Automated organism detection for wastewater treatment.

PATENT

A patented system for capturing microscopy images of activated sludge, classifying organisms and floc characteristics using deep learning, and generating corrective action recommendations in real time.

End-to-end pipeline from basin to corrective action.

FIG. 1 - SYSTEM ARCHITECTURE

An operator collects a sludge sample and places it on a slide, which is then imaged through a microscope and sensor. That data routes through the cloud to a five-component processor stack: Preprocessing (30), Training (32), Attribute Recognition (34), Density Recognition (36), and Correction (38).

A prediction database stores all training data and deployed models. Results are delivered to a user interface device, giving operators actionable output without manual microscopy or lab turnaround.

FIG. 10 shows interfloc bridging at 100× magnification — filamentous organisms growing between and connecting individual floc particles.

This is a key early indicator of filamentous bulking, where the sludge mass knits together, impairing settling performance in the clarifier before the problem becomes severe.

Filaments connecting floc particles.

FIG. 10 - INTERFLOC BRIDGING

Normal floc vs. Zooglea.

FIG. 11 - ELEVATED POLYSACCHARIDE

FIG. 11 contrasts normal stain penetration (healthy floc, left) with lack of stain penetration (right), which indicates elevated polysaccharide - a condition associated with Zooglea-dominant sludge.

Elevated polysaccharide is typically caused by high organic acid loading or low dissolved oxygen, and results in a viscous, poorly settling sludge.

FIG. 12 shows a real sample with lack of stain penetration, indicating elevated polysaccharide within the floc matrix itself.

The system detects this characteristic automatically from the image - without requiring an operator to manually interpret staining results or compare against reference slides.

Detecting polysaccharide in the floc matrix.

FIG. 12 - STAIN PENETRATION ANALYSIS

The system quantifies filamentous organism density across six levels: Few (13A), Some (13B), Common (13C), Very Common (13D), Abundant (13E), and Excessive (13F).

Density grading determines the urgency and specificity of the corrective recommendation. A rating of Abundant or Excessive triggers immediate operational intervention guidance.

A six-level grading scale.

FIG. 13 - FILAMENT DENSITY

Organisms identified and labeled in real time.

FIG. 14 - LIVE OBJECT DETECTION

FIG. 14 shows live detection output on an actual sludge sample. The system draws bounding boxes around detected organisms, classifying each one by type and filtering out irrelevant objects in the same frame.

Here, Nocardia is identified across multiple locations within a single microscopy frame. Each detection feeds directly into the root cause and corrective action determination.

US 11,565,946 B2

Application No. 16/701,725

Filed: December 3, 2019

Granted: January 31, 2023

Expires: December 3, 2039

Status: Active

PATENT DETAILS

Edward Bryan Arndt

Francis John DeOrio

Patrick Joseph Campbell

INVENTORS

Systems & Methods for Treating Wastewater

Computer Vision · Deep Learning

Microscopy Image Analysis

Real-Time Process Control

CLASSIFICATION