sorting area, each corresponding to a material flow channel. Upon receiving a “reject command,” they instantly release high-pressure air to blow rejected beans or impurities away from the normal material flow path.
2. Workflow: “Precise Sorting” in Four Steps
Once beans enter the optical sorter, they are sorted in a rigorous four-step process. The entire process is seamless and efficient, with sorting a single bean taking only approximately 0.05 seconds:
Step 1: Feeding and Smoothing to Create an Observable Flow
The operator pours the beans to be sorted into the top hopper, where they enter the vibrating feeder under gravity.
The vibrating feeder vibrates at a high frequency of 50-100Hz, evenly conveying the beans to the sorting trough. The slits in the trough’s discharge port “sort” the beans into a single, thin stream, preventing overlapping and obstruction.
The thin stream descends along the chute at a steady speed (approximately 1.5-2m/s) into the optical inspection area, preparing for subsequent image capture. Step 2: Optical Capture to Capture the Bean’s “Subtle Features”
When beans enter the optical inspection area, LED light sources on both sides are activated simultaneously. Sidelighting highlights the bean’s shape and outline (such as the irregular edges of a damaged bean), while backlighting enhances color differences (such as the contrast between the tones of green beans and normal yellow beans).
A high-resolution CCD image sensor (with a resolution of up to 2048 pixels) captures each bean from multiple angles as it falls, converting its optical information (RGB color values, shape and outline, and surface gloss) into digital signals, which are then transmitted in real time to the AI intelligent analysis system. Step 3: Intelligent Analysis to Determine “Pass or Fail”
Upon receiving the digital signal, the AI intelligent analysis system immediately invokes a pre-set “bean sorting model” (different bean types, such as soybeans and red beans, have different model parameters, which can be switched with a single click on the touchscreen).
The system compares the optical characteristics of the current bean with those of “qualified beans” in the model. Beans are classified as “qualified” if their RGB color values are within the normal range, their shape is regular, and they lack impurity characteristics (such as the high reflectivity of stones or the unusual shape of grass seeds). Beans with discoloration, mold, damage, or impurity characteristics are identified as “to be rejected.”
After the analysis is complete, the system sends instructions to the rejection execution system: if the bean is “qualified,” no action is triggered. If it is “to be rejected,” the bean’s falling trajectory is precisely calculated to determine the corresponding nozzle position and jet timing.
Step 4: Precise Rejection, Separating Good and Defective Beans
When the “to-be-rejected” beans continue to fall toward the rejection zone, the system controls the high-frequency solenoid valve at the corresponding position to instantly activate, driving the air nozzle to spray out a high-pressure airflow of 0.3-0.5 MPa.
In a very short time (0.02 seconds), the high-pressure airflow changes the trajectory of the “to-be-rejected” beans, causing them to fall into the “defective collection bin” on the side.
The “good beans,” unaffected by the airflow, continue their original trajectory and enter the “finished product collection bin.”
The entire rejection process is contactless, preventing damage to the beans caused by collisions. The high-frequency response of the solenoid valve (over 500 times per second) ensures precise rejection, ensuring that both good beans are rejected and defective beans are not missed.
Through the collaborative system of “hardware + algorithm + process”, the bean color sorter has broken away from the limitations of manual sorting and achieved the core value of “accurate identification, efficient processing, and low loss”.
Post time: Sep-11-2025
