Project

Motion Detection Program

Multithreaded video motion detection with client-server distributed processing

Overview

A multi-functional application that processes video files to detect motion using frame differencing. Features video-to-frame extraction, multithreaded processing, and distributed server-client architecture for parallel processing.

Highlights

Multithreaded motion detection using pthreads for efficient frame processing across CPU cores
Client-server TCP socket communication for distributed motion detection workloads
Frame differencing algorithm with RGB-to-grayscale conversion and binary thresholding
FFmpeg integration for video-to-frame extraction and frame-to-video reconstruction
OpenCV-based frame extraction with customizable resolution and framerate

Tech Stack

CC++pthreadsTCP SocketsFFmpegOpenCVlibjpegImage Processing

User Interface

The motion detection program features a menu-driven interface providing six core options: video-to-frame conversion, motion detection processing, frame-to-video reconstruction, server mode for distributed processing, client mode for connecting to servers, and exit. The interface includes comprehensive input validation, automatic directory creation, and 'home' keyword support for returning to the main menu at any prompt. Users can specify custom framerates and resolutions for output videos, making the system flexible for various use cases from security surveillance to scientific analysis.

Terminal-based menu interface for motion detection operations

Original Video Input

The program accepts standard video formats including MP4, AVI, and MOV through FFmpeg and OpenCV integration. This example shows a nighttime outdoor scene with moving subjects. The video is first decomposed into individual JPEG frames using OpenCV's VideoCapture class, with each frame numbered sequentially (frame_0.jpg, frame_1.jpg, etc.). Frame extraction preserves original resolution and color information, enabling accurate motion detection in subsequent processing stages.

Original night-time video before motion detection processing

Motion Detection Algorithm

Each frame undergoes a six-step processing pipeline: (1) Load current frame as RGB image using libjpeg, (2) Convert to grayscale using weighted average (R×0.3 + G×0.59 + B×0.11) to reduce dimensionality, (3) Load previous frame and convert to grayscale, (4) Compute absolute pixel-wise difference between frames to generate motion map, (5) Apply binary threshold (default 20) to isolate significant motion areas, (6) Save motion-detected frame as JPEG. This approach highlights regions with pixel intensity changes exceeding the threshold, effectively isolating moving objects while suppressing static background elements.

Motion Detection Output

The processed video shows motion-detected regions highlighted in white against a black background. This binary representation clearly identifies moving objects (people walking) while eliminating static elements (buildings, trees, ground). The frame differencing algorithm successfully detects subtle movements even in low-light conditions. Each output frame is saved as motion_frame_N.jpg and can be reconstructed into a video using FFmpeg with user-specified framerate and resolution settings.

Motion-detected output showing only moving objects in white

System Architecture

Menu-driven interface (main.c) for video-to-frame conversion, motion detection, and frame-to-video conversion
Multithreaded frame processing (handle_motion.c) using pthreads to distribute workload across CPU cores
Image utilities (image_utils.c) for JPEG loading/saving, RGB-to-grayscale conversion, frame differencing, and thresholding
Network utilities (network_utils.c) implementing TCP server-client communication for distributed processing
OpenCV integration (vid_to_jpg.cpp) for extracting frames from video files
FFmpeg integration for reconstructing processed frames into output video with customizable resolution and framerate

Multithreading Implementation

Dynamic thread allocation based on CPU core count using sysconf(_SC_NPROCESSORS_ONLN)
Frame batching: total frames divided evenly among threads for parallel processing
Each thread processes assigned frame range independently using process_frame_batch()
Thread synchronization using pthread_join() to ensure all frames complete before proceeding
Prevents race conditions through independent frame file I/O per thread
Significantly reduces processing time for large video datasets

Distributed Processing Architecture

The server mode splits the workload in half: server processes first half of frames locally while sending second half instructions to connected client via TCP socket (port 8080). The client receives processing command with input/output paths and frame range, processes assigned frames using multithreading, and sends completion confirmation back to server. This distributed approach doubles processing throughput when two machines are available. TCP sockets use AF_INET for IPv4 communication with SO_REUSEADDR to prevent address-in-use errors during testing.

Video Processing Pipeline

Video-to-Frame: OpenCV VideoCapture extracts frames and saves as sequential JPEG files (frame_0.jpg, frame_1.jpg, ...)
Motion Detection: Multithreaded processing compares consecutive frames and generates motion_frame_N.jpg outputs
Frame-to-Video: FFmpeg command reconstructs frames into MP4 using libx264 codec with customizable framerate and resolution
User specifies framerate (e.g., 30 fps) and resolution (e.g., 1280x720) for output video
Supports return-to-menu functionality via 'home' keyword during all input prompts
Automatic directory creation and validation for output paths

Key Learnings

Multithreading with pthreads significantly improves performance for CPU-bound image processing tasks
TCP socket programming enables distributed workload processing across networked machines
Frame differencing provides effective motion detection without complex computer vision algorithms
Proper memory management (malloc/free) is critical to prevent leaks in long-running image processing loops
Integration of C and C++ code allows leveraging both low-level performance and high-level libraries like OpenCV